00

AGARD~raph No. 246

Survey of Methods to Assess Workload

BUS1 WC*MT 16 BJ.ST QtUALITV MIRCTDC~ABLE.
TWIE yfFMSaU TIP DDC X'N ~T.NY A /
17 t 9IMICOT OM~ER O~F PAGES MaQ-I1 DUO W4

fTh& 4xaixO Lew bwei appromvd

~for piabIc reklww cmd sala: Itm

*4$ThISJTICN AN A'#AUARROTY

L~.
CO ACK COYVAf

VJim
 Im1RqRp
S/ 'AGARD-AG-246

NORTH ATLANTIC TREATY ORGANIZATION

ADVISORY GROUP FOR AEROSPACE RESEARCH AND DEVELOPMENT

(ORGANISATION DU TRAITE DE L'ATLANTIQUE NORD)

SAGARDograph No. 246

VEY OF.NM.HODS TO SSSESS WORKLOAD)

USAF School of Aerospace Medicine

Brooks Air Force Base, Texas, 78235 "
USA \\

Ric.h•d E, McKenzie, Ph.D.

San Antonio, Texas, USA

ent has Abemro

This " Ap sdb P e
d
Mrpublc reoocd919t 0

This AGARi~ograph was sponsored by the AGARD Aerospace Me5dical Panel.

 SWPM

- 11HE MISSION OF AGARD

The minsion of AGARD is to bringi together the leading personalities of the NATO nations in the fields of science
and technology relating to aerospace for the following purposes:

- Exchanging of scientific and technical information;

-- Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence
posture;

- Imuproving the co-operation among member nations in aerospace research and development;

- Providing scientific and technical advice and assistance to the North Atlantic Military Committee in the tio'ld
of aerospace research and development;
Rendering scientifin and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field;

- Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

- Recommending effective ways for the member nations to use their research and development capabilities for
the common benefit of the NATO community.

The highest authority witnin AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are
composed of experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace
Applications Studies Programme. The results of AGARD work are reported to the member nations and the NATO
Authorities through the AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

The content of this publication has been ruproduced

directly from material supplied by AGARD or the authors.

Pubhished August 1979

Copyright © AGARD 1979
All Rilghtr Resented

"•SBN92-S35-.3.2-0-

Printedby Technical Editi'tg and Reproduction Ltd

Harford House, 7-9 Charlo,'teSt, London, WIP IHDI
 PREFACE

This multi-authored AGARDograph represents the preliminary survey of the Working Group (AMP-WG-)8).
"Evaluation of Methods to Assess Workload" was initiated by the AGARD Aerospace Madical Panel in January 1977,
following approval by the National Delegates Board (NDB) in the fall of 1976. Wcrking Group meetings were held at
Cologne (April 1977), London (October 1977), Fort Rucker, Alabama (May 1978), and Paris (November 1978),
concurrent with symposia conducted by the Aerospace Medical Panel. Early meetings focused, as would be expected,
on the scope of the tusk. While it was evident that the broad outline could be described with a high degree of a~reement,
it was also apparent that tasking individual members with sub-areas would require that they prepare manuscripts de novo,
a level of effort clearly not desired, given the substantial burden each of them alre-4y had in his own laboratory. It was
therefore decided at Fort Rucker to seek contributed chapters from Working Group members and others in the NATO
scientific community who had on hand materials which could be readily adapted to the objectives of the Working Group.
As the reade,' will see, numerous ,-.,atributions were received. The editors feel that the result is e wide-ranging
compendium of workload measurement methodology, though most certainly some methods have been either missed or
are under mrpresented.

The objectives and scope of the effort, as approved by the NDB, were as follows:

OBJECTIVES: Military aircraft are becoming increasingly complex, the associated avionics systems
more sophisticated, and the mismion profiles more demanding. The objective of the
Working Group is to study if such an increase in aircrew wvorkload has become a
limiting factor in the operationai employment of some aircraft and to determine
valuable methods to evaluate this workload.

SCOPE OF WORK: The measurement domain will be broken down into sensory threshold function
tests, motor funct;on, and responses to psycho, physio, and chemical excitation.

The methodology wii .nclude a wide range of instrumentation, laboratories,

inflight measurement and modelling methods, with the goal of compiling
systematically and evaluating the multiplicity of approaches and techniques
implied.

A companion document, AGARD Advisory Report 139 (AR-l 39) gives the conclusions drawn t y the Wor!'ing
Group within the bcunds of the above objectives and scope.

The members of the Working Group were:

PANEL MEMBERS NON-PANEL MEMBERS

R. Auffret. FR K.A. Kimball, US
E.P. Beck, UK M. Lees, US
B.O. Hartmait, US (Chairman) G.S. Malecki, US
A.N. Nicholson, UK R.D. O'Donnell, US
G. Rotondo, IT F.S. Pe'tyjohn, US
P. Woodward, US A. Roscoe, IJK
M.G. Sande!'s, US
H.M. Wegmann, GE

Numerous other members of the Aerospace Medical Panel attended Working Group meetings. because of the high
lcel of interest in this toi;.within the panel.

Accossion For
DNTI.hS Gm`ý"UI

u I Gt 1

- ,i.btiityv Cnd9s
- : ' Avail andj/or
 LIST OF CONT IBIJTORS

Richard A. Albanese Richard E. McKenzie

USAF School of Aerospace Medicine Bryce 0. Hartman
Brooks AFB, Texas, 78235, USA Crew Technology Division
USAF School of Aerospace Medicine (AFSC)
Jackson Bctay Brooks AFB, Texas, 78235, USA
Department of Psychology
University of California at Los Angeles Carl F. Melton
Los Angeles, California, 90024 Aviation Physiology Laboratory
Civil Aeromedical institute
Clyde A. Brictsor, FAA Aeronautical Center
Anthony P. Ciavarelli Oklahoma City, Oklahoma, 73125
Dunlap and Associated, Inc.
We.stern Division Laynw, P. Perelli
7765 Girard Ave., Suite 04 Crew Technology Division
La Jolla, Califonda, 92037 USAF School of Aerospdce Medicine (AFSC)
Brooks AFB, Texas, 78235, USA
Edward P. Buckley
William F. O'Connor Alan H. Roscoe
Tom Beebe Royal Aircraft Establishment
Feieral Aviation Agency Bedford, UK
National Aviation Facilities Experinental Center
Atlantic City, New Jersey, 08405 Prof. Gaetano Rotondo
Italian Air Force Medical Service HQ
R. Cannings Via P. Gobetti 2 - 00185 Rome, Italy
Royal Air Force Institute of Aviation Medicine
Famborough GU14 6SZ R. Simmons
Hants, UK M. Sanders
K. Kimball
Billy M. Crawford US Army Aeiomedical Research Laboratory
Systems Research Branch Fort Rucker, Alabama, 36362, USA
Human Engineering Division
Wright-Patterson AFB, Ohio, 45433 Walter W. Wierwille
Robert C. Williges
Walter B. Gartner Virginia Polytechnic Institute and State University
Miles R. Mukphy Blacksburg, Virginia, 24060, USA
333 Ravens Wood
Menlo Park, California, 94025 Samu21 G. Schiflett
Naval Air Test Center
G.H. Lawrence; Patuxent River, Maryland, 20670, USA
Office of Nava, Research
Physiology Prog~ram (Code 441)
800 Noith Quincy Street - Room 433
Arlington, Virginia, 22217

Richard E. McKenzie
Edward P. Buckloy
Kiriako Sarlanis
Federal Aviation Agency
National Aviation Facilities Experimental Center
Atlantic City, New, Jersey, 08405

',.... .;
...
 -- CONTENTS,

Page

PREFACE Hi

LIST OF CONTRIBUTORS IT

INTRODUCTION- vi

Chapter I CONCEPTS OF WORKLOAD;)

ahapter:IO by W.B.Gartner and M.R.Murphy
L.hapter 2 -CONCEPTS OF FATIGUE•
by W.B.Ga'tner and M.LR.Murphy3
•:• Chapter 3 CONCEPTS• OF' STRESS

by R.E.McI(enzie , 7

Chapter 4 SOME CONSIDERATIONS CONCERNING METHODS TO YEVALUATE

"AND ASSESS WORKLOAD IN AIRCRAFT PILOTS,,
g' by G.Rotondo

Chapter 5 PHYSIOLOGIC ASPECTS OF WORKLOAD/FATIGUE/Si RESS3

by L.P.Pereli / 13

Chapter 6 SOME INSIGHTS RELATIVE TO THE MAN-MACHINE SYSTEM -

AN OVERVIEW OF TEN YEARS OF RESEARCH,
by R.E.McKenzie and B.O.Hartman 17

Chapter 7 kAIRCREW WORKLOAD ASSESSMENT TECHNIQUES;

by W.W.WierwilUe, R.C.Williges and S.G.Schiflett 19

Chapter 8 WORKLOAD ASSESSMENT METHODOLOGY DEVELOPMENT,.

by B.M.Crawford 55

Chapter 9 QUANTITATIVE MILITARY WORKLOAD ANAL YSIS 6

by R.A.Albanese 69

Chapter 10 A, VISUAL PERFORMANCEt A METHOD TO ASSESS WORKLOAD

IN THE FLIGHT ENVIRON'MENT;)
by R.Simmons, M.Sanders and -I.Kimball 73

Chapter 11 HANDLING QUALITIES, WORKLOAD, AND HEART RATE ,

by A.H.Roscoe 83

Chapter 12 BRAIN WAVES AND THE ENHANCEMFNT OF PILOT PERFORMANCE,;

by G.H.Lawrence 93

Chapter 13 PUPILLOMETRIC METHODS OF WORKLOAD EVALUATION:

PRESENT STATUS AND FUTURE POSSIBILITIES;
by J.Be.atty 103

Chapter 14 '*'AIRCREW PERFORMANCE RESEAKCH OPPORTUNITIES USING

.4. THE AIR COMBAT MANEUVERING RANGE (ACMR)'
by C.A.Brictson and A.P.Ciavarelli ) 1l

Chapter 15 A SPEECH PATTERNS AND AIRCREW WORKLOAD j

* by R.Cannings 115
Chapter 16 AN EXPLORATORY STUDY OF PSYCHOPHYSIOLOGICAL
MEASUREMENT AS INDICATORS OF AIR TRAFFIC CONTROL

SECTOR WORKLOAD ____ .

by R.E.McKenzie,.P.Bilckley and K.Sarlanis 129

0?J
 Chapter 17 VNDIVIDUAL AND SYSTEM PERFORMANCE INDICES FOR THE
ARTRAI-FYC CONTROL SYSTEM
by E.P.Buddey, W.F.O Connor t' i.Beebe 135
Chapter 18 WORKLOAD AND SiTRESS IN AIR TRAFFIC CONTROLLERS ,
Cp1
by C.E.Mton 137
Chapter 19 ASSESSMENT CORRELATES OF WORKLOAD AND PERFORMANCE .
by R.E.McKenzie
145
SUMMARY
163

II

VI
 :• INTRODUCT1ION
Task complexity is everywhere in the environment within the operational pilot functions: avionics systems,

commonly with a digital computer core and a wide range of sensor. and information displays; a cockpit packed
with flight displays and controls; capabilities and, at timne, requirements for multiple missions, confrontation with a
Svariety of threat systems; crowded airspace; multiple command/control/target designation systems and techniques;
a host of environmental burdens, inside and outside the cockpit. In addition, the NATO nations have seep the
emergence of multi-role aircraft and an expansion in the tactical tznployment of the helicopter. One result of these
technological and operational advancei has been a marked increase in aircrew workload. This increase in workload
has become a problem of operational significance, to the point where, in some cases, aircrew capability has become a
liniting factor in the operational employment of some aircraft in the more demanding missions. As a consequence,
problems of aircrew workload have assumed increasing importance in the NATO researcr community.

Methods of measuring work.oad have a smbstantial history in the NATO researcn community. Disciplines
represented include systems design engineering, operations research, the be'havioral sciences, aerospace mediicine,
physiology, 'iochemistry, and biotechnology in general. There has been considerable variation in the kinds of
experimental tasks employed, the measures obtained, the instrumenta-ion used, the analytic models and methods
employed, the ratio of synthetic modelling versus enmpirical data used, and the kinds of laboratory facilities required.
The measurement domains include measures of sensory threshold, measures of sensory integration, cognitive function
tests, measures of motor function, vigilance, reaction time, psychophysiologic responses, physiologic and biochemical
changes. Methodology includes a wide range of initrumentation, laboratory facilities and environme~its, inflight
measurement methods, and modelling methodF. Analysis models and experim ital design requirements also vary
considerably. Coraputer utilization in the areas of experimental programming and data processing has become comm, n-
place. Periodic overviews of current findings are necessary. There is a need for summary matrices, as wcll as a widely
endorsed taxonomy of human performince.

This AGARDograph is one such periodic overview. It is current in the sense that each chapter is a condensation or
modificant of recent papers, prepared specifically by each author to fit the objectives of this Working Group. Ongoing
research involving advances in workload measurement technology obviously cannot be represented in this report, since
the editors avoided tasking contributors with the preparation of chapters "de novo." Such is the nature of "periodic
overviews."

It will be helpful to the reader to have a "road-map" of this report. Diagrammatically, it looks like this:

CONCEPTS
Chapter 2 7Chapter
3] e 8ter
eWorkloadlD[eFatioue M[oStdress

OVERVIEWS

Chapter 4 Chapter 5 ] Chapter 61

sMethods in Physiologic Psightsi
FGenr Aspectres froBm one Lab
S~SYSTEMATIC APPROACHES

Viua Har 1 I8 do
ChapterBai
7 J.plnercFedViePit
Chapter
r[ Measurement ]Methodology Modelling
t Scher I Developmentg
AIRCREW APPLICATIONS
Chapterl10 Chapter 11I Chaple- 12 Ch~apter 13 [Chapter14 - Capel
SVisual Heart Bra~n Ptupilonwetric [ Fiekd .Voice Prints
Performance Rite - Waves [ [Opportunities

ATC AFPLICATIONS
Chptr 6 Chapter 17 Chapter 18
Psycho-Phys. Performance Physiol.
!iFactors Measures Biochem-
L " [Chapter 19 Where do
S• •-Assessment we go
"Correlates from here?

vii
 CONCEPTS OF WO&MOAD*

by

SValter B.Gartner, Ph.D.

"Miles R.Murphy, rh.D.
333 Ravens Wood
Menlo Park California 94025

In ordinary uncritical discourse, the phenomoea referred to by the terms "pilot workload" and "fatigue"
are easily distinguinhed. In its broadest and simplest aspect, pilot workload refers to how much a pilot
must do to perform a specified flight operation. Fatigue is widely understood as a feeling of tension or
wearincss, often accompanied by an obvious unwillingness or inability to continue to vokk or perform.
However, "heu attempts ere made to quantify the workload imposed on a pilot by a particular aircraft
design, or operational piocedure, or to access the effects of fatigue upon sysLem performence, important
unresolved issues arise in rejard to the a.cre precise specification of workload and fatigue concepts and
to the adequacy of asissament criteria and cnchniqttes. This chapter and the nei.t address the principle
unresolved issues in conceptuali :ing and measring pilot workload and fatigue. In a survey of the origins
of operator workload concep'ts, Johns (1) has f-und it useful to chAracterize workload as "an Integrative
concept for evaluating the ,ifezts on the human operator asfociated with multiple stresses occuring within
Lan-isachine environments." Further, he proposed to partition this broad conception of workload under
three functionally related components: 'I) input load, (2) operator effort, and (3) work result.

While Lrnader conceptioi.s mny be considered useful for indicating the range and diversity of workload
reference, the nurpose htre to to outline the principle wyoa in which investigators have elecLed to
restrict the u-s of the term. Therefore, we will discuss Jahns' basic classification scheme with only
some minor changes in terminology.

Workload as a Set of Task Demands: The compon attribute of task-demand concepts ci workload is the use
of the term to refer to requirements for task performance which can be specified without reference to any
opera .or response or activity actually applied to satisfy these requirements. The distinction between
demands, a& such, and any actual operator response including capabilities, readiness to respond, etc., is
a very important on'. One approach 1o the traatment of workload ar demand is exemplified by Klein's (2)
attempt to quantify and predict "design-specific instantaneous worklod Icvpls imp.sed upon the pilot
while in flight." In distinguisiting this approach from traditional workload quantification methods, Klein
emphasized that "workload. is addressed from the standpoint of predicting human performance requirements
as demanded by the system and its operational environment rather tban from the standpoint of measurement
of human 7usponses to those demands."

The application of task analysis techniques within a designated system-mission-environment context to

determineappitono requirementst is
k fahiliar and widely used human factors prectnce. Gartner,
ete.al, (3) proposed that demands be more strictly defined as inputs to the crew, which actually serve
(directly or indirectly), to establish crew performance objectives or to represent operational conditions
and events which in an actual flight situation would be expected to initiate crew activity or modify
ongoing crew responses. Task demands are identified using functional criteria, that ia, they are Inputs
that operate as .csponse programs or as action requirements for the crew. This diatinction between respoose
and stimulus-ariented expressions of task demand is considered to underlie some of the problems in work-
load assesameut and practical application, because different "inds of demands are often confounded. In
other wcrds, system-oriented and situation-specific demands are often confused with perceived demands by
the operator a* with the behaviorial or psychophysiological demands imposed on an operator by an asdigned
task. This task-demand concept is closely related to Jahns' (1) input lead component which he deWines
operationýlly as "a vector (L) of input data which must be transformed by the operator into a vector (P)
of output dohi to satisfy a specified performance criterion function and/or maintain a homeostatic operator
state." Th•n inout load characterization of task demand fits a vaiiety of operatnr-loading con, -rts that
dis..inguish one or mo,-e sensory channnels or modalities as important to task performance, and addresses
such concerns as channeW capacity, perceptual overloaJ, and so forth. For example, in hi9 review of task
load factors, Eartman (4) defined load as "the sum of all requirements imposed on the operator at any
instant by the system," and later distinguished load as the number of information channels affecting
operator performance.

The defining ieature of demand oriented expressions of workload is simply that they be free of any
dependence upon considEratlons of operator response or response capabilities. In view of the apparent
diff'culty in sustaining this distinction in practice, it is probably advisable to associate task derand
only with iiput or stumulus-oriented variables and to reserve workload for the response-oriented variables

Workload as Tffort: The focus of the conceptualization of workload as effort relates to how much a
operator has to do, and/or how he nust work to satisfy a epecified set cf demands. A general character-
ization of this concept of workload hao been given by Cooper and Harper (5): 'The term workload is
Intended to convey the amount of effort and attention, both physical and mental that the pilot must
provide t( attain a given level of performance."

A somewhat different emphasis is provided by Welford (6) 1..characterizing effort as "the intensity
with which action is carried out. A man may work either more or less hard at a job.' Here, the emphasis
shifts from effort required to the consideration of the effort a human operator acIually does exert in
the performance of a task.

* * This chapter was abstracted by the editor from NASA TN D-8365, Pilot workload and fatigue: a critical
survey of ccncepts aid assessment techniques with permission of the authors.
2

In his elaooration of tho operator-effort component of workload, Jahns emphasizes the operator's
readiness to respond and he ic'entifies such factors as experience, motivation, set, physiological
readiness and physical factors, as well as the general background and personality of the operator as
determinants of this operator's state.

The concept of effort is most often tined simply to r~foer to how hard a man is working and not to
the actual task performance or to the difficulty or demands of tha task. Singleton (7) has argued for
the separation of performance and effort b) invoking the familiar obser.:ation that "an operator may be
performing better in one of two tasks as compared in an experiment because he is trying harder rather
than because one task is easier than the other." Whatever it is that occurs when a man is working
harder is referred to as effort.

Workload as Activity or Accompliflment: The conceptualization of workload as activity to actual task

performance or the products of this activity. It is often used in operational studies of the effects of
operating procedures or system design on aircrew performance. In a summary report (8) on a UAL-ALPA
Joint projert to evaluate pilot workload the authors (who are not named) defined workload in terms of
the total activity of the captain and co-pilot In performing such tasks as flight-path control, vigilance,
communications, navigation, and system opecation during each phase of a actual flight. The actual
activities engaged in by crew memb,?rs have also been used as workload referents in 'ong-term studies of
crew performance factors. For example, Cantrell and Hartman (9) recorded typical flight-crew activities
over 20 consecutive days, including off-duity and administrative activities as well as those carried out
in flight, to be used as an index of workload.

Workload Assessment Techniques: A critical review of workload assessment techniques Gartner and Murphy
(10) indicates that dispite conceptual and practical difficulties the attempt to develop and apply useful
measures of pilot workload is being vigorously pursued. The wrkload techniques which they examined
included task-demand analysis, measures of task performance, psychophysiological measures and subjective
reports. None of these assEssment technijues were found to be free of significant limitations in their
sensitivity to differences in task diff!.clty, in distinguishing between physical and mental effort, or
in the reliability of data acquisition and interpretation procedures.

With respect to workload, Gartner and Murphy recommend that significant improvements in both
measurement ane management can best b! accomplished by refinements and innovations in the analysis and
measurepment of pilot effort. They state, "human-factors engineering activities are already being applied
to task-demand analysis, and effective techniques are available for this application." However, systematic
sttempti to assess effort pke se are considerable less in evidence, despite the fact that such assess-
ments are needed for the empirical evaluation and adjustment of task demands. Innovations in the direct
assessment of effort would also provide a basis for developing more effective "effort control" techniques.
They also point out that there are directly sssessible neurowuscular tension patterns which can be
reliably related to both central neurophysiologicel states ana the task-relevant phenomena of attention
and perception.

In sumary then, It can be seen tha. there are several ways of conceptualizing workload, though in
geueral they unlsht be divided inLuto a emphasis on the lipuL side (task deftasuds) or the output side (the
work output). Similarly, there are variations in the appropriate measurement techniques though here we
see no obvious simplification. The diversity of definitions and approaches accounts for this workinag
group (AMP WG-08) report, and is a condition which should be kept in mind as the reader proceeds thru
this document.

REFERENCES
1. Jahns, D. W. Operator Workload: What is it and how should it be measured? Crew System Design.
Proceed of an Interagency Conference on Management and Technology in the Crew System Design
Proces.3, Los Angeles CA., September 12-14, 1972.

2. Klein, T. J. A workload simulation model for predicting human performance requirements in the
pilot-aircraft env!ronment. Paper presented, Human Fhctors Society's 14th Annual Convention,
San Francisco, CA., October 13-16, 1970.

3. Gartner, W. B., Frenetz, H. J. and Donohue, V.R. A iull miusion simulation scenario in support
of SST crew factors tascarch, NASA, CR-2150, 19Y2.

4. Hartman, B. 0. Time and load factors in astronaut proficiercy. Symposium oa Psychiophysiological

Aspects of Space Flight, B. E. Flakerty, ed., Columbia Univ. Press, NY., 1961.

5. Cooler, C. E. and Harper, R. P. Jr. The use of pilot rating it •' evaluation of aircraft handling
qualities, NASA, TN D-5153, 1969.

-. Welford, A. T. Fundamentals of skill, Methuen and Co., London, 1968.

7. Singleton, W. T. Prologue to Section 1: Measurement of man at work, W. T. Singleton, J. G. Fox

and D. Whitfield, eds., Taylor aind Francis, London, 1971.

8. Anon. Some worklcad and enviroriencal characteriatlcs of an air carrier short haul turbo-jet
operation. Summary reports or . UAL-ALPA joint project to evaluate pilot workload on B-737
flight operations, United Air]uies, 1969.

9. Cantrell, G. K. and llartman, B. 0. Application of time and workload inalysis technics to transport
flyers, SAM-TR-67-71, AFSC, Brooes AFB, TA., April 1967.

10. Gartner, W. B. and Murphy, M. R. Pilot workload and fatigue: a critical survey of concepts and
assesment cechnqiues, NASA TN D-0365, No -ember 1976.
 CONCEPTS OF PATTMEO

by

Walter B. Gartner, Ph.D.

Miles R. Murphy, Ph.D.
333 Ravens Wood
Menlo Park California 94025

A good sumamry statement uf the recurring theme that the cen'ral difficulty in dealing effectively
with the problem of fatigue A.sone of definition was preseutAd by Wilford (I). Accordtng to him, fatipue
means a subjective state following some kind of physical or mental strair in ordinary "man-in-the-streat"
4, constructs. However, to the physiologist, fatigue me.,.ns some ki. d of reduction of response following
more or less prolonged activity. However. the psychol-qtsw, is placed in the middle and charged with
the responsibility of tackling the problem oi fatigue relative to practical human nffairs. Unfortunately,
according to Welford, tk" often evades this respoisibhlicy by dismissing fatigue as unacientific or by
rejefining the phenonena.

In another reference, Welford (2) note: that, "difficulties have led some wish to abandon the term
fatigite, yet there is a need for a term to cover those changes in performance which take place over a
period of time during which some part of the mechAnim•, whether sensory, central, or muscular, becomes
chronically overloaded." Bartley (3) eevetops the position that the inherent utilicy of the concept
will be realized only when it is clearly distinguished from such considerations as: (1) situation in
which it occurs, (2) the bodily expreesion of fatigue, and (3) the effects of fatigue on performance,
work output, and so forth. However. it will be apparent in the following overview of fatigue concepta
that such phenomenon have not been excluded for more restrictive definitions of fatigue, and that
considorable diversity in the contemporary use of the term remains. Part of the problem seems to be
related to the wide overlap between ti'r concepts of workload and those of fatigue. In the detailed
overview of these concepts, Gartner and Murphy (4) demonstrate that within an average of 30.5 workload
and fatigue indicators, well over 50% of the indicators are directly or indirectly related or overlapped
to a significant if not indistinguishnble degree.

Fatigue as a Feeling of ;.eariness or Tiredness: This conceptualization of fatigue an been characterized

by Bartley (3) as experimental or sensory-cogitative. Experimental concepts seem to be favored in
operational studies of fatigue wherein extensive use is made of subjective asseisments. In his review
of operational studies, Schreuder (5) elaborates on the subjective aspects of fatigue to suggest that,
"The ordinary sense of weariness which the pilot subjectively feels after a hard day's work should not
be labeled as fatigue." Schreuder would insist on a level of intensity of this feeling of weariness
"which is an excess of the expected normal fatigue and which is cumulative and of such amount as Lu
alter the pilot's judgement and ability." Factor analytic studies of fatigue indic.ate that the
sensation of fatigue has three major components: (1) bodily tiredness end drowsiness, (2) weake'ned
motivation or concentration and, (3) a group of physical complaints, not unlike those of psychosowatic
disorders. Other investigators are satisfied with more global and unqualified definitions: Yoshitake
(6) "The feeling of fatigue signifies overall unpleasantness experienced by workc:! and is not quite
the same as cumplainlt uf symptoms of fatigue."

Fatigue as a Clinical Syndrome: In clinical practice, subjective complaints and/or spe,!ific sets of
signs and symptoms are regarded as useful working definitions for fatigue. Mohler (7) has outlined an
exte JAve list of signs and symptoms for both physical and mental fatigue, with the physical signs
expressed primarily in terms of physiolcaical functions, i.e., increased blood glucose, increaaed lag
in pu, tllary respons', inst-.bility of neuromascular coordination, etc. Mohler's mental symptoms are
expres. ',l i-!terms r psychogenic and emotional dysfunction and include increascd irritability and
intolerance, tendency i.odepreesion and withdrawal, and increased sex drive, etc.

Hartman (8) suggests a three-category classification of fatigue (acute, Lumulative and chronic)
characteilzing acutt fatigue as thnt normally occurring between a pair of sleep periods, and cumulative
fatigue as occurring over a period of day or weeks as a result of inadequate recovery from successive
periods of acute fatigue. Hartran urges a clinical definition of chronic fatigue as "a psychoneurotic
syndrome characterized by difficulty in committing oneself to a active or aggressive course of action,
and by a generalized withdrawal or retreat from conflict which is intolerable for situational or
personality reasons."

Fatigue as Perfoimance Decrement or Skill Impairment: Fatigue concept referents in this category, like
the clinical signs and symptoms just cited, are often treated as indicators or effects of fatigue rather
than a distinguishable state. For example, Bartlett (9) states "Fatigue is a term to cover all those
determinable changes in the expression of en activity which can be traced to the continuing exercise of
that activity under its normal operating conditions, and which can be shown to lead, either to deteriora-
tion in the expression of that activity, or more simply, to results within the activity that are not
wanted."

A more formal expreasion of these changes in performance is provided by Hull's development of the
reactive-inhibition construct (10). Hull's behavioral restatement of Spearman's general law of fatigue
and Pavlov's concept of conditioned inhibition is: "Whenever any reaction is evoked in an organism
there ih lfif a conditio.' or state which octs as a primary, negative in that it as an innate capecity to
prouce a cessation of the activity wh!i'h produced the state, we shall call this stete or condition
reactive inhibition".

1 This chapter was abstracted by the editor from N,'.SA TN D-8365, Pilot workload and fatigue: a critical
survey of concepts and assessment technqiues vith permission of the authors.
4

One of the more interesting variants is Bartlett's (11) concept of "skiid fatigue." On the basis,
of studies pilot performance in the Cambridge psycbological laboratory he suggests that it is, "Necessary
to draw a broad distinction between fatigue produced by continued hard physical work and that produced
by work which calls for little continuous muscular effort, but demands persistent concentration and a
high degree of skill." Skill-fatigue, also distinguished from mental fatigue, is said to occur when a
task, such as piloting a plane, requires complex, coordinated, and accurately timed activities. In other
Cambridge studies, deterioration of skill performance was apparent after about 2½ to 3 hours of stimulated
flying~manifesting primarily as a progressive lowering of standards of performance, the missing important
information displays, and the gross mistiming of Interrelated control actions.

FatiAue as a Neurophysiological Condition or State: In traditional or classical studies, fatigue was

ieferred to a particular neuromuscular site, that is, to specific motor units or muscle groups or orpans
or tissce structures and then defined in terms of specific biochemical and/or response capability changes.
This comparatively narrow tocus is now generally recognized as only one aspect of fatigue. In a dis-
cussion of neuromuscular fatigue as a special instance of a more general condition as Basmajian (12)
points out "I shall observe, at once the traditional and necessary warninq that fatigue is a complex
phenomenon and perhaps a complex of numerous phenomenon. The fatigue of strenuous effort is probably
quite different from the weariness felt after a long day's routine sedentary work. Undoubtedly, the
following types exist: emotional fatigue, central nervous system fatigue, general fatigue, and
perpherial neuromuscular fatigue of special kinds."

Welford (13) feels that fatigue is best conceptualized as a local neural impairment.

While Grandjean (14) shares the view of many investigators that fatigue is a central neuruphysio-
logical condition and is located in the central nervous system more spccifically, in the brain stem
reticulor activation system. His conu.eptualization of fatigue as a conditon of the central nervous
system is based on early studies of the role of the brain-stem reticular formation in producing and
maintaining various levels of inactivity, airousal, and activation.

Welford has also suggested that consaLdering fatigue as a central phenomenon attempts to integrate
the comparatively less accessible condition of mental fatigue with the more readily observed condition
of neuromuscular fatigue. He states, "It appears that in the intact organism changes in the muscles
brought about by prolonged or repeated contractions can, according to circumstances, have one of two
limiting effects. Either the muscles themselves become temp,:arily incapable of further contraction or
the coi,ýition of the muscles produces afferent stimuli and these in term affect the central mechanisms
and lead Lo the cessation oi efferent impulses." If the term mental fatigue is to have a meaning in
line with that of neuromuscular fatigue, it must denote the impairment of some brain mechanism as a
result of long continued use. The impairment must be reversible in the sense that it disappears with
rest, and may take the form of lowered sensitivity, or lowered responsiveness, or lowered capacity.

This definition by Welford permits a distinction to be made between m',ntat fatigue and other central
conditions such as adaptation, habituation, and monotony or boredom which also lead to a decrement in
performance over time. However, others see no significance differences in operational definitions of
reactive inhibition, habituation, and central fatigue. Grandjean (13) expresues the popular view that
boredom are components of the fatigue condition a.,d are related to the task situation: "If the workload
is too heavy, fatigue due to physical or mental effort is to be expected; if the worker is underloaded
or forced to conduct repetitive work, fatigue due to monotony will be produced."

Fatigue as a Level of Energy Expenditure: The energy expenditure approach to fatigue focuses on the cost
of protracted effort, wnether mental or physical, in terms of the energy investments or transformations
required t, sustain it. A formulation of the energistic approach by Dukes-Dobos (15) defines fatigue as
a term to denote a normal psychophyaiological process which starts immediately after the beginning of
any physical or mental activity and which consists of the utilization of the bodies' energy stores, the
accumulation of the breakdown products, and the activation of adaptive mechanisms which maintain the
homeostatis of the organism."

Cameron (16) considers the term of fatigue to be no more than a useful descriptive term for a
generalized stress response over a period of time. "The human stress response is generalized in character,
involving the whole system of biological emergency mechanisms. Since it implies, by definition, an
abnormal demand on the energy resources of the system, iL is fatiguing. The degree of fatigue experienced
may depend to some extent on the level of the stress responsc, but will depend primarily on its duration."
Here, he emphasizes the duration of the stress response of the organism, not necessarily the duration of
the stressful condition. This is a critical distinction, because he argues that the length of time
needed to return to a normal arousal level, that is a normal level of biological emergency mechanism
activity is good index of the severity of fatigue.

McFarland (17) has criticized the focus on physiolo6ical factors and fatigue citing the familiar
arguments that effects observed in the laboratory are not always found in actual work situations and that
other factors often influence energy reserves and utilization capacities, mainly physical condition and
motivation, and that the metabolic costs of mental work are very slight. In this argument, characteriza-
tions of the pilot's job as predominantly cognative and not physical or muscular are frequently cited
to question the relevance of physiological factors, especially those derived from studies of heavy phy Acal
workloads. It would seem that if this concept of fatigue as a level of energy expenditure is to bear
frujit we must have a clear focus on the higher order concepts of energy mobilization and channeling in
the individual, rather than focusing on the localization and reduction of this response to metabolic
activities in particular muscles or tissues.

Welford, who views both mental and neuromusclar fatigue as effects of loading, would agree that
fatigue is a consequence or concomitant of workload. Bartley would also agree with this relationship
while insisting that fatigue is a condition of the individual and is not to be defined !n terms of
e7ternal situations or even work products. In this situation, he consider3 energy expenditure, paced
 5

performance, prolonged activity, and demands upon particular body mechanisms to be typically fatigue-
producing.
The primary difficulty in applying fatigue assessment techniques more explicitly is the multi-
dimensional character of fatigue phenomena and their interaction with even more complex phenomena of
individual motivation and stress tolerance. The approach of Bartlett to fatigue assessment utilizing
the application of the concept of skill fatigue is important since observable changes in pilot behavior
during prlivry task performance can be clearly and directly related to the accomplishment of flight
management and/or aircraft control objectives. It must be concluded that factors other than task
demands or protracted effort are more significant in the occurrence of fatigue. These other factors
include individual differences in personality, motivation, physical fittness, and life style, as well
as such situatioual factors as operational management policies, disruption of established biorhythms,
sleep patterns, and exposure to various environmental stressors. The relative contribution of personal
versus task-specific fatigue factors is an important unresolved issue.

REFERENCES

1. Welford, A. T. The psychologist's problem in measuring fatigue. Symposium on Fatigue, W. F.

Floyd and A. T. Welford, eds., H. R. Lewis and Co., London, 1953.

2. Welford, A. T. Fundamentals of Skill, Methuen and Co., London, 1968.

3. Bartley, S. H. Fatigue: Mechanisms and Management, Charles C. Thomas, Springfield IL., 1965.

4. Gartner, W. B. and Murphy, M. R. Pilot workload and fatigue: a critical survey of concel.,d
and assessment techniques, NASA TN D-8365, November 1976.
5. Schreuder, 0. B. Medical aspects of aircraft pilot fatigue with special reference to the
commercial jet pilot, Aerosp. had. Special Report Pub. 37, (4) 1966.

6. Yoshitake, H. Relations between the symptoms and the feelings of fatigue. Methodology in
Human Fatigue Assessment, Taylor and Francis, London, 1971.

7. Mohler, S. R. Fatigue in aviation activities, Aerosp. Med., Vol, 37, 1966.

8. Hartman, B. 0. Psychological factors in flying fatigue. Psychiatry in Aerospace Medicine,

Vol. 4, C. Perry, ed., Little Brown and Co., Boston MA., 1967.

9. Bartlett, F. C. Fatigue in the air pilot, Rept. 488, Air Ministry, Flying Personnel Research
Committee, 1942.

10. Hull, C. L. Principles of Behnvior, Appleton-Century-Crofts, NY., 1943.

11. Bartlett, F. C. Fatigue following highly skilled work, Proc. Roy. Soc. Ser. B. Vol. 131, 1943.

12. Baaujian, J. V. Muscles alive: their function revealed J- electromyography, The Williams and
William3 Co., Baltimore, 1974.
13. Welford, A. T. Fundamentals of Skill, Methuen and Co., London, 1968.

14. GrandJean, E. Introductory Rimarks. Metbodology in Human Fatigue Assessment, Taylor and
Francis, London, 1971.

15. Dukes-Dobos, F. N. Fatigue from the point of view of urinary metabolites. Methodology in Human
Fatigue Assessment, Taylor and Francis, London, 1971.

16. Cameron, C. A theory of fatigue, Ergonomics, Vol. 16, 1976.

17. M..arland, R. A. Understanding fatigue in modern life. Methodology in Human Fatigue Assessment,
-aylor and Francis, London, 1971.

Ic

L........
 - - CONCEPTS OF STRESS

by

Richard E. McKenzie, Ph.D.

Crew Technology Division
USAF School of Aerospace Medicine (AFSC)
Brooks Air Force Base, Texas 78235
USA
Te paraphase the Webster Dictionary definition of stress, we find that stress is a physical,
chemical, or emotional factor to which an individual fails to make a satisfactory adaption and which
causes physiologic tensions that may be a contribut.ry cause of disease. While this publication is not
the format for a clinical discussion of the psychological, psychiatric, and biological aspects of the
effects of stress on the airmen and its related effects upon the man-machine interface, the concept of
stress and its decrementing role is an Important one in modern aerospace flight.

In his discussion of psychology and flying fatigue, Hartman (1) defines acute fatigue as tb - which
occurs in a single flight, during a single day, or more appropriately, between a pair of sleep periods.
Here the recovery from acute fativue is a function of the adequate amount of rest available. But even
without prolonged or even relatively short rest periods, the fatigued flier can mobilize his resources
and return briefly to near rested levels of eficiency when the occasion demands. Hartman also defines
cumulative fatigue as that which occurs over a period of days or weeks, and is the result of inadequate
recovery from successive periods of acute fatigue. Recovery from cumulative fatigue is also dependent
upon adequate rest. However, without an adequate recovery schedule, the pilot finds himself fighting an
enhanced workload self-generated by his own loss in airmanship and efficiency, and finds that the longer
cumulative fatigue continues to build un, the longer it will take for him to recover his reserve and his
capacity to mobilize himself to meet high demand situations.

The term "chronic fatigue" has a special psychiatric meaning and is defined as a neuropsychiatric
disease. Chronic fatigue is a psychoneurotic symptom characterized primarily by difficulty in committing
oneself to an active or aggressive course of action and by a generalized withdrawal or retreat from a
conflict which is intolerable for situational or personality reacons. Thus, this entity is rarely seen
in either the military or civilian pilot/aircrew member.

Like fatigue, stress has its acute phases, one of which is an alerting arousal response enabling
the person to perform better and to otherwise adapt himself to an emergency. Cumulative stress, on the
other hand, is a build-up of physiological, chemical and emotional factors over a period time until
some kind of maladaption occurs. As Selye (2) points out, stress is a reasonably normal component of
modern every day life and can be adaptive, but cumulative stress becomes maladaptive and ultimately then,
stress becomes dist:ass.

Selye (3) has also discussed his general adaptation syndrome which he conceptualizes as the
defensive response of the body, through the endocrine system, to systemic injury evoked by stress. This
is worked out by an initial stage of shock, like an arousal or surprise reaction, followed by a stage of
growing resistance to the injur,, (adaptation), followed in turn by a final stage of healing or exhausLion
and death if adaptation fails. Note that there is no alternative course of action, one must either
resist the bodily effects of stress by healing or one must become exhausted, and ultimately be defeated
by the effects of stress. In short, then, one cannot ignore the effects of cumulative stless.

•parks (4) in a chapter entitled "The Clinical Aspects of Psychiatriy" Illness in Fliers," brings to
light a relatively interesting aspect of a pilot's career progress. One could call this situation one
of selective screening, because flying personnel, through the almost automatic process by which they
learn the skills required for modern pilotage, be it military or civil aviation, are screened for emotional
stability. During their initial period of training, they have close supervision, with exposure to moderate
levels of stress together wit', the requirements of rigid discipline demr.nded by attention to procedures
and the awareness that tL_ aerospace environment is an unforgiving mistress. In order to successfully
pursue his chosen career, the pilot must adapt to the early stresses of flight training. Flight itself
poses additional stressors to which he must form adaptive methodologies or strategies which act as a
further screening process. Following the completion of flying training, additional periods of flying
duties, upgrading and so forth, further cause him to adapt more and more strategies for coping to the
point that almost any unstable individual would be self-eliminated prior to any operation assignment.
For the military pilot, combat poses additional and unique streasors to which he must adapt, ground
himself, or be ultimately surrendered. It is a small wonder that we expect to find any psychiatric
casualties once this screening process is completed.

However, the very aspect of recognizing this screening process implies that we also recognize that
we have a highly selected individual who will almost invariably stand up to most of life's stresses.
Therefore, we are Anclined to form an almost mythical concept of the pilot as being inviolate and
unaffected by ordinary and inordinate stressors. This of course, is far from true. Carlos Perry in a
chabap"' on aerospace psychiatry (5), examines some the stressors of aerospace operation. He points out
that "potential danger, physical discomfort, energy demands, attitude, and enforced physical passivity
.av- seer well recognized as stresses and need little further elaboration. He also shows that increased
saecialiation in aerospace operations is the source of stressors that were not apparent when flying
sz.:ivitiea were more generally uniform. Thus, a given airman may well tolerate stresses associated with
flying long conventional type cargo hauls in the company of a crew, but not be able to successfully cope
with stresses of solitary, short duration, high altitude intercept flights in high performnce single
engine jetcraft. He points out the incongruity involved in the military concept of alert. Here, we have
aircrew men who are interested in flying, going places, and seeing things and we force them to sit for
long periods of time in an alert facility, away from family and other satisfying activities. Perry states
out that boredom has also become a major stress factor. Here automation, lack of diversification, endlaes
a8

routines, and increasing length of individual flights contribute to the production of boredom. He states
that "while boredom may be considered to be a benign type of stress, that these feelings are not far
removed from the more serious faelings of lack of ambition, futility, or even depression." He points out
that the nature of aerospace operations are source of another major category of stress, with the
necessities of long and frequent travel, varying periods of absences which can be a ..ource of severe
stress to maritial and parental activities.
Even the complexities posed by the various types of aircrew equipment for providing a livable
environment for man imposes their own constraints and physiologic stresses. Mission and operational
requiraments present the modern pilot and crew with everchanging complex tasks which provide another form
of stress. These major sources of aircrew stress are compounded by tbe individual's internal psycho-
physiologic reaction to stress and to general external wtressors, such as personal, career or family
problems.

The human body is known to adapt to or to withstand severe conditions of immediate or acute stress.
However, it does not respond as well to long-term or cumulative stre3s, whatever its source. While
aircrew personnel are subject to special forms of stress the basic reaction to stress is uniform., All
stress, physical, emotional and so forth, is respondf.A to by some kini of an adaptive or avoidance
reaction. The basic need is to protect oneself from sore and more stress. Physically the initial
response is a musculoskeletal tension/arousal response, which together with the corresponding changes in
glandular, organ, and nervous systems prepare the individual to retreat from the situation or to comfront
it, the classic flight or flight reaction.

In the case of cumulative stress, the muaGuloskeletal and organ system of the body tend to be
continually activiated to the point where the individual is now stressed even when the original source
of stress is absent. The stress becomes intarnalized and most stimuli, either internal or external,
become sources of stress. We now have a chronically tense, irritable, agitated, disturbed person.

As cumulative stress continues, the musculoskeletal and organ systems of the body may start to
undergo pathologic changes. We begin to see psychosomatic systems of stress in the muscles of the neck
and shoulders or other parts of the body. Chronic muscle tension produces decreaseJ blood flow in these
tissues, with pain and joint pathology. Chronic organ reactions yield typical symptoms of the gastro-
intestionel tract, such as st--mach cramrn, ulcer, colitis, and so forth.

It is obvious that chronic stress and its related pathology cannot be ignored. The aircrew member
who experiences cumulative stress from one or multiple sources and who is then further subjected to
additioral increwents of stress from personal or aircrrft equipment, from the demands of the mission, or
from fatigue or external 9tressors will find his best skills and efforts decremented. This degrading
of performance is obviously related to the disaster or near-disaster of the aircraft accident, but not
necessarily in the causal sense. A recent survey of USAF accidents fails to support degraded performance
or stress as a causative agent. Instead, strcss and decremented performance are seen as factors which
are contributory in that they act to "set the stage", preparing the psychologic and physiologic world
of the pilot in such a manner that he is not able to respond effectively to one or more addiuional untoward
events. This is the insidious danger of stress pathology and constitutes an excellent reason for including
a pre-accident mental status investigation as part of the accident review process. However, as in most
things, "an ounce of prevention is worth a pound of cure." How can we prevent stress, or better yet, how
can we prevent dis streas as the result of cumulative stress?

There is a solution for cumulative stress. It has long been known, in a simple minded way, that one
cannot be tense and relaxed at the same time. Thus, the adaptive response to stress/tension is one of
relaxation. Adequate recovery times from periods of cumulative stress with provision for recreation are
important. Perhaps even more important, however, would be a conditioned learning program wherein the
individuel is taught to avoid the effects of cumulative stress by keeping himself in a relatively relaxed
state. There are several. ong-time approaches to this type of training. One of the earliest being that
of Schultz with his autogenic training followed by Jacobson's progressive relaxation training and more
recently by such mediative techniques as transendental mediation (TH). An intriguing modern day addition
to these forms of relaxation methodologies is that of biofeedback techniques where an electronically
generated signal from the muscle or organ system involved is available to the individual as a learning
technique. This is based on an axiom in information theory that states "the controllor giving information
about the state of the system can then exercise control over that system." It has been demonstrated that
the central nervous system can exercise exquisite control over the CNS, the spinotha]amic ard automonic
nervous systems. Biofeedback is merely one way of giving the controller information about the state
of the system so that he can learn to exercise the necessary control. While biofeedback simply utilizes
modern electronic technology, we must realize that there are many adaptive strategies which could be
employed and also realize that human beings come supplied with internal biofeedback signals which
undoubtedly play an important role in both adaptive and maladaptive behaviors.

REFERENCES

1. Hartuan, B. 0. Psychological factors in flying faZigue, Psychiatry in Aerospace Medicine,

J. G. Perry, (ed.), Little, Brown and Company, Boston, 1967.

2. Selye, H. Stress Without Distress, J. B. Lippincott Co., 1974.

3. Selye, H. Stress, Acts Inc. Medical Publishers, Montreal, 1950.

4. Sparks, J. C. Clinical aspects of psychiatric illness in flyers, Psychiatry in Aerospace Medicine,

J. G. Perry, (ed.), little, Brown and Company, Boston, 1967.
 5. Perry, C. 1, Z. Aerompaca psychiat'ov Aroup. Pld., 2nd. Rd., Hugh W. Handel, (ed.), The
William & Williams Company, Baltimore, 1971.
6. McKenzie, R. E. and Bainesa, R. H. 3iofeedb&ck t"arapy. In: What's psychotherapy a-d who needs
it. (ed) by H. L. Collier, O'Sullivan, Wood&,de md Company. Phoenix, AZ., 197?.

IT
 ""OKJ CONSID!RATIOS C"CWOCUN'4C WT TO E'AWAATE A1D ASSIESS
VORXEA)AD IN AIEZLAIN ?ILOTS

by

Professor Gaetano Rouonro, Brigadier General LAY, MC

Italian Air Force Medical Service H.Q. - Italy
Via P. Gobetti 2 - 00185 Ros

If the nature and entity are analyzed of the various stressing and fatiguirS factors that are ac"ng
on the body and psyche of aircrafts' pilots during their specific activity, it appears obvious that under-
lyi-g the exercise of a piloc's profession to a basic situation which ultimately peraeates the whole of
his activity and exerts a multiplicity of effects on the physique and psyche of the same pilot.

The fundameatal ch~racteristics of such a situation may be sumarized as !ollows:

1. Flying involves the use of a nachine that is required, unlike other machines, to respect certain
aerodyntmc laws. Any infraction of these lafe involves an hmediate risk of crash and accident. In the
pilot's profedsichl activity, therefore, laife depends on the mathera and Its continuous efficiency, a
situation that in actual practice expresses itself in the form of a permanent image of potential "vulner-
ability"*undoubtedly present In the subconscious of each and every pilot.

2. Pilot's activity depends a great deal on the spatial environment of the aircraft, the three-
dimensional displacement and rapid trancition of the aircraft, and, indirectly, on the various conditionis
that have repercussions on th-e human body (i.e., accelerations, acoustic and non-acoustic vibrationn,
equipment, sensorial stress, etc.). All these conditions constitute links in a chair of factovs which
readily explain the wealth of interferences that act on the som.ato-psychic balance and, consequently, on
performance, adaptability and, in the lcng run, on individual fatigue.

3. Flying does no?: just represent a technical or operative activity, i.e., a job, but rather "a
vital activity and an 'in toto' reaction of the ego to the environment."

Upon such a basic substrate, which is in itself potentially stressing and qualitatively conmon to all
pilots irrespective of their specialization and the type of aircraft they fly, there then act interferences
due to the various physical and psychic factors, each of which plays a specific and individualizing role
both in connection with general and particular aircraft (fighter, transport, reconnaissance, rescue, or
helicopter, etc.).

It wou5ld certainly be interesting and Important if ii.were possible to define the degree and limits
of stch psychophysical workload by mean4 of technically valid and acceptable scientific methods with a
view to obtaining differential qualitative and quantitative assessments of the various flying speciali-
zations. In fact, numerous methods have been proposed peovodically for obtaining a measure of workload
by quantitatively evaluating the functional changes that fatirue can produce. As known, such changes may
consist of an increase of the duration and inconstancy oI. the psychomotorial reaction times; an increase
of the latency time of the pupillar reflex; a diminution of the capacity for rapid binocular fusion; an
increase of the accomodation time for near and distant vision; a reduction of the critical flicker fusion
frequency (L, 12) and changes of other ophthalmic iiidex.s; modifications of the characters and duration of
the monosynaptic spinal reflxes produced, for exampsl, in the area of the sciatic nerve (1); variations
in the duration of the central nervous time of the orbicular blinking reflex under light stimulation, and
the time needed for a complex montal process (11); redtiction in muscular force and muscular tone; increased
instability in neuromuscular coordination; increased loss of electrolytes through cutaneous sweating;
reduced circulating plasma volume; variations in the urinary excretio-a of curticosteroids (7) and cathe-
colanine (2, 6); variations in the lactacidemia, glycesia, and cholesterolemia values, the ratio between
alpha and beta lipoproteins, the number of the eosinohiles, and the hematocrit index; and, finasly,
electrocardiographic changes and variations in the Ruffier and Dickson index of cardiac resistance (4).

Quite obviously, however, all these methods lend themselves very readily to criticivi. Indeed, none
of the results obtaineJ by these methods are capable of being interpreted in a unique manner. In fact,
these methods measure functional cha-gee that are or can be influenced considerably by a wealth of other
factors, both endogenou3 and exogenous, including first and foremost, the subject's age. Therefore, if
one wanted to make a comparative evaluation of the amount and precocity of the stress and the psycho-
physical workload produced by the individual streas.ng factors connected with flying, one would admit that
it is extremely difficult to find a precise differential criterion that could .e used to obtain a quanti-
tativ3 graduation of this workloae. This is not only because the subjor . element, here understood as
tha individuality and extreme variability of the response of the single subject to every type of stimulus,
has a predominant weight in this particular activity; It also depends on the nature and entity of the
reaction to any kine of stimulus which are, in turn, conditioned by numerous and extremely variable
individual, environmental, and circumstantial factors.

After those necessary premises concernIng the- difficulties of a unique interpretation of all pro-
posed dltgnostic methods and the preponderance of rsychical workload on physical one in the piloting
aircrafts, it is oi,r opinion that between the above-mentioned functional chances eventually produced by
emotional and psychic fatigue, a pazticular attention could be reserved - in Aviation Medicine - to
var•a•io3s in the utinary excretion of corticosteroids and especially catecholamine.

It is well known that every stress - no matter if physiological or emotional - is capable of inducing
organic reactions due to the inerease of cortic.osteroids and catecholamines in the blood circulation.
According ta many authors who have studied the phenomenon in the aviation field, there are increasss also
in particular flight conditions, particularly those likely to set up a state of stress.
12

Therefore, 4t is possible to conecluse that the determination of the urinary excretion of catecho-
lemines in particular, as indication of a possible psychic stress, could be used &M a methoa; to objectify
"motions." This uould have a usefui application in practice to reveal emotional states undergone in
flight, particularly duvirig the phase of training and other all conditions of considerable psychical
engagement in the couroe of aeronavigation.

In other wordi , the deteruination of such substances would then give useful information about the
presence of stress and would also allow to evaluate the intensity of the latter (and of consequent work-
load). The same evaluation may also be obtaived by determining the quantity of venilmandelic acid (VNA
excreted with urine, such an acid taking its origin from the metabolism of catecholamines (5).

These methosa sight be usefully and practicaily employed with the purpose of obtaining an objective
measuremenc of the emotional aspects of the human personality in real conditions, and then quantitatively
evaluating the workload (especially psychic, but also physical and physiological workload) in the pilot's
professional activity.

REFERENCES

1. Gualtierotti, T., R. Nargaria, and D. Spinelli. 1958. Effects of stress on lower neuron activity.
Exper. Ned. Surg. 16:166.

2. Dlepping, J., 0. Buisson, J. Gserrin, A. Escousse, and J. P. Didier. 1963. Evaluation de

lVelimination urinaire des catecholamines ches des pilotes dWavions a reaction. C.R. Soc. Biol.
157:1727.

3. Krugman, H. E. 1947. CFF as a function of anxiety reaction: an exploratory study. Psychosom.

Med. 9:269.

4. Le Roux, R. 1960. La fatigue operationelle des pilotes d'helicopteres. Revue des Corps de Santa
1:493.

5. Peolucci, G., and G. Blundo. 1973. Determination of emotional condition in student pilots during
air-navigation by dosing vanilmandelic acid (VNA) excreted with urine. Riv. Ned. Aer. Spaz.
36:184.

6. Paolucci, G., and G. Blundo. 1975. Catecholaninic excretion in student pilots. Riv. Ned. Aer.
Spas. 38:27.

7. Rotondo, G. 1955. On the treatment of pilots affected by operational fatigue with dehydroiso-
androsterone. Riv. Ned. Aer. 18:78.

8. Rotondo, G., and A. M. De Angelis. 19o6 Acetil-aspartic acid and citrulline in treatment and
prevention of flight fatigue. Riv. Mkd. Aer. Spaz. 29:85.
9. Rotondo, G. 1969- Experimental contribution to preventive and therapeutic treatment of flight
fatigue. Riv. Ned. Aer. Spas. 32:231.

10. Rotondo, G. 1977. Workload and operational fatigue in helicopter pilots. Aviat. Space Fnvironmental
Ned. (in print).

11. Spinelli, D., and P. Cerretelli. 1961. Analysis of central nervoub functions In particular
physiological conditions. Ned. Sport. 1:128.

12. Vozza, R. 1955. Flicker fusion frequency as a test of operational fatigue in jet pilots. lRiv. Med.
Aer. 18:771.

/1 ai
 13

PHYSIOLOGIC ASPECTS OF WoRKLOAD/FATIGUE/ST.ESS*

Ly

Layne P. Perelli, Captain, USAF

Crew Technology Division
USAF School of Aerospace Medicine (AFSC)
Brooks Air Force Base, Texas 78235
USA

It is important to recognize that the physiological mechanisms of the organism do not particularly
care nor are they necessarily aware that they are reacting to the effects of workload, the effects of
fatigue or the effects of stress. Physiological mechanisms provide a common link between the concepts
of workload, fatigue and stress. Traditionally the basic physiological approach to fatigue involves
the measurement of energy expended in performing a given amount of work. As early as 1919 to 1920,
Waller and De Decker (1) measured the carbon dioxide production of workers and were able to relate
increases in carbon dioxide production to a reduction of work output during a night's activity. They
use the term "physiological cost" to describe the increased metabolic demands resulting from increased
fatigue and related lowered performance. Page (2) (3) has suggested that the concept of fatigue be
replaced with the concept c• metabolic cost, and Bitterr an (4) has suggested that the concept of fatigue
be defined as a reduced efficiency resulting from continued work and reversible by rest, with efficiency
defined as the ratio of performance to expended effort. Effort was to be determined from metabolic cost
indices.

Concepts of physiolog7c cost are related o Selye's concept of the general adaptation syndrome in
which any stress to which the body is exposed creates an overall non-specific, systemic reaction to cope
with or reduce the stress (5). It is theorized that fatigue creates a stressful condition to which the
body tries to adapt, and in so doing produces an abnormal set of physiologic indicators which can be
evaluated as to the severity of the fatirute/stressor.

After reviewing several fatigue studies showing no significiant or dramatic performance decrement
and one study with a performance increase, Cameron (6) concludes that performance measures are too erratic
and unreliable to serve as indicators of fatigue. He feels that the term fatigue should be used as no
more than a descriptive ter.i !or generalized stress response ever a period of time, and that the best
index of acute anl chronic effects would be the time required for biologic emeigency mechanisms to return
to a normal arousal level.

Pursuing this same line of thinking, Harris and O'Hanlon (7) provide a review of what is known about
the recovery of man from exposure to certain adverse conditions, such as sleep deprivation, abnormal
work/rest cycles, prolonged physical work, and environmental and situational stressors. Their purposa
was to determine if recuvery functions can predict how long a man can maintain effective performance
before he must be relieved and how long a rest period is required before he is ready again to perform
effectively during continuous military operations. They conclude that while there is insufficient
knowledge now available to make such predictions, the following list of potential physiological failures
seems most important to consider and reversal of these impairmentb may provide practical indications
that recover,, has taken place: 1) Degraded physical workinng capacit;, 2) Inadequate iron reclamation,
3) Mild cardiel fatigue, 4) Paroxysmal cerebral cortical activity, 5) L'paired carbohydrate metabolism,
6) Thiamine deiiciency, 71 Involuntary hypohydration, 8) Glycogen exhaustion, 9) Increased susceptibility
to infection, 1G) Imbalanced protein metabolism and 11) Adrenal cortical and medullary exhauction. They
feel like Cameron that changes dut. to fatigue will become apparent in tht physiological systems before
p.rformance degradetion occurs. This implies, of course, that even though a given schedule of work has
not yet produced performance decrement, work-rest cycles should be strc'tured so that severe changes in
the physiological systems ire prevented.

Following this same conc-pt that the physiologic cost of fatigue is generally not an immediate
problem providing the individual receives sufficient recovery time, Hartman and Cantrell (8) have taken
the position that the best approach to maintain man's capacity for skillful work is to engineer the
system so that physiological degradation is eliminated. This implies that if physiological indicators
known to be associated with stress reactions axt found to be within normal limits, then it is presumed
that no performance decrement of operational consequence has occured. Thus, the problem is to quantify
these physiologic limits in relation to a criteria of perfornance degradation in such a way as to cause
system managers to design, man, and use an operational system in such a way that these limits are not
exceeded. One of the difficulities in using physiological indicators for evaluating workload, fatigue,
or stress is of a temporal nature in that some physiological responses can be observed only after periods
of hours or even days while other responses occur almost instantaneously. Some measures are unobtrusive
which could be used in operational situations yhii.; others are somewhat impractical or often impossible
to obtain.

We will first discuss the long term physiological indicators of stress workload and fatigue recovered
from the organism and measured as urinary metabolites, namely the 17-hydroxycortico steroids (17-OHCS)
and the catecholemines (epinephrine and norepinephrine). The following general review of 17-cortico
steroids and catecholemines is taken from Guyton (9).

Steroids, namely cor-tisol are excreted into the blood stream from the adrenal cortex in response
to a wide variety of stresses. Steroids enable the body to cope with stress through its effects on
carbohydrate, fat and protein metabolism. It causes a stimulation of gluconeogensis by the liver and a
.4 •decrease in glucose utilization by the cells which in turn raises the blood glucose concentration. At

* This material was abstracted from a chapter of Captain Perelli's draft doctoral dissertation by
Richard E. McKenzie, Ph.D.
the same time it causes a reduction in protein storas in all parts of the body except the liver. Blood
amino acid concentration goes up, transport of amino acids into extra hepatic cells Is diminished and
transport of amino acids to the liver is enhanced. Amino acids are thus mobilized from the tissues to
the liver. Finally, fatty acidu are brought out of adipose tissue Increasing their blood concentration
which increases their u,.ili-tion for en3rgy. The adrsmnl cortex secretes steroids in response to
adrenocorticotrophic hormones from the adenohypophysis which is under direct control of the hypothalmus.
With this indirect feedback mechanism, levels of cortisol can continue to rise to very high blood
concentrations as long as the stress agent continues to stimuiate the hypothalmus in some way. Cortisol
fixes to its target tissues in atout 20 minutes after release. The normal blood concentration is about
12 micrograms per 100 milliliters and its half life in the blood is 100 minutes. The normal secretory
rate is 15 milligrams per day of which approximately 75% is excreted in the urine.

At this point it should be obvious that one can measure 17 keto-steroid production from either
blood or urine sampling. The only problem one should be aware of is that there in a difference in the
concentration time of 17-ORCS found in blood plasma as opposed to urine by about two hours. Increases
in 17-ORCS excretion have been found for various anxiety producing situations, such as electroshock
treatment and with the use or administration of hallucinogenic drugs and in the viewing of mildly stress-
ful motion pictures. Berkum, Bialek, Kern and Yagi (10) performed an extensive series of experiements
simulating five stressful military situations in which the subject was led to believe that he was in
immediate danger of losing his life or of being seriously injured, or that by his actions he has seriously
injured one of his colleagues. All of these stress situations resulted in elevated 17-ORCS excretion
and the level of the increase was related to the presumed level of stress induced for each s8tua'lion.

Miller (11) provides a review of the many studies in which 17 keto-steroids have been found to
increase due to the stress of military flying. In 1943, Pincus and Hoagland (12) conducted three sets oi
experiments which related steroid excretion and flying stress. They reported not only significantly
increased steroid production, but that individual performance scores were positively related to the level
of steroid increase. They also reported that increases in steroid production were found to be related to
independent rating given by the pilot's squadron commuander on their individual su.ceptability to fatigue.

Catecholamines: Catecnolamines are secreted by the adrenal medulla in response to stimulation from the
sympathetic nervous system. The relation-ship between the adrenal medulla and a threating situation
was first demonstrated by Cannon and de la Paz (13). While the proportions of catecholamines which are
excreted depend upon the physiologic conditions, on the average, 751 epinephrine and 25% norepinephrine
are excreted. Their effects on the body are the same as those caused by direct stimulation ,f the
sympathetic nervous system, but the effects last about ten times longer since the circulating catechol-
amines are only slowly removed from the blood. It should be noted that the sympathetic nerve endings
excrete norepinephrine, but in a matter of seconds It is reabsorbed or destroyed at the cellular level
by 0-methyl transferase or wmoamine oxidase. These enzymes are similar to cholinesterase which destroys
acetycholine, the agent excreted by the parasympathetic nervous system. While both the sympathetic
nervous system and the excretions of the adrenal medulla have general nonspecific effects, the
catecholamines stimulate and increase the metabolic rate of every cell in the body. However, it must be
noted that circulating catecholamines do not readily pass the blood-brain barrier. This means that
central nervous system physiology is not ab reactive to these circulating substances as is the rest of
the body physiology.
The general result of stimllation of the sympathetic nervous system is to mobilze the body for
action. Noreprnephrine causes general vasoconstriction, increased cardiac activity, increased basal
metabolism, sweating, inhibition of the gastrointestinal tract, glucose release from the liver, decreased
kidney output, and adrenocortical secretion. Epinephrine has similar effects but bas a greater stimulat-
ing effect on cardiac activity and basal metabolism and has a less constricting effect on the vascular
system of the skeletal muscular system. Normal resting secretion rates are .2 micrograms per kilogram
of body weight per minute for epinephrine and .07 micrograms per kilogram of body weight per minute for
norepinephrine.

While there is some indication that catecholamines are excreted due to stress, they are generally
released in relation to the overall activity level or performance level. In a review of catecholamine
response to various activities, Euler (14) reports that mental stress associated with ager, aggression,
or exhilaration will increase norepinephrine excretion while emotional states characterized by appre-
hension, discomfort, painful or unpleasant feelings, will increase epinephrine excretion. As an example
of what one may expect to find in measures of catecholamine levels, Euler and Lundberg (15) found that
urinary epinephrine levels were elevated in pilots as well as inexperienced passengers during one or one
and one-half hours of moderately stressful flights. The pilots also had elevated norepinephrine levels
while the passengers did not. Melton and Fiorica (16) found that both epinephrine and norepbrine
excretions were elevated during cross-country flights in private pilots with less than 100 hours flying
experience. However, the levels of excretion were not related to the length of flyl, g time. A more
recent study by Krahenbuhl, Marett and King (17) explored catecholamine production during various phases
of Air Force flying training in the T-37 jet aircraft. They found that the emergency procedure phase
whibch was given in a Link trainer was essentially rudtressful, but that both epinephrine and norepineph-
rine were significiantly ele'ated from control valuc.c during actual spin, solo and check flights. Here
again, the assessment using epinephrine appears to be more responsive than the use of norepinephrine as
an indicator.
Even though there does not appear to be any functional relationship between. the adrenal medulla and
the adrenal cortex, there is an interaction of catecholamine in steroids effects within the body.
Broverman, Klaiber and Vogel (18) have attempted to differentiate the effects of short-term versus
long-term stress relative to the interaction of catecholamine and steroids. Short-term stress is
hypothesized to facilitate performance on serially repetitive, overlearned tasks and to impair performance
on novel tasks requiring perceptual restructuring. Long-term stress is hypotherized to have the opposite
effects. They attempt to acco- for these finding by argusing that during short-term stress behavior
is dominated by the sympatheti vous system. However, with increasing exposure of the central nervous
 15

system to the stress-elicited adrenal hormiones, dominance shifts to the parasympathetic system causing
an overali depression of activity.
The Cardiac Indicators: The cardiac activity indicators heart rate (HR) and heart rate variability (WRV)
S \ have been used extensively to analyse inflight pilot activity probably because the data can be collected
without gross interference of flight activities. In addition, heart rate can be measured for specific
segments of performanre during rela;ively short time spans. There is no way to precisely determine
the relative contributions of any sement of behavior during a urine collection reriod aad thus, urine
analysis is confined to relatively gross edtimates of when perfcrmance decrement has occurred. In
addition, heart rate and heart rate variance appears to be more closely related to activity levels and
performance quality than does information on catecholawine production revealed by urine analysis. One
other advantage of heart rate activity is that data reduction can be almost immediately and easily
performed while urine analysis requires one or two days of chemical analysis in the laboratory under
fairly optimal conditions. A study by Bateman, et. al., (19) shows that heart rate for commercial
pilots on routine flights, upgrade training flights, and simulator flights are very similar and higher
than resting rates. However, basic training flights were found to be significantly higher. Heart rate
increased In respoase to speciff inflight stresses and when pilots were demonstrating maneuvers requiring
a high degree of skill. Opmeer aind Krol (20) found that increases in heart rate and decrease in heart
rate variance matched the predicted order of increasing difficulty of four phases of flight, namely
baseline, level flight, take-off, and approach. When pilots were required to fly realistic flight plans
in a simulator, the same relative changes were found. They found heart rate variance to be a more
sensitive measure than heart rate alone and they concluded that heart rate variance appeared to be more
related to cogi tive tasks where heart rate was more responsive to anxiety inducing tasks.
Roscoe (21) has demon.ti'ated that heart rate is a useful tool in evaluating pilot workload changes
created by nýw aircraft instrumentation and advanced control systems. Heart rate was found tn vary as
changes in weather conditions and different runways created more stressful landings. While these inflight
cardiac indicators have yielied some information on cognitive workload and stress levels experienced by
pilots, laboratory studies in which the stimulus presentations can be ura precisely controiled have
been much mo--. successful in relating these indices to performance and workload.

The normal re.ing heart rate exhibits a relatively large degree of beat to beat irregularity (HRV)
referred to as sinus arrythmia. Ettema and Zielhuis (22) fovrnd that sinus arrythmia was significantly
depressed and heart rate, blood pressurq and respiration rate were significantly increased as workload
increased. They concluded that this effect is due to a change in both the breathing pattern and a rise
in vagal tone and sympathetic nervous activity induced by the mental load. Boyce (23) found essentially
the same increase for heart rate and decrease in HRV Yor increasing mental loads. A series of studies
by Thackray (24) has shown 11RV to be a useful measure tor separatin3 rest peroids from mental work periods
on a variety of tasks. Using a two dimensional compensatory pursuit tracking task, he found that heart
rate variance along with heart rate, blink rate, respiration rate, respiration period variability, and
skin conductance were all capable of differentiating the rest period from the work periods. In a
simulated radar control task, heart rate variance was found to be higher for subjects reporting high
boredom. In addition, the performance of the subjects in the higher boredom group also significantly
declined over the rest period. This would suggest that HRV reflects a level of attentiveness wbhn is
related to overall performance capability.

A fairly comprehensive view of the relationship between cardiac indicators and performance has been
stated in the broader framework of arousal theory. It is known that the level ci performance quality is
related to the degree of arousal or activation level of the operator in te,-ms of an inverted, U-shaped
function which implies that an optimal level of activation will producs maximum performance capability.
This in turn is related to the reticular activating system which ift effect mediates the sleep/wakefullness
dimension. This of course is related to increasing levels of fatigue. Heart rate can be expected to
decrease as the subject's level of arousal falls or to inciesse as extra effort is put forth to stay
awake. The seemingly paradoxical increase in heart rate with fatigue is normally seen with physical
exertion as well, where heart rate continues to increase under vigorous exercise up to the point of the
collapse of the organism. Thus, the task demands of the systems operal.or job must be teken into account
if one is to predict the arousal level of a long duration flight. Corcoran (25) attempted to separate
the concept of arousal from task demand by requiring minimal activity from subjects during a 60-hour
period without sleep. In this case, both heart rute and performance on an unarousing, nonphysical,
30-minute vigilance task fell consistently. He argues that perfovAance will follow the inverted "U"
previously described with decreasing arousal, And that arousal wi.l fall with lack of sleep or increased
fatigue, but the effort to remain awake which is what is being miasured by physiological indicators will
be a function of task demand and subjective j.otivation to remela awake.

Extrapolating from these research findings, the filowing changes in heart rate and heart rate
variances can be predicted for long duration flights. First, heart rate and heart rate variance would
tend to increase with moderate levels of fatigue. With very high levels of fatigue, heart rate would be
expected to fall and heart rate variance to increase still further. We would also find that tasks which
created greater levels of arousal because of their complexity or the amount of concentration required
would be initially more resistant to fatigue effects. From this we can hypothesize that straight and
level periods of flight requiring minimal control input and instrument monitoring should show greater
performance decrement with fatigue than periods when maneuvers must be performed. Heart rate should be
higher ane heart rate variance should be lower as the arousal value of the task increases. Tasks
requiring maximum levels of information and concentration should shnw least performance decrement and
greatest heart rate increases and greatest heart rate variance decreases.

Thus, we appear to be at a point where the important pilotage aspects of information processing,
decision making, pattern recognition and so forth, are the important task variables and cardiac indicators
are one of the important measures of workload, fatigue, and stress relative to the man-machine system.
However, it would be a mistake to focus upon single physiologic variables. We have the present capability
to collect and evaluate multiple physiologic variables and weigh them by means of regression analysis sa

" " !1/

 16

as to investigate wheher maeaningful physiolodic profiles can identify specific reacLions to specific
aspects of workload, fatigue or stress.

While we are considering the present state of the art for both grour4-based and in1light physiologic
measurements, some exciting treakthroughs are on the horizon which may allow us to measure and utilize
cortical indicators of dynamic brain activities 4nc.uding decision making and Infor-acion processing.
But first, we shall explore how we arrived at the wosition that information processing activitiev relative
to required information input io a vitai consideration in tie evaluatL..n cf the aan-machine Interface.

REFERMCES
1. Waller, A.D. and De Decker, G. E. The physiological crat of work in various departments of "The
Times" printing house, J. of Physiol., 1919-9120, 53, vc-cvi.
2. Page, R. M. On supplanting the industrial fatigue concept, J. of Pusinessi, 1929, 2, pp. 137-153.
3. Pane, R. M. Measuring human energy cost in industry: a genaral gnide to the ]Iterature, Genetic

Psycho. Momo, 1932, 11, pp. 321-537.

4. Bitteruan, M. E. Fatigue defined a reduced efficiency, Amer. J. of Ps-chol., 1944, 57, pp. 569-573.

5. Selye, H. The physiclogy and pathology of e•posure to stress, Montreal: Acts, Inc., 1950.
6. Cameron, C. A theory of fatigue, Ergonomics, 1973, 16, pp. 633-648.

7. •airris, W. and O'Hanl, ,, J. F. A study of recovery functions in nan (U.S. Army Tech. Memorandum
10-72), Aberd.et Proving Ground, HD: Aberdeen Research and Development Center, Human Engineering
Laborator,. April, 1972 (UTIS No. AD-741 828).

8. Hartman, B. 0. and Cantrell, G. K. Psychomotor nouitoring for general efficiency, In Medical

Education :or National Defense Symposium on Biomedical Monitoring, Brooks APE, Texas: USAF School
of Aerospace Medicine, November 1964.

9. Guyton, A. C. Textbook of Medical Physiology (4th Ed.), ,'hiladelphla: W. B. Saunders, 1971.

10. Berkum, H. M., Bialek, H. M. Kern, R. P. aid Yagi, K. Experimental studies o' psychological
stress itnman, Psychol. Mono., 1962, 76, pp. 1-39.

11. Miller, R. G. Secretion of 17-.hydroxycortic. steroids (17-OUCS) in milltary aviators as an index

of -r-ponsrto -tress: a review, Aerosp. Mee, 1968, 39, pp. 498-501.

12. •incu t, C. and Hc,&Iand, H. Steri 4 d excrction in the stress of flying, J. of Aviation Med., 1943,
14q pp. !73-!97-o

13. Cannon, W. B. and delaPaz, D. Eotionsl stimulation of adrenal secretion, Amer. J. of Physiol.,
1911, 27, pp. 64-70.

14. Euler, U. S. Quantitation of stress by catecholamine analysis, Clinical Pharmacology and

Theropeutics, 1964, 5, np. 398-,04.

15 EuLer, ?j. S. an- Lundocrg, U. Effect of flying on the epinepnrine excret-:on in Air Force personnel,
.J. af Akp!. Phyvsinl., 1953, 6, pp. 551-555.

16. Melton, C. '. and Fiorici, V. Physiological responses of low-time private piots to cross-country
flying (FAA-AN-71-23), Oklahoma City, Oklahoma: FAA Civil. Aeronedical Insti.i;tes, April 1971.

'7. Krahenbuhl, G. S., Marett, J. R. and King, N. W. Catecholamine excretion in T-37 flight training,
Aviat, Space and Environ. Medicine, 1977, 46, pp. 405-408.

18. Brovernan, R., Klaiber, E. L., Vrge, W. and Kobayashi, Y. Short-term vs. Long-term effects of
adrenal hormuns on behaviors, Psychol, Bull., i97'4, 81, p?. 672-694.

19. Batemar, S. C., Goldsmith, X., Jackson, K. F., Ruffell Smith, H. P. and Mottodes, V. S. Heart
rate of training captains engaged in different activities, Aetosp. Med., 1970, 41, pp. 425-429.

20. Opmeer, C. H. J. M. and Krol, J. P. Towards an objectivs assessment of cockpit workload: 1.

physiological variables during different flight phases, Aeroep. Med., 1973, 44(5), pp. 527-532.

21. Roscoe, A. H. Use of pilot heart rate measurement in•flight evaluation, Clin. Med., 1976, 47(1),
pp. 86-90.

22. Ettema, J. H. and Zielhuis, R. L. Physiological parameters of mental load, Ergonomics, 1971,
14, pp. 137-144.

23. Boyce, P. R. Sinus arrhythmia as a measure of mental load, Ergonomics, 1074, 07, pp. 177-183.

24. Thackray, R. I. Patterns of physiological activity accompanying performance on a perccptual-motor

task (AN 69-8), Oklahoma City, Oklahoma: FAA, Office of Aviation Medicine, Civil Aeromedical
Institute, August 1975.

25. Corcoran, D. W. J. Changes in heart rate and performance as a result of loss of sleep, Brit. J.
of Psychol., 1964, 55, pp. 307-314.

IL.i- j
 17

SC*-Z INSIGHTS RILATIVI TO TER NAPMi(XIAIE OSTSM:

AN OVRVIEW I TOTiYCA OF RNIUUAO

r by
Richard E. McKenzie, th.D., 'eryce 0. Sartman, Ph.D.
Crew Technology Division
USAF School of Aerospace Medicine (APSC)
Brooks Air Yov'ce a-se,Texas 78235
USA

It is known that man-machine systems require certain kinds of operator skills and involve specific
kinds of tasks whether they are ground based, air-borne or in space. Viewing the development of aviation
from its infancy threoh current o"erational aircraft, airborne weapons systema, and space systems we
see a remarkable accelerated aevelopment of automation. With this development, there has been a shift
in the nature of the job performed by the man in this assembly of man and machine. In general, piloting
is really more like "machineamnship," with the number oi subsystems which the pilot muss, control, the
number of cockpit displays and other informational inputs as well as the increased commanications load,
all contributing to a tremendous increase in workload (1). The term workload is a somrehat ambiguous
concept that can be defined in many ways. We feel that workload encompasses he concepts of performance,
--
fatige, and stress, any one of which can be defined in ters of the other. Keeping in mind the pilot's
ýunctton as a ayotems monitor, wherein he initiates occasioxial commands to the system, we know that the
pilot will assume actual control of the system only at intervals against a background of activity at a
lewer level. Thus, we have a highly variable work rate situation and our initial concern is whether or
not the intervals of low activity might alter the efficiency of the operator when he is required to
assume coemand or exercise control over the system. In our first attack on this problem we used four
different workload levels from which the subjects went into a period of overload. The subjects in this
study ware used in a single session, matched group design. There were a total 3f 20 subjects, five in
each of 4 load levels. We obtained pronounced decrements in performance during overload after successively
lower work load levels. Unfortunately, in spite of the matching there were some differences between
Iroups of subjects on initial or baseline proficiency, therefore, we felt that this initial exploratory
study was not an adequate evaluation of the problem. What we needed was a repeated session design using
each subject as his own control. With this refinement in a follow-on experiment, we found no differences
in proficiency related to different base work rates. This confirmed British studies on speed, that is
signal rate stress, wherein the effects obtained are function only of the immediate operator load avd
a&:e independent of the characteristic of the preceding task load levels. So we found that the system
*perator works at a steady systematic rate independent of the more variable rate of signal onset. The
operator tends to ignore a rapid onset of signals, preceding in a methodic fashion to work on each
subtask as he gets to it. We liken this smooching function to the strategy of "queing" proposed by
Mill]r. In this strategy the operator assigns each new input to a kind of conceptual list of -eesponses
t- ",e made whan he gets to them. We looked for Miller's other adaptive strategies which hk called
"filtering" (ignoring some signals in order to process the remaining more effectively) and "two-handed
operation." Instances of filtering cold not be identified and two-handed operation occured only
infrequently; however, this does pose the question as to what conditions cause or promote the use of
such strategies (2).

This initial study was reported in 1961 but in the meanwhile we became involved in evaluating system
operator performance factors in the School of Aerospace Medicine'r space cabin simulator. In evaluating
the operator data, we reported the possible "energizing" effect of an intial high signal rate period had
on a subsequent period of very low signal rates. We alho felt that segnal rate might be a way .3f
manipulating both duty time and diurnal variables (3). With the idea that performance decrement was not
specifically time-anchored, but more of an immediate or instantaneous product related to signal rate, we
continued to gather data in the space cabin. The next series of flights explored a reversal of day/night
operating times. Here we again found that signal rate was a primary factor in performance, with marked
decrement at low signal retes below those of 119 per hour. This effect is attenuated by the dA3/night
cycle in that performance decrement is not as great when the low signal rate periods occurred during the
day (4).

At this point the requirement for evaluating special mission personnel including astronaut candidates
led to the interesting concept of task induced stress. Here competing tasks were used in a manner so as
to cause the operator to psychologically internalize the task stress, rather than to attribute his obvious
performance deteriment to the task itself. Aside from the problems of selection and evaluation the
results of our attempt to induce this k1nd of stress show that the competing task situation produced
significant task stress which could be used to access the relatilve adaptiveness of the individual. In
other wordr, the selected group was better able to perform and was, therefore, less susceptible to the
signal/noise ambibuity produced by the task and less bothered by the induced task stress (5).

A later overview of all of the SAM space cabin flights was aimed at evaluating information input as
a factor in crew performance. We might have called the study "signal rate revisited". In brief, we were
able to show that a constant, fairly high level of signal rate (500 signals per hour) resulted in a
rather remarkable increase in operator performance compared to low and/or variable work loads (6). Thus,
we have placed work/rest cycles, diurnal variations, etc., in the role of secondary or even more remote
factors in human performance, and we are left with information input as a critical variable. We can
infer r•ht any factor affecting man's ability to process information; that is, fatigue, drugs, stress
(physical and psychologic), etc., will be reflected Ln decremented performance.

One easily recognized confounding factor in the work load/performance area is fatigue. Ordinary
fatigue has ne,-er shown up as a significant factor in the space cabin studies. However, operational
requirements often impose aircrew problems related to change of sleep cycles, early awakening, and so
forth, in the face of increased pilotage demands. At times, tactical demands have raised the question
of preflight and/or in-flight pharmacologic support. One such demand indicated the use of preflight
sleep induction and inflight arousal via medications. Two research studies were designed to test the
effects such medication night have on performance. Here, workload vat an approximation L.fthe tactical
mission. Peeformanee was measurea using the *Alt1C-dimemsional Pursuit Test developed by the USt.7
School of Aekcspace Medi-ine years ogo as an aid to pilot selection. Preselection criteria and a
paychologic test battery were used as predictors. Heart rate and respiration oere also monitored. in
this instance, physiologic monitoring and psychologic testinog did not reveal nor predict any systematic
chrnges related to the drug treatments.

The drug treatments involved tte administration of eacobarbitAl (three grains) with the "in fligst"
administration af d-ampethamine (5 milligrain•) with appropriate controls. The results indicated a
hangover effect of three grain. of secobarbital seen at the start of the mission 10 ho4rs later and
Ptill prewent at the end of the miesfo- 12 hours taLer. The effects of d-campethacine are decreased in
individuals taking secobarbital (7).

A follow-on study using only l4 grains of secobarbital showed no apparent psychomotor hangover (8).
While these kind of in-laboratory studies are nededn, the cost of doing more than approximating the task
structure and workload requirements are usually prohibitive. However, we feel that the increased necessity
to consider the use of pharmacologic agents by sit-crew ambers will dictate more studies relating to
pilot performance. What the laboratory lacks is some comparative standard of laboratory task or task
system as it relates to acceptable performance standards for actual aircraft pilotage. In spite oi the
long history of laboratory testing, we still cannot answer the question "Will this particular drug, or
instrument, or device impair or enhance the pilot's ability to performa his required duties?" Theoretically,
it should be possible to state with scientific confidence that performance on laboratory task "X" at a
certain level indicates that pilotage under the experimental conditions being evaluated would be difficult,
dangerous or impossible. We have not yet developed such criteria which can be applied to the airborne
human. lowever, we tryl

Given the strong evidence of the critical nature of the relationship between information processing
ability and aircrew performance, perhaps we should make a dedicated effort to evaluata 'nformatlon
processing ability as our laboratory task and compare these results with simulated aircraft
1X" pilotage
performance.

RERERURCSS

1. Hartman, B. 0. and McKenzie, R. 1. The ctqmplex behavior simulator - a device for studying
psychologic problems in modern weapons systems, USAF School of Aerospace Medicine Report 61-9,
December 1960.

2. .Iartbn, 1. 0. and McKenzie, R. E. Systems operator proficiency: effects of speed stress on

overload performarce, USAF School of Aerospace Medicine Report 61-40, June 1961.

3. McKenzie, R. K.i Hartman, B. 0. and Welch, B. E. Observations in the SAN Two-Man Space Cabin
Simulator: III. System operator performance factors, Aerosp. Mad. 32:603, June 1961.

4. Hartman, B. 0., McKenzie, R E. and Welch, B. E. Performance effects in 17-day simulated space
flights, Aerosp. Med. 33:1098, 1962.

5. McKenzie, R. E. A systems task used in the stress testing of special mission personnel, Human
Factors, December 1965, pp. 585-590.

6. McKenzie, R. E. Crew performance as a factor of information input, Aerosp. Med. 44:3, 1971.

7. McKenzie, R. E. and Elliott, L. L. Effects of Pecobarbital and d-amphetamine on performance

during a simulated air mission, Aeroap, Med 36:8. i965.

8. Hlartman, B. 0. and McKenzie, R. E. Rangover effect of secobarbit&l on simulated pilotage

performance, Aeroap. Mad 39:11, 1966.
 19

AIRCM W Og.LOAD ASSESSMNIT TRC1RIOUBS

i• • by

Walter W. Wlerwille end Robert (,. Wi4iges

Virginia Polytechnic Institute and State university
"Blacksburg, Virginia USA
and

Samuel C. Snhiflett
Naval Air Test Center
Patuxent River, Maryland USA

Abstract
A clafsification schcue is presented which sumarizes a survey and analysis of aircrew workload
asaessment techniques relevant to inflight test and evaluation considerations, Two dimensions consisting
of universal operator behaviors and workload assessment methodologies were used in th-t classification
scheme. The universal operator behaviors were classified according to the 12rliner, Angell, and Shearer
(1964) categories Including perceptual, mediational, communication, and motor processes; wheres the work-
load assessment methodologles were cataloged into 28 procedures under the general categories of subjective
opinion, spare mental capacity, primary task, and phyriological measures. An applicability matrix based
on thia classification scheme is presented which SUMEarizes existing research on workload assessment
methodologies, and a bibliography of over 400 relevait references is provided as an appendix to this paper.
Procedures are described whereby this matrix can be used aa a guide for selecting candidate aircrew work-
load assessment measures for inflight evaluation. A brief overview of the various workload assessment
techniques is presented along with a set of critical criteria that need to be considered in evaluating the
feasibility of these measures for In-flight anvironments. It was concluded that no one single technique
can be racummended as the definitive measure of operator workload, but the resulting classification scheme
and applicability matrix can aid the Investigator in choosing among presently available techniques.

INTRODUCTION

One need only compare the cockpit of a modern jet fighter to its World War II predecessor to appreciate
the dramatic iacrease in cockpit complexity. Technological advances during the past 30 years have resulted
in sophisticated avionics and weapons delivery subsystems which are available to aid the aircrew in com-
pleting a specified mission. The ultimate mission success of today's modern fighter, however, still rests
on a common factor present in its World War II counterpart. This factor is the human operator. To be an
effective weapon, the modern fighter with all its advanced sensors And avionics must be compatible with
the capabilities and limitations of the aircrew operator.

During the design, development, and test and evalua.ion of any new aircraft, care must be taken that
the new system does not place unreasonable demands on the aircrew by overwhelming them with too much
information and too little tine to process that information. Such considerations are often characterized
as assessing the mental workload of the system operator.

When one reviews the research literature pertaining to mental workload, two conclusions are readily
apparent. Namely, there is no single, agreed upon definition of mental workload, and there is no single,
universal metric of it. Mental workload is a theoretical construct, and as such, might beat be defined
operationally. Clearly, it is related to factors such as operator stress and effort, but these concepts
also require operational definitions. Raising (1972) provides an excellent over-view of the difficulties
and complexities involved in defining and measuring workload.

Rather than provide a single definition, one must consider the various operational definitions used
in measeuring operator mental workload. The systems engineer, for example, may emphasize operational
definitions based on time available to perform a task. Psychologists tend to emphasize the information
procossing aspects of mental workload and operationally define it in terms of measures related to channe!
capacity and residual attention. Physiologists on the other hand, emphasize considerations of operator
stress and arousal.

Purpose
The impetus for this report stem•ed from a selective annotated bibliography of 83 references which
represent potential measurement techniques for assessment of operator workload in operational environments
(Schifiett, 1976). This annotated bibliography categorized the various methods in terms of general ref-
erences, system anaiyois, subjective techniques, psychomotor performance, information processing, physio-
logical measures, and combined methodologies. Schiflett concluded that the majority of the methods were
developed for use in the design stage of aircrew systems, thereby making them difficult and/or impractical
to use in the later stages of the operational test and evaluation environment.

This project was undertaken to provide a more comprehensive survey and analysis of the presently
available workload assessment methodologies and was specifically directed toward the flight test and
evaluation environment.

Appioach

*" To accomplish the goals of this project a comprehensive search of the scientific literature was con-
ductei including books, scientific journals, technical reports, and proceedings of technical meetings.
Computerized information retrieval, library searches and direct contacts with the scientific comunity
were used to locate relevant documents. Given the large pool of potential documents obtained by these
 20

combined search procedures, it %as necessary to Pdapt a set of leneral and specific criteria for inclusion
of a reference in the final bibliography of over 400 references appended to this paper. Details on the
search procedures as well as selection criteria are provided in Wierwille and Williges (1978).

following the selection of the appropriate workload literature, a user-oriented classification scheme
which combined workload metholo:Logy with universal aircrew behaviors waai used to generate a catalog of
presently available workload assesmuent techniques. Specifically, this paper provides a description of
this classification scheme and details the use of this scheme for the selection of potential measures of
workload. In addition, an overview of the resulting catalog of methodologies is presented.
CLASSIFICATION SCRDOM

An important prerequisite to developing a catalog of methodologies pertinent to operator workload in

flight environment is a comprehensive classification scheme. This scheme is necessary to form a basis
for selecting documents, to classify the citations listed in the bibliography, and to make possible the
construction of a usable analyisis catalog. Consequently, the resulting classification scheme is central
to a meaningful analysis of the workload literature applicable to aircrew considerations.

One dilemma that must be resolved in developing a classification 'icheme is that of providing a scheme
with a meaningful organization of existing workload assessment methodologies. A second dilmna centers
around providing a classification human operator behaviors which are related to aircrew performance so
that accurate implications can be drawn from -he vast amount of workload research that was not conducted
in a specitic aviation-related context. To solve these dilema, the selected scientific literature was
classified according to both the universal operator behaviors present in aircrew missionm as well as the
specific workload methodologies.

Universal Operator Behaviors

The range of operator behaviors and their taxonomy have been investigated for several years. These
behavior. 'ave been used to obtain an understanding of what functions an operator performs in a system
and as a b~ais for task analysis. One widely used listing of operator behaviors was developed by Berliner,
Angell, and Shearer (1964). This approach W3reak, operator behavior into four major processes (perceptual,
mediational, coemunication, and motor) as shown in Table 1. These four major processes are further sub-
divided into seven activities and then into 47 mutually exclusive operator behaviors. Because the terms
used in this scheme are orthogonal, this classification can be expected to yield good agreement among
K. investigators in determining specific behaviorb for a specific aircrew problem. Consequently, the Ber3 ner,
et al. (1964) approach was used to classify oper&tor behaviors in this report. To facilitate referencing
to this classification, a graduated numbering scheme as listed in Table 1 is used throughout.

Workload Methodolosies

The second dimension of classification is the specific list of available methodologies that are
potentially applicable to aircrew workload assessment. The literature on workload is so diverse that
categorization on the part of the reader of this literature is almost intuitive. It is, however, important
to select a zategorization which groups the various workload techniques in a logical way, so that conflicts
and discrepancies on workload concepts are minimized.
The taxonomy of workload methods that evolved from the documents reviewed was found to be particularly
useful and logical. This listing of methodologiew is presented in Table 2 along with a graduated numbering
designation. Basically, the various methods are grouped into four major categories (subjective opinion,
spare mental capacity, primary task assessment, physiological measures) which are further subdivided into
28 individual techniques.

Literature Classification

The resulting two-dimensional classification scheme used the n~umerical lesignations of workload
methodologies given in Table 2 with a subset of the universal operator behaviora given in Table 1. Early
in the classification of documents according to this two-dimensional analysis it became evident that the
scientific workload literature was addressed primarily to overall human performance as compared to specific,
detailed aspects of performance. Consequently, the literature reviewed couls be classified only according
to the four major processes and seven activities shown in Table 1 instead of the 47 mutually exclusive
behaviors. Even at this less-refined level of analysis, classification of the literature according to the
operator behaviors dimension appeared to be more subjective and unreliable than classification on the
second dimension of various workload methodologies.

Aplicability Matrix

Following the abstracting and classification of the selected documents, all the references were
summarized into a two-dimensional, applicability matrix which indicated the potential use of each of the
28 workload assessment techniques across the seven universal operator behaviors. A four-point rating scale
was used to represent the amount of positive research evidence supportin •he potential use of each work-
load technique for each operator behavior. These ratings included:

0: Workload method is unsuitable for assessing workload of the operator behavior cited. No
research or only negative research support.

1: Workload method is potentially suitable for assessing worklnad of operator behavior c:ited.
Some contradictory evidence exists; further research is needed.
 21

2: Workload method is suitable for assessing workload of the operator behavior cited.
No contradictory evidence exists; further research is needed.
3: Workload method is suitable for assessing workload of the operator behavior cited.
No contradictory evidence exists. Application is proven.
The complete applicability matrix resulting from this analysis is shown in Table 3.

I It should be noted that the ratings in Table 3 are based on all the research reviewed and, as such,
represent data collected in laboratory simulator, field, flight simulator, and flight test environments.
This was done to provide an overview of all the available data so so to suggest potentially applicable
techniques for the aircrew test and evaluation environment. Conceivably, none of the data used for a
particular rating was from the flight test environment. Table 3, therefore, is not totally suggestive
of overall ratings of research supporting the use of a technique in the flight test environment (research
of this type is, in fact, quite limited); rather it merely suggests a potentially applicable approach. To
complete the evaluations for posible selection of a workload assessment technique in the flight test area,
one must carefully consider the critical criteria for selection as well as the detailed evaluation of each
technique. Nevertheless, considerable judgment on the part of the authors was necessary in several cases
in arriving at a rating.
SELECTING A WORKLAD ASSESSMDIT HETHODOLOGY
The literature summarized in Table 3 could be used for a variety of purposes. For axample, cells
resulting in 0 or 1 ratings could suggest areas for additional methodological research. Of primary
importance, however, is the use of the classification scheme and resulting applicability matrix as an aid
in the selection of a workload assessment methodology for aircrew flight test and evaluations.
Stvpx in Selecting a Method
The information summarized in the applicability matrix presented in Table 3 as well as the complete
catalog description of workload estivati•u techniques presented by Wierwille and Williges (1978) can be
used as a guide in selecting a worklonQ assessment methodology in tl.*following six step procedure:
Step 1. Specify the aircrew problem for which mental workload to
'.. be evaluated.

Step 2: Perform a general task analysiF - ng specific operator behaviors.

Step 3: Using the workload method applicability matrix (Table 3), calculate workload methods
weighting. Rank order the methods.
Step 4: Select the first N methods in ranking. Study each of the N methods in the workload
methodology literature review.

Step 5: Select the method to be used.

Step 6: Read referenced documents and plan the workload measurement experiment.
The first step is to define the particular aircrew problems for which of the mission, and particular
aeircrew task. The second step is to relate the aircrew problem to the universal operator behavior dimension
shown in Table 1. This may be done by examining a task analysis which uses these terms or by having the
V investigator directly assess which behaviors are required of the airerew member during the task. With the
completion of Step 2, the aircrew p-,41lem dimension and the operator behavior dimension have been compressed
into a single dimension of specific operator behaviors which can be related to the seven universal operator
behaviors of the applicability matrix (Table 3).
To aid in the completio-, of Steps 2 and 3, a worksheet as presented in Table 4 is useful. The investi-
gator checks the top of the appropriate columns on the worksbeet of the universal operator behaviors which
are gernane to the particular aircrew mission as determiued by Steps 1 and 2. This essentially applies
equal weightings to the various operator behaviors chosen. Alternatively, each dimension can be weighted
according to the importance attributed to each operator behavior present in a particular mission. For
example, searching for and receivinp information (1.1), information processing (2.1), and communication
processes (3.) may be the central 'jperater behaviors in a particular mission. Rather than "checking"
these three dimenuioas on the worksheet, the invewrt4sg determine& that comwnicatiou processas are
perhaps twice as important to the mission as the other two. Conseq"u tly, communication processes are
-Cighted as 2 on the workaheet, and the other two operator behaviors are weighted as 1.
In step 3, the matr•i of Table 3 is used to determine the applicability rating of each workload
assesment technique. This is done by entering the applicable ratings from Table 3 on the worksheet and
adding the row of numbers for each workload technique. If the "check" approach is used, only the rating
values from Table 3 are added on the worksheet for each row and placed in the "SUM"column of the Volk-
sheet. If a weighting a&;roach is used, the weighting is multiplied by each applicability rating number
of the corresponding row; and, subsequently, the rows are added and placed in the "SUM" column of the
worksheet.
Step 4 involves the rank ordering from highest to lowest score for each workload technique. The
techniques with the highest scores are then selected. These N workload techniques are the most applicable
for the particular aircrew problem under study. Also as a part of Step 4, the investigator reads the work-
load catalog stmamry and the bibliography pertinent to each of the N particular techniques that had the
highest scores.
• ,22

It is difficult to state beforehand how large v should be. Most likely, it will be between 3 and 5
for most workload problems. However, judgment on the part of the investigator must determine the value of
N.
Once the Investigator has read the smmary of each technique, it should be possible to select the
technique that is to be used. This is Step 5. Obviously, judgment again plays a major role. More
specifically, practical aspects will have to be taken into consideration. Comparative difficulty of
Implementation, cost of the experiment, and ability to meet space, weight, and power requirements are some
cf t.he factors involved. A feasibility matrix for the selection of workload methodologies for in-flight
envirosments is found in Table 6.

Once the technique is selected, the investigator should obtain and read in detail the docments cited
in the bibliography relating to the specific technique. This will insure that available information is
used in conducting and carrying out the workload assessment experiment. Pitfalls and potential misappli-
cations might also be avoided.
SAMPLE APPLICATION

In this section the procedure for selection of one or more workload techniques will be demonstrated
by a sample problem. After a brief description of the environment associated with the sample problem,
the steps of the selection procedure will be described.

Backaround; The SS-3 Operator's Task

The $5-3 operator's position in the P-3 aircraft is one of control and usage of the aircraft'3 non-
sonar sensors. Several sensor systems are available to the operator, and the corresponding support equip-
ment for them is quite complex.
The SS-3 operator communicates with the remainder ot the crey using an open-line intercom that is
commom among the entire crew. Main cosmunications are with the TACCO (tactical aircraft coordination
officer) and the pilot; however, substantial listening to the "problem being worked" is also performed
by the SS-3 operator.
The SS-3 operator is responsible for ESM (electronic support measures). ESM is essentially the
passive evaluation and identification of incoming radar signals. The corresponding emitters may be ships,
aircraft, or ground-based radars and may be either friendly or hostile. C:dinarily, many radar signals
are impinging on the aircraft. With the aid (f the aircraft's central computer and the MPD (multipurpose
data display) at the SS-3 position, the operator must sort and evaluate them.

The SS-3 position also contains the MAD which is designed to provide precise location information on
partly or fully submerged vessels at short ranges. The system determines anomalies in the earth's
magnetic field resulting from large amounts of magnetic or paramagnetic material.
In some updated P-3 aircraft, the IRDS (infrared display system) has been added. The IRDS operates
at intermediate ranges between those of the radar and the KD. It provides a television-like raster-
scanned Image to the SS-operator. This sensor not only provides directional information on A target or
contact, but also provides an infrared (heat sensitive) picture showing details of platforms such as
supeistructures; rigging, antinnas, snorkels or periscopes. Positive identification of the contact or
target can often be made on the basis of these details.

It is important to recognize that the ESM system, the RADAR, the IRDS, sad the MAD are all tied into
the aircraft's central computer and appear in one form or another on the MPD before the SS-3 operator.
A large portion of the operator's workload involves updating the information, selecting modes, and per-
forming "evaluation" operations. The operator has before him, numerous comsnd and data entry pushbuttons
as well as a trackball (usable with either hand). The trackball allows the positioning of cursors and
symbols on the MPD so that specific coordinates may be inputted in what appears as an analog or digital
mode.

Application of the Procedure; An Example

Step 1. Aircrew workload problem statement. The IRDS is being installed in updated P-3 aircraft.
In making this addition, the mental workload of the S5-3 operator is going to be increased. It is desired,
therefore, to determine the workload of the S8-3 operator both with and without the addition of IRDS.
There are two points in a tactical intercept mission where highest workloads may be presumed to occur:
first, when the 5S-3 operator is attempting to transition from the radar to lIDS, and second, when the
operator is attempting to identify the target using the lIDS (having previously acquired it). For purposes
of 'explanation, discussion will be limited to the second high-workload condition.
When the IRDS is in use, the SS-3 operator controls the position of the sensor, vectors the pilot to
the target, and keeps the TALCO appraised of details becoming visible on the ilDS display, As soon as
positive identification is made by the SS-3 operator, the TACCO is informed.
Depending on the tactical situation the SS-3 operator performs ES4 duties, and keeps the MPD and
aircraft computer updated on the tactical situation. Furthermore the operator listens carefully to crew
comunications over the intercom.
During the same time period, if the aircraft does not contain an lIDS, the SS-3 operator vectors the
"pilot using the radar. Responsibility for positive Identification is then transferred to the cockpit crew.
The ESKM,NPD, computer, and communication tasks remain essentially the same for the SS-3 operator during
this time period.
 23

Stop 2. Determination of oper.tor behaviors. Since the higher workload* are likely to occur with
the IRDS present in the system, and since a comonn workload methodology should be used for both IRDS-
present and IRDS-absent cases, the IRDS-present case will be used to determine operator behavior@ and
weightings. In the IRDS-abeent case, only slight changes would occur, having to do with target identi-
fication.

In terms of the intercom task, the universal operator behaviors category (Table 1) is 3. Coanmnication
Processes. This is weighted with an importance of 4 on a scale of 0 to 5, where 0 is "of no consequence"
and 5 is "absolutely critical" to mission succesc. Whether the SS-3 operator Is verbally vectoring the
pilot or providing details on the identification to the TACCO, the intercom task is very important.

The IRDS aspect of the task involves calculations of vectoring information and visual discrimination
of details in the scene. These two aspects should receive a top rating of 5 because the mission is depend-
ent on the SS-3 operator a abilities at directing and rapid identification. The task consists of 1.2
Identifying Objects, Actions, and Events, and 2.2 Problem Solving and Decision Making. Continuous tracking
would also be performed. But, because slight errors in pointing the IRDS sensor would probably not harm
identification (as long as the target remained in the field of view), the behavior 4.2 Complex/Continuous
Motor Processes could be given a weighting of 3.

For most situations, the MAD would not yet have come into operation in the scenario, so it would be
assumed that it is uot part of the task. Similarly, while the radar might be operating, it would probably
only be used as a back-up (when the IRDS is operating and target acquisition has already been mwde).

The UKM system would continue to operate and to provide information on radar emitters in the area.
Under the assumption that the P-3 is not itself under attack, the SS-3 operator would relegate ESM tasks
to a lower priority. The examination of radar contracts would primarily involve 1.1 Searching for and
Receiving Information, 2.1 Information Processing, and 4.1 Simple/Discrete Motor Processes. This would
be given a priority rating of 2. Obviously, a much higher priority would be given to ESM (probably 5)
if the aircraft were under attack.

The SS-3 operator would also be performing data input duties to the MPD and computer to the extent
possible. However, these aspects would be of a bookkeeping and update nature, since pr:imary communication
would be via the intercom. Nevertheless, the operator would perform the task to ihe extent possible. It
involvesi l.1 Sevrching for and Recei,,ing Information, 2.1 Information Processing, 2.2 Problem Solving and
Decision, lskiig; 4.1 Simple/Discrete Motor Processes and 4.2 Complex/Continuous Motor Processes. When
the lIDS idejitificzation task is being performed, MPD and computer updating might have an importance
weighting of .2.

If the highest priority weighting stated above is used for each operator behavior, the weighting
would appear as shown in the first horizontal line of numbers of the worksheet Lor this example, as shown
in Table 5.

Step 3. Workload methods weightina and rank ordering. Having obtained the necessary universal
operator behavior weighting for the specific SS-3 operator workload problem, it becomes possil-le to compute
the relative veightings of workload tecn,.iques and to rank order them. This is done by multiplying each
number in the "Behavior Check ( ) or Weighting" row of Table 5 by the corresponding number in each row of
Table 3. Each individual product is then entered in Table 5 in the appropriate workload methodology row
and operator behavior column.

All products in each row are then added, and the sum is placed in the right hand column. The workload
methodclogies exhibiting the highest sums are the ones most applicable to the SS-3/IRDS problem.

Step 4. Selection of N techniques; study of the techniques. The results of the selection procedure
indicate that the following six techniques (ranked by numerical score with the highest first) are the most
appropriate for the S8-3 operator workload problem:

2.1.1 Task Analytic; Task Component, Time Sunmation

1.1 Opinion, Rating Scales

1.2 Opinion; Interviews and Questionnaires

2.2.1 Secondary Task; Arithemetic-Logic (Nonadaptive)

2.2.2 Secondary Task; Tracking (Nonadaptive)

4.1.8 Physiological; Pupillary Dilation

The initial selection of six techniques rather than some other number is arbitrary. Howevcr, techniques
having scores substantially below the highest ranking score are not likely to result in ac.turate, reliable
assessment of operator workload, because the corresponding techniques are not fully proven.

It is worth noting that small changes in the weightings of importance of the universal operator
behaviors would probably not have changed the outcome of the selection procedure up to this point. Most
likely the same six weightings sight have resulted in a different set of techniques being selected,
particularly for the fourth, fifth, and sixth ranks.

After studying the six techniques more carefully using section 4 of Wierwille and Williges (1978), it
should become possible to select one or possibly two to be implemented. As a means of carrying the example
through the advantages and disadvantages of the six techniques will be briefly reviewed.
24

The task-component, time summation technique is primarily analytical. However, it could be easf.ly
adapted to the TME efyiroumont by having SS-3 operators perform each will segment of each task separately.
These cauld be timed. Subsequently, time available could be determined from the mission scenario, sad
assessment of workload determined. The apparent drawbacks to such a technique are its complexity and the
fact that SS-3 operators may be capable of performing simultaneous tasks because of their high skill level.
The two opinion techniques are clearly applicable. It in probably true that the technical training
of 88-3 operators In sufficient to make then highly reliable judges of their own mental workload. The
investigator would have to present and specify the problem carefully so that the operators would have a
clear picture of what is expefted. Because of their high level of motivation, it is probable that accurate
assessment of maximum tolerable workload could be obtained.
The two secondary task techniques might also be applicable. Preference should probably be given to
the arithmetic-.ogic task, becauce the operator will have his left hand in use for the trackball and his
right hand in use fo, 'he =DS controller. Introduction of yet another manual control for tracking would
probably cause congebLion and severe intrusion. Even the arithmetic-logic task will to some degree cause
congestion because the operator is already using both hands, one foot, his voice, both ears, and his
vision (with at least two displays). If at all possible, the secondary task should in some way be inte-
grated into the present task through programing. Perhaps the ESK contacts, properly attended by the
SS-3 operator, could be scored as a secondary task. Since the operator would relegate this task to a low
priority anyway during the specified scenario, instructions to the operator would el-eady be similar to
his present method of operating.

The technique of pupil dilation is perhaps the least proven of the six; yet it holds promise. The
SS-3 operator's station in the aircraft is already somewhat Isolated. A curtain can be drawn around the
open side of the station, and the side window can be blocked. Consequently, ambient lighting could be
maintained constant. A small video camera could probably be installed at the upper and right hand corner
of the MPD. Alternatively, a commercially available, headmounted pupillography system could be used.

It should be noted that gathering of pupil dilation information is complicated by eyelid droop when
the observer becomes tired. Normally SS-3 operators are on duty for six to twelve hours. Care would
therefore have to be taken to fly short missions for data taking purposes.
Step 5. Workload method selection. It is believed that any of the initial six methods coald be used
to assess the SS-3 operator's workload. Final selection becomes a matter of ease of implementation, costs,
and other matters of feasibillity as indicated later in Table 55. On the basis of these factors it is most
likely that an opinion approach could be most rapidly and easily Implemented. It would therefore be the
recomended first choice. The task component, time sumation technique, using experimentally derived task,
element times would provide highly quantitative results. Therefore, it would be a good second choice.
However, a great deal of time and effort might have to go into the experiment and the data analysis.
Step 6. Study of documents; planning of experiment. Further study of documents referenced in the
appendix should make possible the construction of an opinion technique that has all the desired attributes
for the particular SS-3 probien under examination. The choice of rating scales, questionnaires, interviews
or soms combination thereof would have to be made.
Preliminary planning of the experiment should include a test of the technique on operators who would
not participate in the later data-taking session. These operators could aid in uncovering confusion terms
in questionnaires or rating scales, and in ironing out problems of terminology, instructions, and scoring.
The final experimental plan should be such that the experiment, when conducted, will yield statisti-
cally significant differences in experimental conditions if in fact there are differences. The most
prized result io significant differences in workload levels. Under these conditions, definite conclusions
can be drawn regarding workload.
OVUMVIEW OF WOULOAD TECHNIQUES
This 3action provides a brief overview of the various workload estimation techniques at the second
level of classification as shown in Table 2. Each procedure is described only in terms of its theory and
background, because of the brevity requirements of this paper. Hiwever, a complete description of method/
apparatus, areas of applications and examples, limitations, and suggesced RDT & E follow-up can be found
in Wierwille
critical and Williges
criteria (1978).
must first To provide
be considered in an
theoverall
inflightevaluation of these various techniques, certain
environment.

The In-Fliaht Aircrew Workload Problem

Whenever an attempt is made to measure workload in-flight, a number of practical considerations become
Important. These considerations deal primarily with the difficulties of introducing or adding anything
to the cockpit or crew station envirenment. These practical considerations go well beyond those involved
in ground simulation and have far-reaching ramifications on workload technique assessment.
Physical space. In most aircraft, crew positions are carefully designed to take maximum advantage
of the space available. This space is usually lipited by airframe and other design considerations in very
complex trade-offs. The introduction of any devict, having substantial *ite will compromise the efficiency
of the original design. Needed controls or displaya may be obscured or made inaccessible. Crew comfort
might also be sacrificed by reducing the already limited freedom of movement available. Therefore, for
most airborue situations, desirability and practicality of the workload measurement equipment increases as
the physical sise of the equipment decreases. Also, there is an upper-bound on tolerable size, %Mich for
sme situations might be as little an one-eighth cubic foot. Hov.var, a study of allowable size could
profitably be performed.
 Obviously, the physical size consideration is not as severe in groun4 simulatInn. Usually in this
camel a way can be found to "fit" another devioe into the simulation. Since the workload meacurement
apparatus need not be self-contained, supporting parts can be "hong" outside the simulsted crew station.

Portability and self-containment. In general, it would be desirable to assess workload using a

single small package that can be easily added to the crew station. Prototypo &ad operational aircraft
urally do not include power sources and telametry or recording equipment for optional equipment in tha
crew station. Furthermore, present practice would probably not permit modifications of operational air-
craft for powering and recording of .workload measurement. The assumption, therefore, must be umde that a
workload assessment system must be largely nelf-contained if il.is to be used ln-fli'/•ht.

Intrusion and safety. It is well known that many methods of workload measuremert tend to intrude on
t'aske at hand (primary tasks). An aspect of intrusion that must be considered separately because of its
importance is that of safety.

Certain types of flight operations are 1.•tthemselves critical. Take-off, londing, ejection, and
any other type of system failare, are examples of critical operations. Two tT i of safety-related
intrusion may possibly occur through introduction of workload measurement equ: "nt: obstruction and
distraction. Obstruction involves the problem of having an extra physical obj...c rithin the spuze needed
to deal with a critical operation. Distraction pertai,,e to the fact that the workload assessment may
draw the crew member's attention away from the critical situation. Unless backup crew stations Pre avail-
able, it may be inadvisable to assess workload of certain critical operations in flight except by S
posteriori techniques which by their nature do not intrude.

Data transmission or recording. It is one problem to design a feasible workload task for in-flight
use--it is yet another to score the task and analyze the results. There appcar to be three alternatives
in the area of data analysis:

1. Perform in-flight analysis and record the processed data output in concise form for later use;

2. Record or store the unprocessed data for later playback and analysis; and

3. Telemeter or otherwise trausmit unprocessed data to a ground station for recording or processing.

Experimental controls. A problem that may arise when performing in-flight experiments is that of
obtaining adequate experimenter controlq. The investigator or experimenter may not be on board when the
workload assessment procedures are conducted. Consequently, radio contact may have to suffice. In those
cases where the experimenter remains on the ground, workload assessment should be obtained by a system
that is procedurally simple to operate. Also this system should be as "'fudge-proof" as possible so that
the effects of biases of the aircrew members are minimized.

Workloe•d assessment integration. Because some modern aircraft incorporate computer graphic displays
with substantial computer capability, the possibility exits that certain workload assessment techniques
may te integrated into crew stations through software. Existing capabilities or near future capabilities
may be suih as to permit special modes of operation of standard displays and controls that would permit
workload assessment. Scoring might be accomolished by the on board computer and the results stored in
condensed form for pcst-flight readout. Not all methods of workload assessment may oe suitable for this
kind of integration; but initially, it appears that certain ones would be applicable A feasibility
study of the programing potential of new aircraft systems for workload measurement eppears to be a fruitful
area of research.

In-flight worLload assessment summary. In-flight measurement of workload represents a challange well
beyond that of ground simulation. Factors such as physical size, weight, intrusion zelated to safety,
portability, and experimental control becoma extremely important. Techniques that work well on the ground
may therefore prove infeasible for in-flight use, particularly during critical mission phases such as
take-off, landing, or subsystem failure (degraded mode). Nevertbeless, newer techniques are becoming
available that can eliminate or at least minimize the in-flight problems, principally the inclusion of
workload assessment as a software change in the aircraft's avionics system (using the ex~nting computers
and graphics capabilities), and the use of microprocessors in self-contained miniaturized modules that
perform all functions involved in workload assessment.

Table 6 provides a s9Ary of the seven critical criteria used to evaluate each of the various workload
measurement approaches for the in-flight environment. This matrix provides some perspective on the relative
feasibility of implementation, provided the measurement te'.hnique could otherwise be perfected. Details of
these feasibility considerations are provided in the descriptions of each method which follow.

WORKLOAD TECHNIQUES SIUMIARY

1. SUPJECTIVE OPINIONS

Subjective opinions are a commonly used measure of workload in flight test and evaluation. Often
this measure is used in conjunction with other indices to provide a broader basis for evaluation and
comparison. A variety of techniques exist toy gathering subjective opinions. These include psychomet-
rically defined rating scales, structured questionnaires with dichotomous or multiple choice responses,
open-end questionnaries, structured interviews, and unstructured interviews.

+
•In workload assessment applicaiAons, primarily two general approaches have been used. The more
systematic approach deals with the use of rating scale procedures for obtaining pilot opinions; whereas,
the second area deals with less structured approaches using a variety of interview and questionnaire
procedures. Often the terms rating scale and questionnaire are used somewhat interchangeably in the
scientific literature. For the purposes of this review, rating scales will be used for procedures which
 26

represent subject opinions gathered by devices with psychometric scaling properties, and questionnaires
used iu structured interviews will refer to procedures that are not based strictly on scaling considera-
tions. Consequently, questionnaires have been grouped with interviews for the purposes of this review.

1.1 Ratins Scales. Over the last twenty years much work has been dedicated to the development of rating
scales for assessing the handling qualities of aircraft. These scales ordinarily contain about ten
categories with descriptors that are not readily subject to confusion. The most widely used of these
scales is the Cooper-Harper scale (1969). It is accepted for use in handling qualities work and is
primarily used by teat pilots. The descriptors of this scale pertain to the "flyability" of an aircraft.
Even though the scale does contain some reference to workload the descriptors would have to be modified
for use in workload applications. If this Cooper-Harper scale were used for workload assessment in its
present form, the assumption must be made that handling difficulty and workload are directly related.
Such an assumption may well be unwarranted.

Recently, some research has been directed toward the development and evaluation of wrkload-opecific
rating scales. Comparisons have been made between the workload measurements obtained from rating scales
and those obtained from primary task performance, secondary tasks, occlusion, and physiological measures
(Hicks and Wierwille, in press). Specifically, the rating scale proved to be a sensitive measure of work-
load and resulted in little intrusion on the primary task. Additional research has been directed toward
developing a research-based, conjoint rating scale of workload for the F-18 aircraft (O'Conner and Suede,
1977; and Donnell and O'Conner, 1978) which vwa a direct outgrowth of the work of Helm (1975, 1976a, and
1976b).

With the ception of the conjoint measurement technique, most previous approaches have failed to
follow rigorous psychometric procedures in developing workload rating scales. Examples of the use of
ratings in this regard can be given both for flight simulator studies (e.g., Johannsen, 1976; Kreifeldt,
Parkin, and Rothschild, 1976; Murphy and Gurman, 1972; and Schultz, Newell, and Whitbeck, 1970) and
flight tests (e.g., Baker and Intano, 1974; Helm, 1975, 1976a; Lebacqz and Aiken, 1975; and Stackhouse,
1973).

1.2 Interviews ard Questionnaires. In contrast to the rather rigorous procedures available for the
development -i rating scales, the procedures used in interviews and questionnaires are not nearly as
structurvi. 4l.'1ation of these procedures to aircrew workload assessment range from completely open-
ended debriefi, tessions after flights (Soliday, 1965), to self-reporting logs of stressful activities
(Soutendam, 1977; Cantell and Hartman, 1967), to carefully chosen questionnaire items (Steininger 1977).
Recent work by Rohmert (1977) demorstrates procedures that can be e• ýIoyed in using questionnaire develop-
ment. This approach, called the "Ergonimic Job Description Questio,..aire," was developed specifically for
workload evaluations of air traffic control activities.

If questionnaires and interviews are used in an unstructured or e-en-ended way, care still needs to
be given to the appropriate topic areas and questions chosen for inclusion. If, on the othbx hand, struc-
tured resporses are used, the choice of response items (e.g., dichotomous or multiple choice) should be
constructed and tested much in the same manner as described for rating scales.

2. SPARE MNMAL CAPACITY

The largest body of research data dealing with the measurement of human operator worklcad is concerned
with the evaluation of the concispt of spare (residual or reserve) mental capacity. This contept is grounded
on the fundamental assumption of a single-channel, sampling model of the human operator (Knooles, 1963; and
Rolfe, 1973b). The approach assumes that an upper bound exists on the ability of the human -perator to
gather and process information. Spare mental capacity, then, is the difference between the total worktload
capacity of the operator and the capacity needed to perform the task. As spare mental capacity decreases,
the operator's workload increases until a point of overload is reached. At this point, the information
processing demands of the task exceed the operator's total workload capacity.

A variety of methods and procedures have been developed to measure, both directly and indirectly,
spare mental capacity. In addition, a great deal of laborptory research data exist on empirical tests of
various ramifications of the single-channel concepts. For example, data are available on the possibility
of multi-channel processing; procedures for switching attention among channels; various points of conflict
or bottlenecks in the human information processing channel; and variations in the upper limit of an indi-
vidual's mental workload capacity due to factors of stress, emotional state, fatigue, and effort. Much
of this human performance research is summarIzed by Kahneman (1973) and will not be reviewed at this time.

Essentially, three general methodological approaches have been advanced for measurement of workload
using the generalized spare mental capacity paradigm. These approaches include task analytic, secondary
task, and occlusion procedures. These methods are presented with the overall caution that even though the
underlying single-channel, sampling model assumptions of the human operator is a viable concept, it is not
a totally unequivocal hypothesis in terms of supporting data.

2.1 Task Analytic. Task analytic-methods assess spare mental capacity by using mathematical/theoretical
methods from systems engineering. The data base used in these techniq•_-. is most often obtained through
laboratory and simulation experiments rather than flight tests. Task analytic methods assume that all
task components, performed serially, require specified lengths of time to complete. As long as the actual
time available for overall completion exceeds the sum of theoretical time durations for performing the
task components, spare mental capacity exists. However, when the actual time available is insuff:oient,
* ~stress and task overloading occur. Task analytic methods consist of either task component/time susmation
computer models (Greening, 1978) or information-theoretic based procedures (Senders, 1970 and Baty, 1971).

2.2 Secondary Task. Most behavioral research approaches to estimating spare mental capacity have used
secondary task procedures. This appropcii provides the human operator with an additional (secondary) task
to be performed nnly when the main (primary) task has been fully attended. Performance on the secondary
 27

task thaoretically decreases as the attentional dmeand of the primary task increases. Secondary task
performance, then, becomes an indirect measure of operator workload.

Choice of the secondary task and procedures used to administer it become central issues in considering
this method of worklord assessment. Knowles (1963), for example, states that a viable secondary task for
workload assessment should not physically interfere %ith the primary task, require little of scoring.
Detailed reviews of the extensive fiterature on secondary tasks are provided by Rolfe (1973b) and Levine,
Ogden and Kisner (1978).

2.3 Occlusion. In many cases where workload is to be estimated the primary information input to the
(Senders, at
• i •.al., 1967).is visual. The ocilusion method of workload estimation can be used in such cases
operator

Occlusion is a time-shiring technique and as such is similar to the secondary task method. However,
in occlusion the time-sharing is accomplished by suppressing information inputs; that is, by giving the
are found time
operator visual information.
samplesandof Gallagher
in Farber Examples
(1972) and Hicks of automobile
and Wierwille press). research using this technique
(in driver

3. PRIMARY TASK MEASURES

It can be hypothesized that as the mental workload of a human operator increases, the performance
of that operator may change, ordinarily in the direction of degradation. If such a change does in fact
occur, its measurement would be an indication of increased workload. Tht.s hypothesis underlies the
primary task performance method of assessing workload.

The use of primary task measures as a means of assessing workload was not particularly popular
during the 1960's and early 1970's, because initial indications were that operators adapt to chauging
conditions, thereby holding performance constant. As Cooper and Harper (1969) put it, "In a specific
task, he (the pilot) is capable of attaining essentially the same performance for a wide range of vehicle
characteristics, at the expense of significant reductions in his capacity to assume other duties..
In this case they were referring to measures such as glide-slope error or flight path error in turbulence.

A somewhat more detailed examination of performance, however, might provide an indication of changes.
As a task becomes more difficult, an operator may sumon more effort, thereby holding performance in a
specific variable or set of variables constant. However, to maintain this performance, the operator may
have to modify his strategy.. By examining measures other than those involving system output, it may be
possible to detect this shift in strategy and thereby obtain a measure of workload.

Another concept in primary task measures was recently put forth by Albanese (1977). He suggests
that "successful mission completion" is a measure of workload. In this case, if an operator is able to
complete a mission successfully, there is no overload. On the other hand, if the operator cannot success-
fully complete the mission, an overload is presumed to have occurred. This rather broad concept has
distinct merit if an investigator is most concerned about the overload/nonoverlord dichotomy. Primary
task measures properly chosen, will indeed make assessment of mission success possible. Measures such
as landing touch-down performance, aiming performance, seeker lock-on, and number of procedural blunders,
can be used. Successful mission completion must be defined in terms of the measures.

3.1 Single Measures (Primary Task). A very large number of workload studies (Murphy, et al. 1974; Price,
1975; and Wickens and Kassel, 1977) have involved the use of one or more primary task measures, indi-
vidually on performance or as a precaution, while main interest was on some other method of assessing
workload (Kalsbeek and Sykes, 1967 and Trumbo, et al. 1967). In a few cases, the primary measures have
been taken specifically as a means of investigating level of workload (Brictson, 1974 a, b).

3.2 Multiple Measures (Primary Task). When a human operator performs a task in an actual system, several
subtasks are o.dinarily involvd. In such cases, a single measure of system perfor•.nce, such as error,
may be inadequate. Considerations such as stores usage, accelerations experienced, anJ operator percep-
tual style and strategy way become important. In other situations, it may be found that single measures
of the primary task do not exhibit adequate sensitivity to operator workload, because of operator
adaptivity. In cases such as these, multiple measures of primary task variables might be considered for
workload assessment. Essentially, the use of multiple measures provides a more complete picture of operator
behavior and operator/system performance.

To obtain the maximum information, the multiple measures should first be subjected to a combined
analysis and V';en subsequently to individual analysis where appropriate. Techniques that can be used for
the combinei analysis include multiple-regression analysis, correlation analysis, and various multivariate
analyses. These techniques provide a sound methodological approach for drawing valid conclusions regard-
ing system performance and workload.

Ordinarily, when using multiple measures, the additional measures used are not simply a greater
number of those used in single measure analysis. Measures such as RMS accelerations, number of control
(stick)
added reversals,
measures dominant et
(Kreifeldt, spectral frequencies,
al. 1976). Usually, and controlsuch
measures surface zero are
as these crossings
intendedareto typical
reflect ofstrategy
the
changes instead of performance scores, because performance scores may not change at lower operator workloads
totally different workload
measures have been taken which combine several
In several cases multiple
assessment techniques. Primary task measures may be combined with any ofdoesthe not
other a problem
methods:
present in
opinion,
spare mental capacity, and physiological of
measures these measures differ
that the units (Clement, 1976; O'Donnell and Spicuzza, 1975; and
The fact
Simmons ez al. 1976).
the analysis. The 3cores can be normalized or similarly treated in the analysis. In fact,
workload a few studies
techniques is
different
have been performed with the purpose of determining which of several
most sensitive. (See Wierwille and Hicks, in press.)
 28

3.3 Mathematical Modeling. Mathematical modeling of human operators in systaea has long been an area of
substantial interest to researchers. Interest began in the area of tracking and manual control system.
Subsequently, it has branched into areas of human operator decision processes, supervisory processes,
and team interactions.
Recently, a few of the researchers (Jex a:d Allen. 1970a; Baron and Lavison, 1975; Weverinke, 1976
and 1977; and Wickena and Gopher, 1977) involved in mode1a-ng have begun to examine the problem of operator
workload. This has usually been done as an attendant examination, with prime interest being in model
stimulus-response accuracy (Phatak, 1973; and Watson, 1972).
Other recent studies have departed from the describing function and optimal control models. Onstott
and Faulkner (1977) (also Faulkner and Onstott, 1977) worked with an urgency model of attention allocation.
Rouse (1977b) employed queuing theory to study human interaction with computers. Also Ravon and Gopher
(1977) postulate a model based on resource allocation. These models all have some bearing on workload;
however, results are preliminary.
4. PHYSIOLOGICAL MA9URES
One of the most widely researched methods of assessing operator workload is the use of physiological
measures. The physiological method generally involves the measurement and data processing of one or more
variables related to human physiological procesaes. The underlying concept in physiological monitoring
is as follows:
As operator workload changes, i.nvoluntary changes take place in the physio-
logical processes of the human body (body chemistry, nervous system activity,
circulatory or respiratory activity, etc.). Consequently, workload may be
assessed by the measurement and processing of the appropriate physiological
variables.

In many cases, there Is an underlying assumption that high workload levels are accompanied by
increased emotional stress. This stress is then measured by physiological recording and is related back
to workload. Stress in this case is assumed to act as an intermediate variable, causing physiological
changes.
In other cases, the underlying assumption involves changes in the state of "arousal." Arousal may
be considered as a state of preparedness of the body of level of activation of the human organism.
Roughly, one may think of arousal as the state of excitedness. Here again, the assumption is that mental
workload changes are accompanied by changes in arousal level that can be measured by appropriate physio-
logical monitoring equipment.
It is worth mentioning that physiological measures of workload do not require the underlying assump-
tion that the human operator is a single-cimannel sa'pliag device. Instead, a rather global defifitic j1
workload may be assumed, in which mental workload is considered a conglomerate of behaviors, similar .o
those enumerated by Berliner, et al. (1964).
'.1 Single Physiological Measures. The majority of work on physiological monitoring for the sake of
r sessing workload has been performed using single measures. In several cases data on more than one
nasure have been taken in a given experiment, but each measure has then been analyzed individually.
Such measures are considered here as single measures. Although perhaps not stated explicitly by the
Investigators, the objective of these studies has been to find a sin3le phyaiological measure that
accurately and reliably refltLts changes in operator mental workload.
In dealing with single mttsures (or any physiological measures for that matter), It must be recognIzed
tviat opermtor behavior other than mental workload may have an effect on the physiological measures.
Physical exertion, for example, may affect the measures being taken. Consequently, the range of potential
applicat.tanna of a measure may be severely limited by the confounding effect of operator behavior in areas
other Vwu mental workload. In specific terms, a measure that varies with physical work as well as
mental work for eximple can only be used if physical work is held constant or its manifestations on the
meacure are known and taken into account.
A :eview of each of the physiological measures as shown in Table 2 is beyond the sccpe of this paper.
lovever, the reader is referred to a discussion of combined physiological measures in the following
eectiou 4.2 and Wierwilile and Williges (1978).

"4.2 CombinedPhysiological Meauvres. Certain investigators have taken the point of view Lnat single
physiological measures may not provide adequate predictive information to allow assessment of workload.
They then proceed to analyze multple physiological measures in a combined analy3is in an effort to
bntter assess and predict workload. The multiple physiological measurement philesophy is the same
approach taken by researchers as was discussed in S9ction 3.2 for multiple primary task measures.
As with primary task measures, a common class of techniques can be applied. These include multiple-
regrmssion analysis, correlation analysis, and multivariate analysis. The rurpose in using toese
statistical techniques is to provide the best prediction and discrimination of worklodd ievels, .sased cn
the physiological measures at hand.
Several reports and papers have been published describing multiple featur'e extraction tecihni~uts
applied to multiple physiological measures (Spyker, at al. 1971; Stachouse, 1973; and Stackhouse, 1978).
SThe technique used is one of selecting a number of features for euch physiological measure and theor
performing a multiple regression. The best weighting of the most highly correlated features in then
•used in the prediction equation.

lul
 29

Storm, at Al. (1976) have performed analysis of multiple compounds in the urine believed correlated
with various aspects of aircrew-member stress. In general, while statistical analyses were not performed.
great care was taken in analyzing .rom a diagnostic point of view the directions and magnitudes of changes
in the levels of the compounds. Moreover, interactions were studied. A study of this type as well as
Storm and Hapenney (1976) gives a general inmi,-essioa of the physiological changes that occur when Air
Force aviators unuerso high workload/stress conditions for extended periods of time. Br~ctson, at al.
(1974) and McHugh, at al., (1974) studied the effects of high workload conditions on the performance of
naval aviators in Ligh-performance aircraft. The approach taken was one which combined stepwise multiple
regression of physiological, psychiatric, and performance measures in carrier landings.

The physiological measures in these studies on naval aviators were primarily those taken from blood
samples and included serum cholestrol, serum uric acid, blood lactate, and pyruvate. Changes in the
levels of biochemical measures were analyzed as a function of alterations in levels of workload, sleep,
performance, and mood.

4.3 Speech Pattern Analysis. Recently, there have been indications that inaudible changes take place in
speech when an individual is under stress. These changes generally are not detectable by an unaided
listener but can be elicited with the proper equipment, e.g., Psychological Stress Evaluator (PSE). The
underlying theory of the PSI has to ao with presence or absence of phyeiologirtal tremor of micro-tremor
in the human voice. In general this micro-tremor is present in an individual who is not under stress.
The tremor results in a frequency modulation effect of certain voice sounds that is only detectable with
equipment. The tremor and frequency modulation of the voice become suppressed when an individual is
under stress, such as when attempting to deceive law enforcement personnel (Kradz, 1974, and Dahb, 1974).

Older and Jenney (1975) analyzed voice commntcations of Skylab astronauts as a means of determining
situational stress. The scores obtained using a PSE were correlated with operational variables known to
represent varying degrees of stress. They found some statistically significant relationships, but
concluded that PSE usage was not sufficiently prodictive of wild stress as to warrant use in future
missions.

Simonov and Frolov (1977), following the work of Older and Jenney, liadertook to determine the emo-
tional state of cosmonauts and others via voice analysis. They indicated that the problem appears very
complex and that substantial further work is required.

Harris, et al (1977), taking a somewhat different approach, using automatic voice recognition and
synthesis equipment, showed that a verbal arithmetic task produced less decrement in concurrent manual
tracking than did a keyboard arithmetic task. They point out that automatic voice recognition equipment
introduces an additional source of error that may be dependent on task difficulty.

It seems clear that extreme stress can be measured by voice analysis. At this time, however, the
usefulness oZ voice analysis for either mild stress or mental workload is unclear. Several investigators
appear on the verge of analysis of voice in regard to workload, but results are not presently available.

CONCLUSIONS

This survey of the workload literature has shown that several approaches are potentially useful for
the aircrew workload problem, but no one single technique can be recommended as the definitive measure
of operator workload. Because of the multidimensionality of workload, it also appears unlikely that any
one single measure will ever suffice completely. Consequently, multiple measures i.•cluding the dimensions
of subjective opinions, spare mental capacity, primary tasks, and physiological correlate& should be
considered. The classification -cheme and applicability matrix developed in this paper should provide
the Investigator with an aid fo, .toosing among the presently available techniques.
RECOMMENDATIONS

This study of the workload literature has provided support for several recommendations, including
implications for future work. Four of the most prominent research recomendations are presented in
brief .

Computerized Information Retrieval System for Workload Literature

This study of the workload literature has been performed in a way that will allow computerizing of
the informatiou. The advantages of computerizing would be numerous. A user would be guided through
important citations based on the needs associated with a given aircrev, workload estimation problem.
More specifically, relevant references could be cross-filed according to:

1. The workload classification scheme.

2. Keyword or combinations of keywords (in title or in abstract).
3. Author or authors.
4. Workload category or subcategory.

If requested by the user, the system would also provide a narrative summary on the N workload
techniques. to provide broad necessary background should the user not already have it.

IL-provement of Rating Scales for Workload Estimation

The two major advantages of subjective opinion ratings are acceptance and lack of intrusion. Pilot
acceptance of opinion ratings has been good and is well documented in the handling quality domain.
Opinion ratings are geaierally not intrusive. However, with the exception of the conjoint measurement
technique, most previous approaches have failed to follow rigorous psychometric procedures in developing
workload rating scales. Additionally, several other limitations of ratings also need to be considered.

• 'L.•..... 1 . .. .. ..... .. ..... ... ,-- • •, , , •, • • •- . ... - . . -• . .• '• • • • • ,,- , :•• •... ... .. • • .. . . •:••J.••:
30

Adaptivity of the pilot, for example, represents a serious problem. Due to adaptivity, ratings may be
either too high or too low. A system that initially provides the impression of being aw•ward to use may
higher ratings than it Sobtain
should, because the crew ember adapts. Other problems ir.clude possible
omotional state, experience, and learning.
Given the wideapread use and general applicability of rating scales as a technique of workload
asaessment, it is eurprising that a rigorous workload rating scale has not been developed. Research is
needed to Aeteraine the underlying scaling dimensions of mental workload and to develop an interval type
metric characteristic of the conjoint measurement procedure. Recent approaches such as behaviorally
anchored response scalas (BARS) may be useful in this regard. Objective anchor points such as semantic
differential such as policy capturing might be applicable in determining the relative Importance of
various dimensions used in subjective estimates of workload. Research is also needed t,.compare the
utility of these various rating procedures and to specify the reliability and validity of the resulting
scales.
Comparison of Methods of WorkLoad Intimation
This literature review has shown that little work has been done on experimental comparison of work-
load estimation methods. To a great extent each research group in the workload estimation are tends to
advocate usually one or possibly two workload estimation techniques. One group advocates time-estimation.
another critical tracking tasks, and still others specific kinds of physiological measures. While all
this work is clearly important, particularly in regard to development, evaluation, and optimisation of
various techniques unauwared.
Hicks and Wierwille (in press) have recently addressed this problem on an initial basis. They
compared five diffevent (specific) workload techniques in a moving-base driving simulator. In',ludod in
their comparison were rating scales, primary task measures, secondary task measures, occlusion, and heart
rate variability. It was found that large differences in techuique senitivity existed when nperator
loading was adjusted under controlled conditions. SensitIvity in this context is defined as the
statistically rignificant differences in operator loading. High sensitivity low variance of the scores
about the means. Iu addition, it was determined that the degree of intrusion varied with the technique,
with sone being nonintrusive while others were highly intrusive.

A similar, more complete study needs to be performed for the aircrew workload estimatiou problem.
At present, the comparative sensitivity of aircrsw workload estimation techniques is unknown. Because
sensitivity has generally not been high, such a study is vital. Selection of a technique without
comparative information may yield results indicating that there is no change in aircrew workload for two
or more eifferent configurations when in fact there is a change. And, since an aircrew eamber's work-
load may already be high, failure to discriminte workload differences in a T&E situation may later
jeopardize mission success.

Update of Literature Search on Workload Evaluation Techniques

Workload evaluatio-, is at present a highly active research area. It is estimated that more than
one hundred resea'.chert, in the United States, Europe, and elsewhere are Iersed in workload research
at this time. Because of the forthcoming results and the extreme diversity of this work, the workload
search described here will need to be updated peridocially if it is to r~main current.

The updating of the search is very important since much of the work presently in progress has dire,:t
bearing on the aircrew worklord problem. More specifically, while much of the earlier research on work-
load was of an exploratory nature or involved development of concepts and constructs, more recent work
has tended toward ';he practical with applications to aircraft and other human-operator systems problems.
 31

APPENDIX

WORILOAD BIBLIOGRAPHY

Albanese, R. A. &Athe'atical analysis and computer sLtlation in military mission worklo.d assessment.
Proceedings of the AVARD Conference on Methods to Assess Workload, ACARD-CPP--216, April, 19i?, A13- -
A13-*6.

Allen, R. W., Jex, H. R., McRuer, D. T. and DP4arco, R. J. Alcohol effects on driving behavior and
performance in a car simulator. IEEE Transactions on Systems. Man, and Gybernetics, 1975, SUC-5,
498-505.

I
Ailport, D. A., Antonis, B. and Reynolds, P. On the division of attentioný a disproof of the single
channel hypothesis. quarterly Journal of Experimental Psychology, 1972, 24, 225-235.
kAlluisi, R. A. and Morgan, B. B., Jr. Effects of practice and work load on the performance of a code
transformation task (COTRAN). Moffett Field, California: National Aeronautics and Space Administration,
Contractor's Report NASA CR-1261, 1969.

Alluisi, E. A. and Morgan, B. B., Jr. Effects of sustained performance of' time-sharing a three-phase
code transformution task (3P-COTRAN). Perceptual and Motor Skills, 1971, 3S, 639-651.

Anderson, P. A. and Toivanen, '4.L. Effects of varying levels of autopilot assistance and workload on
pilot performance in the helicopter formation flight mode. Washington, D.C.: US Office of Haval Research,
JAWAIR 680610, March, 1970.

Armstrong, G. C., Same, D. D., McDowell, J. W. and Winter, F. J., Jr. Pilot factors for helicopter pre-
experimental phase. Randolph AFB, Texas: USAF Instrument Flight Center, IFC-TR-74-2, February, 1975.

Asiala, C. F. Advanced man-machine evaluation techuilquea. Paper presented at tim American Defense
Preparedness Association, Huntsville, Alanama, November 12-13, 1975.

Ascala, C. F., Loy, S. L. and Quinn, T. J. Digital simulation model for fighter pilot workload. St.
Lou!s, Missouri: McDonnell Aircraft Company, MDC A0058, September, 1969.

Auffr.•t, R., Saris, H., Berthoz, A. and Fatras, B. Estimate of the perceptive load by variability of
rate of heartbeat: Application to a piloting task. Le Travail Huaain, 1967, SC, 309-310.

Bahrick, H. P., Noble, M. and Fitts, P. N. Extra-task performance as a measure of learning a primary
task. Journal of Experimental Psychology, 1954, 48, 298-302.

Bainbridge, L. Forgotten allernatives in skill and work-load. Ergonomics, 1978, 21. 169-185.

Baker, D. L., and Intano, G. P. Helicopter yaw axis augmentation investigation - CDG-PFH-4. Randolph
AFB, Texa: USAVW Inetrume'-t flight Center, IFC Test Plan 74-'.1, December 1974.

Barnes, J. A. Use of eye-movement measures to establish design parameters for helicopter inscrument
panels. Proceedings of the AGARD Conference on Methods to Assess Workload, AGARD-CPF-216, April 1977,
A,-1!-A.3-8.

Baron, S., and Levison, W. H. An optimal control methodology for analyzing the effects of display
parameters on performance and workload in manual flight control. IEEE Transactions on Systems, Man, and
Cybernetics, 1975, SMC-5, 423-'30.

Bate, A. J., and Self, H. C, Effeutv of simulated task loading on side-looking radar target recognition.
Wright-Patterson Ab'Y, Ohio: Aerospace Medical Reiearch Laboratory, A14RL-TR-67-1l4, June 1968.

Baty, D. L. Human cransinformation rates during one-to-four axis tracking with a concurrent audio task.
Proceed ink of the 7th Annual NASA-University Conference on Manual Control, University of Southern
California, June 1971, 293-306 (NASA SP-281).

Beatty, J. Pupillometric measurement of cognitive workload. Proceedings of the 12th Annual NASA-
University Conference on Manual Control, University of Illinois, May 1976, 135-143 (NASA TH X-73 70).

Benson, A. J., Huddleston, J. H. F.. and Rolfe, J. N. A psychophysiological study of compensatory

tracking on a digital display. Human Factors, 1965, 7, 457-472.

Bergeron, H. P. Pilot response in combined control tasks. Human Factors, 1968, 10, 277-282.
Bergetroma, B., and Arnberg, P. Heart rate and performance in manual rtssile guidance. Perceptual and
Motor Skills, 1971, 32, 352-354.

Bayer, R. A study of pilot's workload in helicopter operation under simulated INC employing a forward
looking sensor. Proceedinks of AGARD Confersnce on Studies cn Pilot Workload, AGARD-CPP-217, April 1977,
B6-0-B6-10.

Bisseret, A. Analysis of mental processes involved in air traffic control. Ergonomics, 1971, 14.
565-570. _
32

Borg, G. Subjective aspects of physical and mental load. 8raonomics, 1976, 21, 215-220.
Boyre, P. Rt. Sinus arrhythlmia as a measure of mental load. Irgonomics, 1974, 17, 177-183.
Boylan, 1, i., A review of crew systems avnlytic methods. Seattle, Washington: Boeing Aerospace
Cemny, Dlb0-17525-1, January 1974 (a).
Boylan, I. J. Introduction to Blosing operator workload and workspace evaluation models. Seattle,
!7ashhintonm Boeing Aerospace Company, D140-17526-1, January, 1974. (b)
Bradashaw J. L. Load and pupillary changes In continuous processing tasks. British Journal of
PsocboloM, 1968, $9, 265-271.

Brlchcin, N. and RampeJwova, 0. Results of two kinds of mental load measurements. Ceakoolovenska
Psychologie, 1970, 14, 19-31. (In Czechoslovakian).

Brictoon, C. A. Pilot landing performance under high workload conditions. Proceedings of the AGARD
Conference on Simulation and Study of High Workload Operations, AGARD-CP-146, April, 1974, A7-1 - A7-10.
(a)
brictson, C. A. Pilot landing performance under high workload conditions. La Jolla, California: Dunlap
and Associates, Contract N00014-73-C-0053, April, 1974. (AD/A 001 802). (b)

Brictoon, C. A. Methods to assess pilot workload and other temporal indicators of pilot perforsance
effectiveness. ProceedinAs of AGARD Conference on Studies on Pilot Workload., AGARD-CPP-217, April, 1974,.
B10-1 - B10-7.

Brictson, C. A., McHugh, W. and Naitoh, P. Prediction of pilot performance: Biochemical and sleep-mood
correlates under high workload conditions. ft'oceedings of the AGARD Conference on Simulation and Study
of High Workload Operations, AGARD CP-146, April, 1974, A13-l - A13-10.

Brigham, F. R. COPTEC-A controller overload prediction technique. Loughborough, England: Human

Sciences and Advanced Technology Researcb Group, Department of Human Sciences, University of Technology,
HUSAT Memo. No. 67, February, 1974.

Broadbent, D. E. and HIeron, A. Effect• of a subsidiary task on performance involving imediate memory
by younger and older men. British Journal of Psychology, 1962, 53, 189-198.
Bromberger, R. A. LAMPS simulations: VI Pilot performance in the LAMPS simulator. Warminster,
Pennsylvania: Naval Air Development Center, NADC-76191-40, October, 1976.
Brown, E. L., Stone, G. and Pearce, W. E. Improving cockpits through flight crew wor"Joad measurement.
Paper presented at the 2nd Advanced Aircrev Display Symposium, U.S. Naval Air Test Center, Patuxent River,
Maryland, April 23-25, 1975. (Douglas Paper 6355).
Brown, I. D. Measuring the spare mental capacity of car drivers by a subsidiary auditory task.
Ertonomics, 1962, 5, 247-250.
Brown, I. D. A comparison of two subsidiary tasks used to measure fatigue in car drivers. Ergonomica,
1965, 8, 467-473.
Brown, I. D. Subjective and objective comparisons of successful and unsuccessful trainee drivers.
Ergonomics, 1966, 9, 49-56.

Brown, I. D. Dual task methods of assessing workload. Prgonýmcs, 1978, 21, 221-224.
Brown, I. D. and Poulton, F. C. Measuring the spare mental capacity of car drivers by a subsidiary task.
Ergonomics, 1961, 4, 35-40.
Burke, J. K. Us. of Eye Mark/Sony Videocorder System and related data reduction. Dallas, Texas: LTV
Aerospace, Vought Systems Division, VSD Report 2-57110/3R-3107, August, 1973.

Burke, J. K. In flight acquistion of task aequences and task times. Paper presented at the meeting of
the Aerospace Medical Association, Las Vegas, Nevada, May, 1977.

Cannings, R., Borland, R. C., Hill, L. E. and Nicholson, A. N. Pitch and formant analysis of the voice
in the investigation of pilot workload. Proceedings of the AGAJD Conference on Methods to Assess
Workload, AARD-CPP-216, April, 1977, A5-1 - A••i-0.

Cantell, C. K. and Hartman, B. 0. Application of time and workload analysis technics to transport
flyers. Brooks AFB, Texas: USAF School of Aviation Medicine, Technical Report SAK-TR-67-71, Augiat,
1967.
Caplan, R. D. and Jones, K. W. Effects of workload, role ambiguity, and type a personality on anxiety,
depression, and heart rate. Journal of Applied Psychology, 1975, 60, 713-719.

Casey, S. N., Brietmaier, V. A. and Meson, W. K. Cerebal activation and the placement of visual displays.
Warminster, Pennsylvani&: U.S. Naval Air Developmeut Center, NADC-77247-40, August, 1977.

LII
 33
J
Catlett, 1. L. Application of information theory concepts to study work complex versun operator action
time. Rad River Army Depot, Texarkana, Texas: USANC Intern Training Center, UIAINC-1IC-2-73-40, March,
1973. (AD 786 286).
Ceder, A. Driver's eye movements as related to attention in simulated traffic flow conditions. IHuman
Factors, 1977, ý.9, 571-581.

Chainova, L. D., Komarova, I. A. and Zonabend. F. I. Complex psychophymiological evaluation of the

readability of symbolic information (Komplekanaya Psikhofiziologichesdaya Ot sanka Chitammosti Znakovoy
Informataii). Vo.rosy Poikhologil, 19 16, 163-168. (Royal Aircraft Rsaablishment Library Transl%tiolt
1777).
Chiles, W. D. Objective methods for developing indices of pilot workload. Oklahoma City, Oklahoma:
Federal Aviation Administration, Civil Aeromadical Institute, FAA-AN-77-15, July, 1977.

Chilee, W. D. Objective methods. Tn A. H. Roscoe (1a.) Assessing pilot workload. ACARD-AG-233,

February, 1978.
Chiles, W. D. and Alluisi, 1. A. A review of methods for specifying operator or occupational workload.
Paper presented at XIXth International Congress of Applied Psychology, Munich, Germany, August, 1978.

Chu, Y. and Rouse, W. B. Optimal adaptive allocation of decision making responeiblity between human and
compute% in multi-task situations. Proceedins of the 1977 International Conference on Cybernetics and
So.iety, Washington, I.C., September, 1977.

Clark, W. I., Jr. and Armstrong, G. C. Three-cue helicopter flight director evaluation. Randolph APB,
Texas: USAF Instrument Flight Center, IFC TR 77-3, July, 1977.

Clazený:, ;. F. Investigating the upe of a moving map display and a horizontal situation indicator in
stimulated powered-lift short-haul operations. Proceedings of the 12th Annual NASA-University Conference
(NASA-TMX-73 70).

Clement. W. Annotated bib. aphy of procedures which assess primary task performance in some manner
as the basic element of a woj... oad Technical Report No. 1104-2, January, 1978.

Clement, W. F., Jex, H. R. and Graham, D. A manual control-display theory applied to instrument landings
of a jet transport. IEEE Transactions on Man-Machine Systems, 1968, )MS-9, 93-109. (a)

Clement, W. F., Jex, H. R. and Graham, D. Application of a systems analysis theory for manual control
displays tc,aircraft instrument landing. Pro.;eedings of the 4th Annual NASA-Universit Conference on
Manual Control, Ann Arbor, Michigan, March, 1968, 69-94. (NASA SP-192)'. (b)

Clement, W. P., McRuer, D. T. and Klein, R. Hi. Systematic manual cantrol display design. Proceedings
of the AGARD Conference on Guidance and Control Displays, AGARD-CP-96, 6-1 - 6-10, 1972.

Cliff, R. C. The effect.; of attention sharing in a dynamic dual-task environment. Proceedings of the
7th Annual NASA-University Conference on Manual Control. University of Southern California, June, 1971.
(NASA SP-281).

Cohen, S. I. and Silverman, A. J. Measurement of pilot mentel effort. Paris, Frnuce: North Atlantic
Treaty Organization, Advisory Group for Aeronautical Research and Developme'rt, Report 148, May, 1957.

Colle, H. A. and Detaio, J. C. The use of dual-ta:sk performance operating -urea to assess workload.
Paper nresented at the 1978 Review of Air Force Sponsored Basic Research in Flight and Technical
Training, U.S. Air Force Academy, Colorado Springs, Colorado, April, 1978.

Control-display pilot factors program. Randolph AFB, Texas: USAF lastrument Pilot Instructor School,
Instrument Evaluation Project NR 63-1, December, 1963.

Cooper. G. E. and Harper, R. P., Jr. The use of pilot rating in the evaluation of aircraft handling
qualities. Moffett Field, California: National Aeronautics and Space Administration, Ames Research
Center, NASA TN-D-5153, April, 1969.

Corkindale, M.G.G. A flight simultor study of missile coatrol petformance as a function of concurrent
workload. Proceedings of the AGARD Conference on Simulation and Study of High Workload Operations,
AGARD-CP-146, April, 1974, AS-1 - A5-6.

Corlett, 1. N. Cardiac arrhythmia as a field technique: Some comments on a receA',t symposium. krxoncnics,
1973, 16, 3-4.

Couluris, C. J., Ratner, R. S., Petracek, S. J., Wong, P. J. and Ketchel, J. N. Capacity and Productivity
implications on en route air traffic control automation, Washington, D.C.: Federal Aviarion Administra-
tion, FAA-RD-74-196, December, 1974. (AD/A 016 622).
Crabtree, M. S. Human factors evaluation of several control system Configurations, including workload
sharing with force wheel steering during approach and flare. Wright-Patterson AFl, Ohio: USAF Flight
Dynamics Laboratory, AFFDL-TR-75-43, April, 1975.

Crawford, B. M., Pearson, W. H. and Roffman, N. Multifunction iwitching and flight control workload.
. Paper presented at the 6th Psychology in the DoD Symposium, U.S. Air Force Academy, Colorado Springs,
Colorado, April, 1978.

t I'•d
rF
Faulkner, W. H. and Onstott, 1. D. Error rate information in attentico allocation pilot models. Pro-
ceedinas of the 13th NASA-University Conference on Manual Control, Massachusetts Inerltute of Technology,
June 15-17, 1S77, 72-78.

Finkelvan, J. N. and Glass, D. C. Reappraisal of the relationship betweer' ioise and human perfocuance by
miana of a subiiidinry task measure. Journal of Applied Psycholosy, 1970, 54, 211-213.

Firth, P. A. Psychological factors i-. ,icing the relationship between caru.e.rt arrhythami and mental
load, Ergonomics, 1973, 16, 5-16.

Flora, C. C., Kriechbaum. G.K.L. and . A flight investigation of asytew-s devieloped for reducing
W4.¼'
pilot workload and improving tracking ac.i.uitcy during noise-abatement landing appro.~ches. Moffett Field,
California: National Aeronautics and Space Administration, Ames Research Canter, Contractor's Report
NASA CR-1427, 1969.

Fowler, R. L., Williams, W. E.. Fowler, N. C. rendYo':ng, D. D. An investigation of the relationihip between

[Medical
operator performance and operator panel layout '\.- .-ontinuoua tasks. Wright-Patterson AFB, Chio: Aerospacu
Research Laboratory, AMRL-TR-68-17fJ, Dec;v~.rýer, 1968.
K. Gabriel, R. F., and Burrows, A. A. Improvinp s:,e-sharing perform-ince of pilots through training. Human
Factors, 1968, 10, 33-40.

V Gardner, R. N., Deltramo, J. S., and Krinsk~y, R. Pupillary chbingos during encoding,
of information. Perceptual and NoLor Skills,' 975, 41, 951-955.
storage, and retrieval

Gartner, W. 3. and Murphy, N. R. Pilot workload and fatigue: A critical survey of concepts and assessment
techniques. Moffett Field, California: National Aeroaautical and Space Administrat~on Ames Research Cen-I
ter, NASA TN D-8365, November, 1976.
Gaume, J. G. and White, R. T. Mental workload assessment. TI. Physiological correlates of mental work-
load: Report of three preliminary laboratory tests. St. Loule, Missouri: Mcl~nne~ll Douglas Corporation,
Report MDC J7023/01, December, 1975.

Geer, C. W. Navy manager's guide for the test and evaluation sectioNns of NTL-H-46855. Seattle Washington:
Boeing Aerospace Com..ary, Technical Report D194-11006-1, June, 1977. (b)

Geiselhart, R., Schiffler, R. J1.and Ivey, L. J. Itstudy of task loadings using a three-man crew on a
KC-135 aircraft. Wright-Patterson A15, Ohio: Aeronautical Systews Diviaiorn, ASD-TR-76-19, October, 1976.

KC-135 aricraft (GIANT BOOM). Wright-Patterson AFB, Ohio: Aeronautical Systems livision, ASD-TR-76-33,
April, 1977.

Gerathewohl, S. J. Definition and measurement of perceptual and mental workload in ali~r'zewa and operators
of Air Force weapon systems: A status report. In Higher menital functioning in operatiovil environments,.
North Atlantic Treaty Organization, April, 1976.

Gerathewohl, S. J., Brown, E. L., Burke, J. E., Kimball, K. A., Lowe, W. P., and Stackhouse, S. P. Inflight
measurement of pilot workload: A panel discussioa. Aviation. Space, and Environmental Medicine, June,
1978, 810-822.

Glenn, F. A., Streib, N. I., and Wherry, R. J., Jr. The human operator simulator volume VIII: Applica-
tions to assessment of operator loading, Willow Grove, Pennsylvania: Analytics, Technical Report 1233-A,
June, 1977.

Goerres, H.P. Subjective stress anssessment as a criterion for measuring the psychophysical workload on
pilots. Proceedings of the AGARD Conference on Studies on Pilot Workload, ACARD-CPP-217, April, 1977,
112-1_______-_ B12-8._______

Gopher, D. Eye movement patterns in selective listening tasks of focused attention. Perception and
Psychophysics, 1973, 14, 259-264.

Gopher, D., Navon, D., Chillag, N. and Dotan, H. Tracking in two dimensions as a function of dimwnsicn
priorities and tracking difficulty. Haifa, Israel: Technion-Isreel Institute of Technology, The Ce.nter

for Industrial Safety Research, Technical Report APOSR-77-2, December, 1977.

Gopher, D. and North, R. A. Manipulating the conditions of training in time-sharing performance.
Human.,Factors. 1977, 19, 583-593.

Graham, D. K. Transport air-plaike flight deck development survey and analysis: Report and recommndations.
Moffott Field, California: National Aeronautical and Space Administration, NASA CR-lAS, 121, January.
1977.

Green, R. and Flux, K.. Auditory communication and workload. Proceedings of the AGARD Coaference on
Methods to Assess Workload, AGAPD-CPP-216, April, 19777, AA-1 - A4-8.

Greening, C. P. Analysis of crew/cockpit models for advanced aircraft. China Lake, California: Naval
Weapons Center, NUC TP 6020, rebruary, 1978.

Gregoire, H. G. Is man the weakest link? Proceedings of AGARD Conference on Methods to Assess Workload,
AGARD-CPP-216, April, 1977, Al-l - Al-3.
 35

Gutmann, H. E., Easterling, R. G. and Webster, R. 0. The effects of flicker on performance as a function
of task-loadivg. Albuquerque, New Mexico: Sandia Laboratories, SC-174-72 0617, November, 1972.
Hacker, W. Determining the psychic workload: Present status and perspectives. Sociallstische
Arbeitswissencchaft, 1974, 18, 17-28. (In German).

I Hacker, W., Plath, H. E., Richter, P. and Zimmer, K. Internal representation of task structure and mental
t load of work: approaches and methods of assessment. Ergonomics, 1978, 187-194.
Hale, H. B., Anderson, C. A., Williams, E. W. and Tanne, E. Endocrine-metabolic effects of unusually long
or fr-quent flying missions in C-130E or C-1355 aircraft. Aerospace Medicine, 1968, 39, 561-570.

Hale, H. B., Hartman, 1. 0., Harris, D. A., Williams, E. W., Miranda, R. E., Hosenfeld, J. M. and Smith,
B. N. Physiologic stress during 50-hour double-crew missions in C-141 aircraft. Brooks AhB, Texas: USAF
School of Aerospace Medicine, SAN-TR-71-487, October, 1971.

Hale, H. B., Hartman, B. 0., Harris, D. A., Williams, E. W., Miranda, R. E., Hosenfeld, J. M. and Smith,
B. N. Physiologic stress during 50-hour double-crew missirms in C-141 aircraft. Aerospace Medicine, 1972,
43, 293-299.

Hale, H. B., Hartman, B. 0., Harris, D. A., Williams, E. W., Miranda, R. E. and Rosenfeld, J. N. Time
zone entraiment and flight stressors as interactants. Aerospace Medicine, 1972, 43, 1089-1094.

Hale, H. B., McNee, R. C., Ellis, J. P., Jr., Bollinger, R. Rt.and Hartman, B. 0. Endocrine-metabolic
indices of aircrew workload: An analysis across studies. Brooks AID, Texas: USAF School of Aerospace
Medicine, Unpublished report.
b. Hale, H. B., McNee, R. C., Ellis, J. P., Jr., Bollinger, ft.R. and Hartman, B. 0. Endocrine-setabolic
indicea of aircrew workload: An analysis acress studies. Proceedings of the AGARD Conference on Simala-
tion and Study of High Workload Operations, A(,AID-CP-146, April, 1974, A10-1 - A10-6.
Hale, H. B., Williams, 9. W., Smith, B. N., and Melton, C. E., Jr. Excretion patterns of air traffic
controllers. Aerospace Medicine, 1971, 42, 127-138.

Hall, T. J., Passey, G. E. and Meighan, T. H. Performance of vigilance and monitoring tasks as a function
of workload. Wright-Patterson AFB, Ohio: Aerospace Medical Research Laboratories, ANRL-TR-65-22, March,
1965.

Hamilton, P. Process entropy and cognitive control: mental load in internalized thought processes.
Position paper prepared for NATO Symposium on Mental Workload. Mati, Greece, September, 1977.

Harris, D. A., Pegram, G. V. and Hartman, B. 0. Performance and fatigue in experimental double-crew
transport missions. Aerospace Medicine, 1971, 42, 980-986.

Harris, S. D., North, R. A. and Owens, J. N. A system for the assessment of human performance in con-
current verbal and manual control tasks. Paper presented at the 7th Annral Meeting of the National
Conference on the Use of On-Line Computers In Psychology, Washington, D.C., November 9, 1977.

Hart, S. C. A cognitive model of time perception. Paper presented at the 56th Annual Meeting of the
Western Psychological Association, Los Angeles, California, April, 1976.

Hart, S. G. and McPherson, D. Airline pilot time estimation during concurrent activit7 including
simulated flight. Paper presented at the 47th Annual Meeting of the Aerospace Medical Association,
Bal Harbour, Florida, May, 1976.

Hart, S. G., McPherson, D., Kriefeldt, J. and Wemps, T. E. Haltiple curved descending approaches and
the air traffic control problem. Moffett Field, California: National Aeronautical and Spa-ie Adminis-
tration, Ames Research Center, NASA TM-78, 430, August, 1977.

Hart, S. G. and Simpson, C. A. Effects of linguistic reduncency on synthesized cockpit warning message
comprehension and concurrent time estimation. Proceedings of the 12th Annual NASA-t.iversity Conference on
Manual Control, University of Illinois, May, 1976. (NASA. TM X-73 70).

Hartman, B. 0., Hale, H. S. and Johnson, W. A. Fatigue in FI-111 crew-smembers. Aerospace Medicine, 1974,
45, 1026-1029.

Helander. M. G. Physiological reactions of drivers as indicators of road traffic demand. In Driver

performance studies: Trenaportation Research Record 530. Washington, D.C.: U.S. Transportation Research
Board, Technical Report TRB/TRR-530, 1975, 1-17.

Helm, W. R. Human factors test and evaluation, functional description inventory as a test and evaluation
tool development and initial validation study. Volume I and I1. Patuxent River, Marylai.d: U.S. Naval
Air Test Center, SY-77R-75, September, 1975.

Helm, W. R. Human factors evaluation

_r__lsne:
of model P-3C UPDATE I_ Third interim report. Patuxent
River, Maryland: U.S. Naval Air Test Center, SY-122R-75, February, 1976(a).

Helm, W. R. Function descripticn inventory as a human factors test and evaluation tool: An empirical
validation study. Patuxent River, Maryland: U.S. Naval Air Test Center, SY-127R-76, July, 1976(b).

Hilm, W. R. The application of computer aic I evaluative techniques to system test and evaluation.
Proceedings of the 21st Annual Meetlna of th :,uman Factors Society, San Francisco, California, October,
1978, ')-!-94.
 36

Cross, K. D. and Cavallero, F. R. Utility of the .- rtical contact analog display for carrier lavdings -
a dignostic evaluation. Proceedings of the AGARD Conference on Guidance and Control Displans,
AGARD-CP-96, 21-1 - 21-11, 1972.
Curry, R. E. Position paper prepared for NATO Symposium on Mental Workload. Kati, Greece, September,
1977.
Dahm, A. E. Study of the field use of the psychological stress evaluator. Dektor Counterintelligence
and Security, Inc., Springfield, Virgins, Unpublished manuscript, 1974.

Damns, D. and Wickens, C. A quasi-linear control theory analysis of time-sharing skills. Proceedings
of the 13th Anuual NASA-University Conference on Manual Control, Massachusetts Institute of Techonology,
June 15-17, 1977. (a)

Dames, D. L. and Wickens, C. D. Dual-task perforu:nce and the lick-Rlyman law of choice reaction time.
Journal of Motor Behavior, 1977, 9, 209-215. (b)

Danev, S. G. and Wartna, G. F. Information load and time stress: Some psychophysiological consequences.
TNO-Nieuws, 1970, 25 389-395.

Daniel, J. Newer approaches to a research of mental load. Proceedings of the 2nd meeting of Psychologists
from the Danubian Countries, Smolenice, Czechoslovakia, September, 1970.

Daniels, A. F. Crew workload sharing assessment in all-wheather, low-level strike aircraft. In Problems
of the Cockpit Environment, ACARD Conference Proceedings No. 55, March, 1970. (AD 705 369).

Defayole, M., Dinand, J. P., and Gentil, M. T. Averaged evoked potentials in relation to attitude,
mental load. and intelligence. In W. T. Singleton, J. G. Fox, and D. Whitfield (Ed$.) Measurement of
man at work. London: Taylor and Francis, 1973, 81-91.
Dick, A. 0. and Bailey, G. A comparison between oculometer data and pilot opinion on the usefulness of
instruments during landing. Rochester, New York: University of Rochester, Center for Visual Science,
Technical Report No. 3-76, 1976.

Dick, A. 0., Brown, J. L. and Bailey, G. Statistical evaluation of control inputs and eye movements in
the use of instrument clusters during aircraft landing. Rochester, New York: University of Rochester,
Center for Visal Science, Technical Report 4-76, 1976.

Donnell, M. L. and O'Connor, M. F. The application of decision analytic techniques to the test and
evaluation phase of the acquisition of a major air system: Phase II. McLean, Virginia: Decisions and
Designs, Technical Report n. 78-3-25, April, 1978.

Dougherty, D. J., Emery, J. H. and Curtin, J. C. Comparison of perceptual workload in flying standard
instrumentation and the contact analog vertical display. Washington, D.C.: Joint Army Navy Aircraft
Instrumentation Research, D228-421-019, December, 1964.
FI
Drennen, T. C., Curtin, J. G. and Warner, H. D. Manual control in target tracking tasks Ls a function
of control type, task loading, and vibration. St. Louis, Missouri: McDonnell Douglas Corporation, MD-CE
1713, August, 1977.

Dunn, R. S., Gilson, R. D. and Sun, P. A simulator study of helicopter pilot workload reduction using
a tactile display. Proceedings of the 12th Annual NASA-University Conference -k Manual Control, I 1
University of Illinois, May, 1976. (NASA TMX-73 70).

Dyer, R. F., Matthews, J. J., Wright, C. E. and Yudawitch, K. L. Questionnaire construction manual.
Fort Hood, Texas: U.S. Army Research Institute for the Behavioral and Social Sciences, Field Unit,
Technical Report P-77-1, July, 1976.

Edson, R. K. The Dektor psychological stress evaluator (voice s5Lit a'nalyzer) as a research instrument.
Unpublished master's thesis, National Graduate University, April, 1976.

Ellis, C. A. Subjective assessment. In A. H. Roscoe (Ed.) Assessing pilot workload. ACARD-AC-233,

February, 19,.;.

Enstrom, K. D. and Rouse, W. B. Telling a computer how a human has allocated his attention between
control and monitoring tasks. Proceedings of the 12th Annual NASA-University Conference on Menial
Control, University of Illinois, May, 1976, 104-123. (NASA TI X-73, 70).

Ephrath, A. R. Pilot performance in zero-visibility precasiot approach (NASA CR-137759). D.ctoral

dissertation, Massachusetts Insitute of Technology, 1975.

Ephrath, A. R. A novel approach to the cross-adaptive auxiliary task. Proceedings of the 12th Annual
NASA-University Conference on Manual Control, University of Illfnois, May, 1976, 63-71, (NASA TMX-73, 170).

Ettemsa, J. H. Blood pressure changes during mental load experiments in man. Psychotherap, and
Psychosomatics, 1969, 17, 191-195.

I •on
Farber, E. and Gallagher, V.
driving task difficulty.
Attentional demands as a measure of the influence of visibility conditions
Highway Research Record, 1972, 414, 1-5.

LI
 37

Henry, P. U., Davis, T. Q., nsgelkan, S. J., Triabvasser, J. H. and Lancaster, M. C. Alcohol-induced
4 performance decrements assessed by two link trainer tasks using experienced pilots. Aerospace Medici..e,
1W74, 45, 1180-1189.

Ease, K. A. and Teichgraber, W. M. Error quantiaetion effects in compensatory tracking tasks. IEEE
Transactions on Syatems, Man, and Cybernetics, 1974, SMC-4, 343-349.

Hickok, J. H. Grip pressure as a measure of task difficulty in compensatory tracking tasks. Master's
thesis, Naval Postgraduate School, Monterey, California, September, 1973.

Hicks, T. G. and Wierwille, W. W. Comparison of five mental workload assessment procedures in a moving-
base driving simulator.. Human Factors, in press.

Hulgendorf, 1. L. Information processing, practice, and spare capacity. Australian Journal of Psychology,
1967, 1__9, 2-251.

Hoffman, E. R. and Joubert, P. N. The effect of changes in some vehicle handling variables on driver
steering performance. Hunma Factors, 1966, 8, 245-26'.

Bolden, F. M., Rogers, D. S. and Roplogle, C. R. Simulation of high workload opera.tions in air to air
combat. Proceedings of the AGARD Conference on Simulation and Study of High Workload Operations, AGARD-
CP-146, April, 1974, A6-1 - A6-4.

Holland, M. K. and Tarlov, G. Blinking and mental load. Psychological Reports, 1972, 31, 119-127.

Hopkin, V. D. Mental workload measurement in air traffic control. Position paper prepared for NATO
Symposium on Mental Workload. Mati, Greece, September, 1977.

Hosman, R.J.A.W. Pilot's trac!ing behavior under additional workload. Delft, The Netherlands: Delft
University of Technology, Department of Aeronautical Engineering, Report UTH-199, June, 1975.

Howitt, J. S. Flight-deck workload studies in civil transport aircraft. In AGARD measure of aircrev
performance. Report No. N70-19780, December, 1969.

Huddleston, H. F. and Wilson, R. V. An evaluation of the usefulness of four secondary tasks in assessing
the effect of a lag in simulated aircraft dynamics. Ergonomics, 1971, 14, 371-380.

Hughes, H. M., Hartman, B. 0., Garcia, R. and Lozano, P. Systems simulation: A global approach to aircrew
workload. Proceedings of the AGARD Conference on Simulation and Study of High Workload Operations, AGARD-
CP-146, April, 1974, Al-l - Al-14.

Jahns, D. W. Operator workload: What is iL and Low should it be measured? In K. D. Cross and J. J.
McGrath (Edo.) Crew System Design. Santa Barbara, California: Anacapa Sciences, July, 1973. (a)

Jahno, D. W. A concept of operator workload in manual vehicle operations. Meckenheim, Germany:

Forschungsinstitufuer Anthropotechnik, Report No. 14, 1973. (b)

Jenney, L. L., Older, H. J. and Cameron, B. J. Measurement of operator workload in an information

processing task. Washington, D.C.: National Aeronautics and Space Administration, Contractor's Report
NASA CR-2150, December, 1972.

Jennings, A. E. and Chiles, W. D. An investigation of time-sharing ability as a factor in complex per-

formance. Human Factors, 1977, 19, 535-547.

Jea, H. R. Two applications of a critical-instability task to secondary workload research. IEE!

Transactions on Human ?actors in Electronics, 1967, HF1-8, 279-282.

Jew, H. R. and Allen, R. W. Research on a new human dynamic response test battery. Part I. Test
development and validation. Proceedings of the 6th Annual NASA-University Conference on Manual Control,
Wright-Patterson AFB, Ohio, April, 1970. (a)

Jex, H. R. ani Allen, R. W. Research on a ne,# humen dynamic response test battery. Part II. P'3ycho-
physiological correlates. Proceedings of the 6th Annual NASA-University Conference on Manual Control,
Wright-Patterson AFB, Ohio, April, 1970. 1b)

Jex, H. R. and Cleent, W. F. Defining and measuring perceptual-motor load in manual control tasks.
Hawthorne, Calif.: Systems Technalogy, Inc., Report No. 1104-1, March, 1978.

Jex, H. R., Jewell, W. F. and Allen, R. W. DcN'elopment of the dual-axis and cross-coupled critical
tasks. Proceedings of the Eighth Annual Conference on Manual Control, University of Michigan, Ann Arbor,
Michigan, May 1972, pp. 525-552.

Jew, H. R., McDonnell, J. D. and Phatak, A. V. A "critical" tracking task for man-machine research related
to operator's effective delay time. Proceedings of the 2nd Annual MASA-Univers..ty Conference on Manual
* Control, Massachusetts Institute of Technology, March, 1966, 361-377. (NASA-SP-128).

Johannsen, C. Nebenaufgaben ala Beanspruchungsmesnserfahren in 7ahrzeugfuhrungsaufSaben. Zeitschrift fur

Arbeitswissenschaft, 1976, 30, 45-50.

A! Johannsen, C. Position paper on mental workload. Prepared for NATO Symposium on Mental Workload. Mati,
Greece, September, 1977.
38

Johannsen, G., Pfendler, C. and Stein, W. Human performance and workload in simulated landing-approaches
with autopilot-failures. In T. B. Sheridan and C. Johannsen (Ed..) Monitoring behavior and supervisory
control. New York: Plenum, 1976, 83-95.

Jones, B. C., Jr. and Schuster, D. R. Design and development of an adaptive, auditory, and distractive
stressor. IIEE Transactions on Man-Machine Systems, 1970, WS-11/3, 161-163.

Kahneman, D., Beatty, J. and Pollack, 1. Perceptual deficit during a mental task. Science, 1967, 157,
218-219.
Kahneman, D., Tureky, B., Shapiro, D. and Crider, A. Pupillary, heart rate, and skin resistance changes
during a mental task. Journal of Experimental Psychology, 1969, 79, 164-167.

Kalebeek, J.W.H. Objective measurement of mental workload: Possible applications to the flight task.
Proceeding, of the 55th AGARD Conference, Amsterdam, The Netherlands, 1968, 4.1 - 4.6.

Kalabeek, J.W.H. Measurement of mental workload and of acceptable load: Possible applications in industry.
International Jout-nal of Production Research, 1969, 2, 33-45.

Kalsbeek, J.W.H. Standards of acceptable load in ATC tasks. Ergonomic, 1971, 14, 641-650.

Kalsbeek, J.W.H. Sinus arrhythnia and the dual task method in measuring mental load. In W. T. Singleton,
J. G. Fox, and D. Whitfield (Eds.) Measurement of Man at Work. London: Taylor and Francis, 1973, lCl-113.
(a)
Kalsbeek, J.W.H. and Sykes, R. N. (, 'ective measurement of mental load. Psychologica, 1967, 27, 253-2bl.

Kantowitz, B. H. and Knight, J. L * Jr. Testing tapping time-sharing, I1. Auditory secondary task.
Acta Psychologica, 1976, 40, 343-362.

Kantov•ltz, B. H. and Knight, J. L., Jr. Testing tapping time-sharing: attention demands of movement
amplitude and target width. In G. E. Stelmach (Ed.) Infoimnation Processing in Motor Learning and Control.
New York: Academic, 1977.

Kelley, C. R. Design applications of adaptive (self-adjusting) simulators. Proceedings of the 2nd Annual
NASA-University Conference on Manual Control, Massachusetts Institute of Technology, March, 19',6, 379-401.
(NASA SP-128).

Kelley, C. R. and Wargo, M. J. Cross-adaptive operator loading tasks. Human Factors, 1967, 9, 395-404.

Kennedy, J. P. Time-sharing effects on pilot tracking performance. Mastnr's thesis, Naval Postgraduate
School, Monterey, California, September, 1975. (AD A016 378).

Kennedy, R. S. Two procedures for applied and experimental studies of stress. Ft. Rucker, Alabana; U.S.
Army Aeromedical Research Laboratory, 70-11 NANI 1099, February, 1970.

Kerr, B. Processing demands during mental operations. Memory and Cognition, 1973, 1, 401-412.

Kirchner, J. H. and Laurig, W. The human operator in air traffic control systems. Ergonomics, 1971, 14,
549-556.

Klein, T. J. A workload simulation model for predicting human performance requirements in the pilot-
aircraft environment. Paper presented at the 14th Annual Meeting of the Human Factors Society, San
Francisco, California, October 13-16, 1970.

Klei-, T. J. and Cassidy, W. B. Relating operator capabilities to system demands. Dallas, Texas: LTV
Aerospace, Vought Systems Division, 1972.

Klein, T. J. and Hall, A. A.. An analysis of pilot performance requirements in the A-7E team. Dallas,
Texas: LTV Aerospace, Vought Systems Division, VSD Report No. 2-542201 5R-5777, April, 1975.

Knowles, W. B. Operator loading tasks. Human Factors, 1963, 5, 155-161.

Kornstadt, H. J. and Pfennigslorf, J. Evaluation of an integrated flight display for the manual IFR-
landing of VTOL-aircraft. Froceedings oi the AGARD Conference on Guidance and Control Displays,
AGARD-CP-96, 10-1 - 10-8, 1972.

Koym, K. G. Familarity effects on task difficulty ratings. Brooks AF3, Texas: USAF Human Resources
Laboratory, APHRL-TR-77-25, June, 1977.

Kradz, H. P. The psychological stress evaluator. Ellicott City, Maryland: Howard County Police Depart-
ment, Unpublished manuscript, 1974.

Kraft, C. L. and Elworth, C. L. Flight deck workload and night visual approach performance. In AGARD
measure of aircrew performance. Report No. N70-19786, December, 1969.

Krause, E. F. and Roscoe, S. N. Reorganization of airplane manual flight control dynamics. U. B.

Knowles, M. S. Sanders, and F. A. Muckler (Ede.) Proceedings of the Sixteenth Annual Meeting of the
Human Factors Society. Santa Monica, California: Htman Factors Society, 1972, 117-126.

. . ..... 1 a...at
 39

[rebs, M. J. and Winaert, J. W. Use of the ocusometer in pilot workload measurement. Washington, D.C.:
National Aeronautics and Space Administration, NASA CR-144951, February, 1976.

"Krebs, N.J., Wingert, J. W. and Cunlnghm, T. Exploratton of an oculometer-based model of pilot workload.
Washington, D.C.: National Aeronautics and Space Administration, NASA CR-145153, March, 1977.

Kreif•ult, J., Parkin, and Rothschild, P. Implications of a mixture of aircraft with and without traffic
situation displays for air traffic management. Proced!%@ of the 12th annual NASA-University Conference
on Manual Control, University of Illinois, Nay, 1976, 179-200. (NASA T1 X-73 70).

Krivohlavy, J. Phyoiological correlates of the Informational performance. Activitas Nervosa Superior,

1968, O0, 165-171. (In Csechoslovakian). (a)

Irivohlavy, J. Pulse rate and information load during typing. Activitas Nervosa Superior, 1968, 10,
172-176. (In Czechoslovaklau). (b)

Krol, J. P. Variations in ATC-workload as a function of variations in cockpit workload. Ergonomics, 1971,

1.4, 385-590.

Krzanowaki, W•.J. and Nicholson, A. N. Analysis of pilot assessment of workl1oad. Aerospace Medicine,
1972, A_3, 9-007.

Kuhar, W. R., Gavel, P. and Moreland, J. A. Impact of automation 'ipon air traffic control system
productivity/capacity (A~rS-111). Washington, D.C.: Federal Aviation Administration, FAA-RD-77-39,
November, 1976.

Lane, N. E. and Streib, N. 1. The human operator simulator: Vcrkload estimation using a simulated
secondary task. Paper presented at NATO/AGARD Aerospace Medical Panel, Cologne, Germany, April, 1977.

Lane, N. E., Wherry, R. J., Jr. and Streib, M. The human operator simulator: Estimation of workload
reserve using a simulated secondary task. Proceedings of the AGAiD Conference on Methods to Assess
Werkload, AGARD-CPP-216, April, 1977, All-1.

Laurell, H. and Lispar, H. A. A validntion of subsidiary reaction time against detection of roadoide
obstacles during prolonged driving. Ergonomics, 1978, in press.
Laurig, W. and Phillip, U. Changes in the pulse frequency rhythm in relation to the workload.

CVeranderunden der Pulsfrequenzarrhythmie in Abhangigkeit von der Arteitsschwere). Arbeitsmdeizin

Sozilamedizin, ArbeltshygLene, 1970, 5, 184-188. (Royal Aircraft Establishment Library Translation 1586).

Lauschner, E. A. Measurement of aircrew performance: The flight deck workload and its relation to pilot
perforLmnce. AGARD Aerospace Medical Panel, AGARD-CP-56, May, 1969.

Laville, A.. Teiger, C., and Duraffourg, J. An attempt to evaluate workload in a repetitive task. Paper
prescnted at the annual conference of the Ergonomics Research Society, April, 1972.

Lebacqz, J. V. and ALiken, E. W. A flight investigation of control, display, and guidance requirements for
decelerating descending VTOL instrument transitions using the X-22A variable stability aircraft. Volume I.
Buffalo, New York: Calspan Corp~ration, Ak-5336-F-l, September, 1975.

Leplat, J. Factors determining workload. Ergonomics, 1978, 21, 143-149.

Leplat, J. and Pailhous, J. The analysis and evaluation of mental work. In W. T. Singleton, J. G. Fox,
and D. Whitfield (Eds.) MIeasurement of man at work. London: Taylor and Francis, 1973, 51-56.

Levine, J. M., Ogden, G. D. and Eisner, E. J. Nersurement of workload by secondary tasks. Washington,
D.C.: Advanced Research Resources Organization, Contract No. NAS2-9637, January, 1978.

Levison, W. H. A model for task interference. Proceedings of the 6th Annual NASA-University Conference
on Manual Control, Wright-Patterson AFB, Ohio, April 7-9, 1970, 585-616.

Levison, W. H. Position paper prepared foý A4ATO Symposium on Mental Workload. Mati, Greece, September,
1977.

Lindquist, 0. H. Design implicatiors of a better view of thu multichannel capacity of a pilot. Proceedings
of the AGARD Conference on Guidance and Control Displays, AGARD-CP-96, 5-1 - 5-6, February, 1972.

Linn, V. C., Jr. The parotid fluid technique for the evaluation of mental stress in a production situa-
tion. Texarkana, Texas: US Army Logistics Managemer.t Center, USAMC Intern Training Center, USAMC-ITC
Report No. 2-72-05, July, 1972.

Linton, P. N. VFA-V/STOL crew loading analysis. Warminster, Pennsylvania: U.S. Naral Air Development
Center, NADC-75209-40, May, 1975.

Linton, P. N., Jahns, D. W. and Chatelier, P. R. Operator workload assessment model: An evaluation of a
SVFI/VA-V/STOL system. Proceedings of the AGARD Conference on Methous to Assess Workload, AGARD-CPP-216,
April, 1977, A12-1 - A12-11.

Lisper, H. 0., Laurell, H. and Stening, G. Effects of experience of the driver on heart-rate, respiration-
rate, and subsidiary reaction time in a three hours continuous driving task. Ergonomics, 1973, 16,
501-506.
40

Lovsesy, I. J. Inflight recording of helicopter pilot activity. Proceedings of AGARD Conference on

Studies on Pilot Workload, AGARD-CPP-217, April, 1977, B3-1 - B3-10.

Luceak, H. and Laurig, W. An .analysis of heart rate var•ibility. Ergonomics, 1973, 16, 85-97.

Nachac, M. Mental load, fatigue, and recovering. Psychologic, 1971, 6, 72-79. (In Czechoslovakian).

Mabhhour, M. The effect of motion on attention in man-machine systems. Stockholm, Sweden: University
of Stockholm, The Psychological Laboratories, April, 1969.

McDonald, L. B. and Ellis, N. C. Stress threshold for drivers under various combinations of di$creLe
and tracking workload. Proceedings of the 19th annual meeting of the Human Factors Society, Dallas,
Texas, October, 1975, 488-493. (a)

"McDonald, L. B. and Ellis, N. C. Driver workload for various turn radii and speeds. In Driver perform-
ance studies: Transportation Research Record 530. Washington, D.C.: Transportation Research Board,
TRR 530, 1975, 18-30. (b)

McFeely, T. E. Pupil dismeter and the cross-adaptive critical tracking task; A method of workload
measurement. Master's thesis, Naval Post-graduate School, Monterey, California, June, 1972. (AD 749 075).

McGrath, J. J. Temporal orientation and task performance. Goleta, California: Human Factors Research,
719-IC, January, 1969. (AD 758 909).

McHugh, V. B., Brictson, C. A. and Naitoh, P. Emotional and biochemical effects of high workload.
Proceedings of the AGARD Conference on Simulation and Study of High Workload Operations, AGARD-CP-146,
April, 1974, A12-1 - A12-9.

McLean, J. R. and Hoffmann, E. R. Steering reversals as a measure of driver performance and steering
task difficulty. Human Factors, 1975, 17, 248-256.

Merhav, S. J. and Ya'acov, 0. B. Control augmentation and workload reduction by kinesthetic information
from the manipulator. Proceedings of the 12th Annual NASA-University Conference on Manual r_-trcl,
University of Illinois, May, 1976. (NASA 7M T-73 70).

Michon, J. A. A note on the measurement of perceptual motor load. Ergonomics, 1964, 7, 461-463.

Michon, J. A. Tapping regularity as a measure of perceptual motor load. Erionomics, 1966, 9, 401-412.

Nichon, J. A. and Doorne, H. van Equipment note: A semi-portable apparatus for the measurement of
perceptual motor load. Ergonomics, 1967, 10, 67-22.

Miller, R. G. and Rubin, R. T. Stress of airczaft carrier landings. 1. Corticosteroid responses in

Naval aviators. San Diego, California: Navy Medical Neuropsychiatric Research Unit, Report No.
NNMNU-70-16, April, 1970.

Mobbs, R. F., David, G. C. and Thomas, J. M. An evaluation of the use of heart rate irregularity as a
measure of mental workload in the steel industry. London, England: British Steel Corporation, KISRA,
OR/HR/25/71, August, 1971.

Monty, R. A. and Ruby, V. J. Effects of added workload on compensatory tracking for maxium terrain
following. Human Factors, 1965, 7, 207-214.

Moray, N. Mental workload position paper. Position paper prepared for NATO Syr-,osium on Mental Workload.
Mati, Greece, September, 1977.

Morgan, T. R. Inflight physiological data acquistion system. Brooks AFB, Texas: USAF School of Aero-
space Medicine, SAM-TR-75-46, December, 1975.

Norrissette, J. 0., Crannell, C. W. and Switzer, S. A. Group performance under various conditions of
workload and information redundancy. Wright-Patterson AFB, Ohio: Aerospace Medical Research Laboratory,
A•U.-TR-65-16, April, 1965.

Molder, G. The heart of mental effort. Position paper prepared for NATO Symposium on Mantal Workload.
Mati, Greece, September, 1977.
Mulder, G. and Mulder-Hajonides van der Meulen, W.R.E.H. Mental load and the measurement. of heart rate
variability. Ergonomics, 1973, 16, 69-83.

Murphy, J. V. and Gurman, B. S. The integrated cockpit procedure for identifying control and display
requirements of aircraft in advanced time periods. Proceedings of the AGARD Conference on Guidance and
Control Displays, AGARD-CP-96, 4-1 - 4-7, 1972.

Morphy. N. R. Coordinated crew performance in commercial aircraft operations. Proceedings of the 21st
Annual Meeting of the Human Factors Society, San Francisco, California, October, 1978, 416-420.

Morphy, M. A., McGee, L. A., Palmer, E. A., Paulk, C. H. and Wempe, T. E. Simulator evaluation of three
situation and guidance displays for V/STOL zero/zero landings. Proceedings of the 10th Annual NASA-
University Conference on Manual Control, Wright-Patterson AFB, Ohio, April 9-11, 1974.
 41

)Mrrell, J. F. Pilot's assessment of their cockpit enviroument. In Problem. of the Cochit hvironment,
AOARD Conferwvice Proceedings No. 55, March, 1970.

Nagaraja Rao, B. K. and Griffin, J. J. Secondary task performance of helicopter pilots during low-level
flight. University of Southampton, Institute of Sound and Vibration Research, Report No. iV-TR-54,
December, 1971.
Navon, D. and Gopher, D. On the economy of the human processing system: A model of multiple capacity.
Haif a, Israel: Technion Technical Report AhOSR-77-l, 1977.

Nicholson, A. N. Aircrew workload during the approach and landing. Aeronautical Journal, 1973, 77, 286-289

Nicholson, A. N., Hill, L. I., lorland, R. G., and Krzanowski, W. J. Influence of workload on the
neurological state of a pilot during the approach and landing. Aerospace Medicine, 1973, 44, 146-152.

Noble, M. and Trumbo, D. The organization of skilled response. Organizational Behavior and Human
Performance, 1967, 2, 1-25.

Noel, C. 3. Pupil diameter versus task layout. Master's thesis, Naval Postgraduate School, Monterey,
California, September, 1974.

North, R. A. Task components and demands as factors in dual-task performance. Savoy, Illinois:
University of Illinois at Urgana-Champaign, ARL-77-2/A1OSR-77-2, January, 1977.

North, R. A. and Goper, D. Measures of attention as predictors of flight performance. Hman Factors,
1976, 18, 1-14.

Boyer, A. Mental fatigue and palmar skin resiastance. Travail Humain, 1971, 34, 289-298. (In French).

O'Connor, M. F. and Suede, B. N. The application of a decision analytic taebniques to the test and
evaluation phase of the acquisition cf a iajcr air system. McLean, Virginia: Decisions and Designs,
Technical Report 77-3, April, 1977.

O'Donnell, R. D. Secondary task assessment of cognitive workload in alternative cockpit configurations.

Wright-Patterson AFB, Ohio: Aerospace Medical Research Laboratory, ANRL-TR-75-49, 1975. (AGARD-CPP-181).

O'Donnell, R. D. Secondary task assessment of cognitive workload in alternative cockpit configurations.

In Higher mental functioning in operational environments. North Atlantic Treaty Organization, April,
1976.

O'Donnell, R. D. and Spicuzza, R. J. Pilot performance assessment in systems using integrated digital
avionics. Proceedings of the 46th Annual Meeting of the Aerospace Medical Association, San Francisco,
California, 1975.

Ohhara, S. Changes of tracking performance, respiration, and heart rate during experimentally induced
anxiety. Japan Air Self Defease Force. Aeromedical Laboratory Reports, 1970, 11, 198-205. (In Japanese).

Older, H. J. and Jenney, L. L. Psychological stress measurement thrmugh voice output analysis. Alexandria,
Virginia: The Planar Corporation, Contract NASA 9-14146, March, 197,

Olson, B. A. Display and control requirements study for a V/STOL tactical aircraft. Wrigbt-Patterson
AFB, Ohio: USAF Flight Dynamic. Laboratory, AFFDL-TR-66-114, December, 1966.
Onstott, E. D. Task interference in multi-axis aircraft stabilization. Proceedings of the 12th Annual
NASA-University Conference on Manual Control, University of Illinois, May, 1976, 80-103. (NASA ]M X-73,
70).

Onstott, E. D. and Faulkner, W. H. Predication of pilot reserve attention capacity during air-to-air
target tracking. froceodings of the 13th Annual NASA-University Conference on Manual Control,
Massachusetts Institute of Technology, June 15-17, 1977, 136-142.

Opmesr, C.H.J.M. The Information content of successive Ri-interval times in the ECG. Preliminary
results using factcr analysis and frequency analysis. Ergonomics, 1973, 16, 105-112.

Parks, D. L. Current workload methods and emerging challenges. Seattle, Wash.: The Bosing Co.,
Document No. D6-44563T3, July, 1977.

Parks, D. L. and Springer, W. E. Human factors engineering analytic process definition and criterion
development for Computer Aided Function-allocation Evaluation System (CAFES). Seattle, Vashington:
Boeing Aerospace Company, D180-18750-1, January, 1976.

Pettyjohn, F. S., McNeil, R. J., Akers, L. A. and Faber, J. M. Use of inspiratory minute volumes in
evaluation of rotary and fixed wiug pilot workload. Proceedings of the AGARD Conference en Mtithods to
Assess Workload, AGARD-CPP-216, April, 1977, A9-1 - A9-2. (a)

Pettyjohn, F. S., McNeil, R. J., Akers, L. A. and Faber, J. M. Use of inspiratory minute volumes in
evaluation of rotary and fixed wing pilot workload. Fort Rucker, Alabama: U.S. Army Aeromedical
Research Laboratory, USAA1L Report No. 77-9, April, 1977. (b)

Pew, R. W. Position paper on workload. Prepared for NATO Symposium on Mental Workload. Mati, Greece,
September, 1977.
Photo&., A. V. Improvememt In weapon system effecttvaness by application of identification setbods for

determdiing hkman operator performance decrements under stress conditions. Palo Alto, Californias
Systams Control, December, 1973.

Phililop, U., Reich*, D. and Kirchner, J. H. The use of subjeccive rating. 1_.aon .ce, 1971, 14,
611-616.
Phillips, J. F. The feasibility of short interval time estimation as a methodology to forecast human
performance of a specifled task. Red River Army Depot, Texarkana, Tezas: UnBCWM Intern TraJning Center,
DAJRa(-ITC-02-08-76-010, April, 1976.

Poston, A. M. A survey of existing computer programs for sircrew workload sesesament. Aberdeen Proving
Ground, Maryland: U.S. Army Human Enigneering Laboratory, Technical Momorandua 13-78, May, 1978.
Potempa, K. W. A catalog of human-factcrs techniqucs for testing new systems. Wright-Patcerson AlB,
Ohio: USAF Humae Resources Laboratory, AIRIL-TI-68-15, February, 1969.

Price, D. L. The effects of certain glabal orders on target acquisition and workload. Human Factors,
1975, 17, 571-576.
Price, B. E. Development of potential roles of supersonic transport crews. Chatsworth, California:
Serendipity AssocLiate., T1 20-66-3, December, 1965.

Price, H. E., Honabergar, W. D., and Ereneta, W. J. A study of potential roles of supersonic transport
crews and some implications for the flight deck, Volume I: Workload, crew roles, flight deck concepts,
and conclucions. Noffeett Field, California: National Aeronautica NASA CR-561, October, 1966.

Pritaker, A.A.B., Wortumn, D. R., Seum, C. S., Chubb, G. P. and Siefert, D. J. SAINT: Volume I. Systems
analysis of integrated network of tasks. Wright-Patterson AFB, Ohio: Aerospace Medical Research
Laboratory, AHRL-Ti-73-126, April, 1974.
Rasmussen, J. Reflections on the concept of perator workload. Position paper prepared for NATO
Symposium on Mental Workload. Kati, Greect Seotembfr, 1977.
Rasing, J. K. The definition and measurement of pilot workload. Wright-Patterson AFB, Ohio: USAF
Flight Dynamics Laboratory, AFFDL-TH-72-4-FGR, February, 1977.

Raps, B. 3. knd Wierwille, W. W. Driver performance in controlling a driving simulator with varying
vehicle response characteristics. SAE Paper No. 760779, October, 1976.
pko, J. D., Toeb, N. and Brown, B. 1. Behavioral effects of prolonged exposure to continuous and
.ntermittent noise. Louisville, Kentucky: University of Louisville, Performance Research Laboratory,
ýTR-74-29, June, 1974.
Replogle, C. R., Holden, F. M., Gold, R. Z., Kalak, L. L., Jonas, F. and Potor, C., Jr. Human operator
performance in hypoxic stress. Wright-Patterson APB, Ohio: USAF Aerospace Medical Research Laboratory,
AMRL-TR-71-29, Paper No. 31, December, 1971.

Rohmert, W. An international symposium on objective assessment of workload in air traffic control tasks:
Held at the Institute of Arbeitswissenschaft, The University of Technology, Darmstadt, German Federal
Republic. Areonomics, 1971, 14, 545-547.

Rohmert, H. Deternination of stress and strain of air traffic control officers. Proceedings of the
AQMA DCon' ice on Methods to Assess Workload, AGARD-CPP-216, April, 1977, A6-1 - Af-8.

.amterP, .'., Laurig, W., Philipp, U. and Luczak, H. Heart rate variability and workload measurement.
Ergonomics, 1973, 16, 33-44.

Rolfe, J. N. Multiple task performance: Operator overload. Occupational Psychology, 1971, 45, 125-132.
Rolfe, J. M. Whither workload. Applied Ergonomics, 1973, 4, 8-10. (a)

Rolfe, J. V secondary task as a measure of mental load. In W. T. Singleton, J. G. Fox, and D.

Whitfield easurement of Man at Work. London: Taylor and Francis, 1973, 135-148.
M.) (b)

Rolfe, J. H., Chappelow, J. W., Evans, R. L., Lindsay, S.J.E. and Browning, A. C. Evaluating measures
of workload using a flight simulator. Proceedings of the AGARD Conference on Simulation and Study of
High Workload Operations, AGARD-CP-146, April, 1974, A4-1 - A4-13.

Rolfe, J. N. and Lindsay, S.J.E. Flight deck environment and pilot workload: Biological measures of
w.orkload. Applied Ergonomics, 1973, 4, 199-206.

Rosch, E. and Wempe, T. Secondary task for full flight simulation incorporating tasks that comonly
cause pilot error: Time estimation Moffett Field, California: NASA-AMES Research Center, NASA-Th-X-
74153, October, 1975.
Roscoe, A. H, Pilot vorkload during steep gradient approaches. lVarnsborough, England: Royal Aircraft

Establishment, Flight Systems Department, Technical Memorandum No. Th FS 78, 1976.

Roscoe, A. H. (Ed.) Assessing pilot workload. AGARD-AG-233, February, 1978. (a)
 43

Roscoe., A. H. Physiological uethods. In A. H. Roscoe (Ed.) Asses 1i pilot workload. AGARD-Ah-233,

February, 1978 (b).

Rosco*, S. N. Assessment of pilotage error in airborne area navigation procedures. Human Factors, 1974,
16, 223-228.

Rosenubvck, F. Hardware problems in ergonomics measurements. Ergononlcs, 1971, 14, 617-623.

Roult, A. Outlines of a position paper. Position paper prepared for IATO Symposium on Mental WorkJoad.
Nati, Greece, September, 1977.

Rouse, W. B. Approaches to mental workload. Position paper prepared for NATO Symposium on Mental
Workload. Mati, Greece, September, 1977. (a)

Rouse, W. B. Hutmn-computer interaction in multitauk situations. IRKE Transactions on Systems, Man,

and Cybernatica. 1977, IC-7, 384-392. (b)

Sander-9, A. F. Some remarks on mental load. Pusition paper prepared for NATO Symposium on Mental
Workload. Mati, Greece, Septmqber, 1977.

Sanders, M. S., Jankovich, J. J. and Goodpaster, P. R. Task analysis for the jobs of train conductor
and brakeman. Crane, Indiana: Naval Ammunition Depot, RDTR-No. 623, July, 1974.

Sanders, N. G., Simmons, R. I., Hofmann, M. A. and DeBonis, J. N. Visual workload of the co-pilot/
navigator during terrain flight. Proceedings of the Human Factors Twenty-First Annual Meeting.
San Francisco, Calif crnia: human Factors Society, October 1977, 262-266.

Savage, R. K., Wierwille, W. W. and Cordes, R. E. Evaluating the sensitivity of various measures of
operator workload using random digits as a secondary task. Human Factors, in press.

Sayers, B. M. Analysis of heart rate variability. Ergonomics, 1973, 16, 17-32.

Schiffler, R. J., Geiselhart, R. and Ivey, L. Crej composition study for an Advanced Tanker/Cargo
Aircraft (ATCA). Wright-Patterson, AFB, Ohio: USAF Aeronautical Systems Division, ASD-TR-76-20,
October, 1976.

Schiflett, S. U. Operator workload: An annotated bibliography. Patuxent River, Maryland: US Naval

Air Test Center, SY-257R-76, December, 1976.
Schori, T. R. A comparison of visual, auditory, and cutaneous tracking displays when divides attention
is required to a cross-adaptive loading task. Ergonomics, 19)3, 16, 153-158.

Schori, T. R. and Jones, B. W. Smwking and workload. Journal of Motor behavior, 1975, 7, 113-120.

Schouten, J. F., Kalebeek, J.W.H., and Leopold, F. F. On the evaluation of perceptual and mental load.
Ergonomics, 1962, 5, 251-260.
Schultz, V. C., Newell. F. D. and Whitbeck, R. F. A study of relationships between aircraft system
performance and pilot ratings. Proceedings of the 6th Annual NASA-University Conference on Manual
Control, Wright-Patterson AFB, Ohio, April 7-9, 1970, 339-340.

Schwartz, J. J. and Ekkers, C. L. Estimation of task loading by observing and regulating complex
technical systems. Mans en onderneming, 1976, 76, 85-108. (In Dutch).

Seibel, R., Christ, 11.E. and Teichner, W. H. Perception and short term memory under workload stress.
Port Washington, New York: U.S. Naval Training Device Center, NAVTRADEVECEN 1303-2, June, 1964.

Seifert, R., Daniels, A. F. and Schlmidt, K. A method of man-display/control system evaluation.

Proceedings of the AGARD Conference on Guidance and Control Displays, AGARD-CP-96, 8.-1 - 8-8.

Senders, J. W. The human operator as a monitor and controller of aultidegree of freedom systems. IEEE
Transactions on Human Factors in Electronics, 1964, HWE-5, 2-5.
Senders, J. W. The estimation of operator workload in complex systems. In K. B. DeGreene (Ed.) Systems
Psychology. New York: McGraw-hill, 1970.

Senders, J. W., Kristofferson, A. B., Levison, W. H., Pietrich, C. W. and Ward, J. L. The attentional
demand of automobile driving. Highway Research Record, 1967, No. 195, 15-33.

Sheridan, T. B. and Stassen, H. G. Definitions, models and measures of human workload. Position paper
prepared for NATO Symposium on Mental Workload. Kati, Greece, September, 1977.

Sharman, N. R. The relationship of eye behavior, cardiac activity and electromyographic responses to
subjective reports of mental fatigue and performance on a Doppler identification task. Master's thesis,
Naval Postgraduate School, Monterey, California, September, 1973. (AD 769 754).

Shulman, H. G. and Briggs, G. E. Studies of performance in complex aircrew tasks. Columbus. Ohio: The
Ohio State University, Research Foundation, RF Project 2718, Final Report, December, 1971.
 Siegel, A. I., Lanterman, R. S., Flatzer, H. L. and Wolf, J. J. Techniques for evaluating operator
loading In smn-machine systems: Developmen,:: of a method for real time assessment of operator overloading.
Wayne, Pennsylvania: Applied Psychological Services, January, 1976.
Siegel, A. I. and Williams, A. R., Jr. Identification and measurement of intellective load carrying

thresholds. Wayne, Pennsylvania: Applied Psychological ServLces, AFOSR-TR-75-0593, December, 1974.

Siegel, A. I. and Wolf, J. J. Man-machine samllation models: Psychosocial and performance interaction.
New York: Wiley, 1969.

Siegel, A. I., Wolf, J. J., Fischl, M. A., Miehle, W. and Chubb, G. P. Modification of the Sicgel-Wolf
operator simulation model for on-line experimentation. Wright-Patterson AFB, Ohio: USAF Aerospace
Medical Research Laboratory, ANRL-TR-71-60, June, 1971.

Siegel, A. I., Wolf, J. J. and Sorenson, R. T. Technicians for evaluating operator loading in man-
machine system: Evaluation of a one or a two-otorator system evaluating model through a controe ! d
laboratory test. Wayne, Pennsylvania: Applied Psychological Services, Contract Nonr 2-492(00),
July, 1962. (AD 284 182)

Simmons, R. R. and Kimball, K. A. Methodological considerations of visual workloads of helicopter pilots.

Proceedings of the AGAlD Conference on Methods to Assess Workload, AGARD-CPP-216, April, 1977, A2-l.

Simon@, R. R., Kimball, K. A. and Diaz, J. J. Measurement of aviator visual performance and workload
during helicopter operations. Ft. Rucker, Alabama: U.S. Army Aeromedical Research Laboratory, 77-4.
December, 1976.

Simonov, P. V. and Frolov, M. V. Analysis of the human voice as a method of controlling emotional state:
Achievements and goals. Aviation, Space, and Environmental Medicine, 1977, 48, 23-25.

Simpson, C. A. and Hart, S. G. Required attention for synthesized speech perception for two levels of
linguistic redundancy. Paper presented at the 93rd meeting of the Acoustical Society oi America, SLate
College, Pennsylvania, June 7-10, 1977.

Simpson, C. G. Improved displays and stabilization in general aviation aircraft. Moffetr Field, California
National Aeronautical and Space Administration, N69-24238, November, 1968. (AGARD Symposium paper).

Sinaiko, H. W. Third international congress on ergonomics. London, England: Office of daval Research,
ONtL-C-19-67, November, 1967.

Smith, W. S., Jr. Effects of nsuromuscular tension in the use of an isometric hand controller. Monterey,
California: U.S. Naval Post-graduate School, Master's thesis, December, 1972.

Soede, M. Reduced mental capacity and behavior of a rider of a bicycle s*Lulator under alchohol stress
or under dual task load. Proceedings of the 13th Annual NASA-University Couference on Manual Control,
Massachusetts Institute of Technology, June 15-17, 1977, 143-151. (a)

Soede, N. On mental load and reduced mental capacity; some considerations concerning labormtory and
field investigations. Position paper prepared for NATO Symposium on Mental Workload. Mati, Graeca,
September, 1977. (b)

Soliday, S. M. Effects of task loading on pilot performance during simulated low-altitude high-speed
flight. Fort Eustis, Vicginia: U.S. Army Transportation Research Center, USATRECOM 64-69, February,
1965.

Soliday, S. M. and Schonan, B. Task loading of pilots in simulated low-altitude high-speed flight.
Human Factors, 1965, 7, 45-53.
Sontend•m, J. Insetruments and methodology for the assessment of physiological cost of performance in
stressful continuous operations the air traffic services tower environment. Proceedings of thL AGARD
Conference on Methods to Assess Workload, AGARD-CPP-216, April, 1.977, A7-1 - A7-32.

Iperandio, J. C. Variation of operator's strategies and regulating effects on workload. Ergonomics,

1971, 14, 571-577.

Sperandio, J. The regulation of working methods as a function of workload among air traffic controllers.
Ergonomics, 1978, 21, 195-202.

Spicuzza, R. J., Pinkus, A. R. and O'Donnell, R. Z. Development of performance assessment methodology

for the digital avionics information system. Dayton, Ohio: Systems Research Laboratc-ies Final Report,
1 December 1973 - 30 June 1974, August, 1974.

Spyker, D. A., Stackhouse, S. P., Khalafalla, A. S. and McLane, R. C. Development of techniques for
measuring pilot workload. Washington D.C.: National Aeronautics and Space Administration, Contractor's
Report NASA CR-1888, November, 1971.

Stackhouse, S. Workload evaluation of LLNO display. Minneapolis, Minnesota: Honeywell, 7201-3408,

October, 1973.

f_,• Stackhouse, S. P. The measurement of pilot workload in manual control system. 14inneapolls, Minnesota:
Honeywell, Inc., FP398 FRI, January, 1976.

Stackhouse, S. P. Meastirement of aircrew informaL'ion processing workload. Minneapolis, Minnesota:

Honeywell, Unpublished manuscript, 1978.
 45

in a complex task. Journal of 4

Stager, P. and Moter, P. Instructions and information processing
E•perlmumtal Poychology, 1971, 87, 291-294-.
Stager, P. and Zufalt, K. Dual-taskctethod in determining load differences. Journal of Experimental
Psycholo•y, 1972, 94, 113-115.

Stanford, B. A. Validity and relia~iliULy of subjective rating of perceived exertion during work.
reonomics, 1976, 19, 53-60.

Steininger, K. Subjective ratings of flying qualities and pilot workload In the operation of a short
haul jet transport aircraft. Proeoedjs. of AGARD Conference on Studies on Pilot Workload, AGARD-CPP-217,
April, 1977, B11-1.

Steininger, K. and Wistu'a, C. Minimum flight crew of transport aircraft. Methods for measuring workload
of flight crews. Hamburg, West Germany: Deutsche Porachungs-and VWrsuchoanotatt fuer Luft-und Raunfahrt,
Repor- No. DLR-1S-355-74/3, 1974. (In German).

Stephens, B. W. and Michael@, R. N. Time-sharing between two driving tasks: Simulated steering and
recr.gnition of road signs. Paper presented at the 43rd Annual Meeting of the highway Research Board,
Washington, D.C., January, 1964.

Sternberg, S. High-speed scanning in himan memory. In R. 4. Haber (Ed.) New York: Dolt, Rinehart and
Winston, 1969.

Stors, W. P. and Hapeaney, J. D. Mission-Crew fatigue during rivet joint operations. Brooks AFB, Texas:
USAY School of Aerospace Medicine, SAh-TR-76-36, September, 1976.

Storm, W. P., Hartman, 1. 0., Intono, G. P. and Peters, G. L. Endocrine-metabolic effects In short-
duartion high-workload mission@s: Feaslb'.lity study. Brooks AFB, Texas: USAF School of Aerospace
.edicine. SAM-Ti-76-30, Augnst, 1976.

Street, R. L., Singh, H. and Hale, P. N., Jr. The evaluation of mental stress through the analysis of
parotid fluid. Human Factors, 1970, 12, 453-455.

Strieb, i. 1. The hben operator slmlalor volume 1: Introduction and overview. Willow Grove,
Pennsylvania: Ana•.ytles, Augist, 1975.

Strieb, M. I., Glenn, F. A., Fisher, C. and Fitte, L. B. Chapter VII from the human operator simulator
volume VII LAMPS air tactica1 officer simulation. 4illow Grove, Pennsylvania: Analytics, November,
1976.

Strothee, D. D. Alrcrew pcrformance in asty aviation. Proceedings of Conference 27-29 Nuvember 1973,
U.S. Army •viation Center, Fort Rucker, Alabama, November, 1973, 188-192.

Strother, D. D. Visual and manual workload of the helicop!er pilot. Paper presented at toe Annual
National Forum of the American Helicopter Society, Washington, D.C., May, 1974. (Preprint No. 821).

Sun, P. B., Keane, W. P. and Stackhouse, S. P. The measurement of pilot workload in manual control
systems. Proceedinfe of Aviation Electronice Symposium, Fort Monmouth, New Jersey, April, 1976.

Telcbgraber, W. N. The effects of signal quantization on compensatory tracking performance. Unpublit.hed

naster's thesis, Naval Postgraduate School, December, 1972.

Teiger, C. Regulation of activity: an analytical tool for studying workload In perceptual motor tasks.
Ergonomic*, 1978, 21, 203-213.

Terbraak, P. High workload tasks of aircrew in the tactical strike, attack, and reconnaissance roles.
In AGARD smulation and study of high workload operations, AGARD-CP-146, October, 1974.

Thorne, R. 0. Pilot "orkload: A conceptual model. AGARD Conference Proceed!ng, No. 119 on Stability
and Control, Braunschweig, Germany, April 10-13, 1972, 21-1-21-6. (AGARD-CP-119).

Triges, T. 1. Aspects of mental workload. Proceedinse of the IEE-QMISRS Internmtional Syfposima

on Men-Machine Systems, Cambrige, nagland, September, 1969.

Trumba, D. and Noble, N. Response uncertainty in dual-task performance. Organizational Pehavior and
Human Performance, 1972, 7, 203-215.

Trumbo, D., Noblc, N. and Swrinik'J. Secondary task interference in the performance of tracking tasks.
Journal of xprimental Psychology, 1967, 73, 232-240.

Ursin, H. and Ursin, R. Physiological indicators of mental load. Position paper prepared for NATO
Symposium on Mental Workload. Hati, Greece, September, 1977.

van Gigch, J. P. A model for measuring the Information processing rates and mental load of complex
activities. Canadian Operational Research Society Journal, 1970, 8, 116-128. (a)

V van Gich, J. P. Applications of a model used in calculating the mental load of workers in industry.
Canadian operational Research Society Journal, 1970, 8 176-184. (a)

.1i
46

Verplank, W. L. Is there an optmm workload in manual control? Proceedinas of the 12th Annual NASA-
University Conforence on Manual COntrol, 'Uiversity of Illinois, May, 1976, 72-79. (1NAM T 1-73, 70).

Weller, K. C. An investigation of correlation between pilot scanning behavior and workload using stopwise
regression analysis. Hampton, Virginia% NIASA Langley Research Center, NASA IN X-3344) March, 1976.

Watson, B. L. The effect of secondary tasks on pilot describing functions In a compensatory tracking task.
Toronto, Canada: University of Toronto, Institute for Aerospace Studies, UTIAS Technical Note No. 178,
June, 1972.

Waugh, J. D. Pilot performance in helicopter simulator. Aberdeen Proving Ground, Maryland: U.S. Army
Hunginsering Laboratory, Technical Memorandum 23-75, September, 1975.

Weir, D. H. and Klein, R. H. Measurement and analysis of pilot scanning behavior during simulated
instrument approaches. Proceedings of the 6th Annual NASA-University Conference on Manual Control,
Wright-Patterson AFB, April, 1970, 83-108.

Welfor4, A. T. Mental workload as a function of demand, capacity, strategy and skill. txponomics, 1978,
21, 151-167.

Wampe, T. S. and Bate, D. L. Himan information processing rates during certain, multiaxis tracking tasks
with a concurrent a-iditory task. 1199 Transactions on Man-Machine Systems, 1968, K-_99, 129-138.

Westbrook, C. B., Anderson, R. 0. and Pietrzak, P. I. Handling qualities and pilot workload. Wright-
Pattersou AFB, Ohio: AF Flight Dynamics Laboratory, APFDL-FDCC-TD-66-5, September, 1966.

Weverinke, P. H. Human control and monitoring-models and experiments. Proceedings of the 12th Annual
NASA-University Conference on Manual Control, University of Illinois, May, 1976, 14-38. (NASA TMI-73, 170).

Wewerinke, P. H. Performance and workload analysis of inflight helicopter tasks. Proceedinge of the 13th
Annual NASA-University Conference on Manual Control, Massachusetts Institute of Technology, June 15-17,
1977, 106-117.

Wewerinke, P. H. and Swit, J. A simulator study to investigate human operator workload. Proceodings of
the ALARD Conference on Simulation and Study of Hizh Workload Operations, AGARD-CP-146, April, 1974,
A2-1 -A2-6.

White, Rt.T. Task analysis methods: Review and development of techniques for analyzing mental workload
in multiple-task situations. St. Louis, Missouri: McDonnell Douglas Corporation, NDC J 5291, September,
1971.

White, R. T. Mental workload assessment, I. Laboratory investigation of decision-making And short-term

memory in multiple-task situation. St. Louis, Missouri: McDonnell Dougals, MDC J6662/01, December, 1975.

White, R. T. and Gaume, J. G. Mantal workload assesment, III. Laboratory evaluation of one subjective
and two physiological measures of mental workload. St. Louis, Missouri: McDonnell Douglas Corporation,
Report MDC J7024/0l, December, 1975.

White,
systems.R. T. and presented
Ware, C. T. Prediction of human
Paper at the Anrual Meeting of operator performance
the Western in theAssociation,
Psychological design of comand and control
Vancouver,
British Columbia, June 21, 1969. (Douglas Paper 5539).

Wickens, C. D. The effect of time-sharing on the performance of information processing tasks: A feedback
control analysis. Ann Arbor, Michigan: The University of Michigan, Human Performance Ccnter, Technical
Report No. 51, August, 1974.

Wickens, C. D. The ,effects of divided attention on information processinA in manual tracking. Journal
ftExperimental Psycholony, 1976, 2, 1-13.

Wickens, C. D. Position paper. Paper presented at NATO Conference on Workload. Matapan, Greece,
September, 1977.

Wickens, C. D. and Gopher, D. Control theory measures of tracking as indices of attention allocation
strategies. Human Factors, 1977, 19, 349-365.

Wickens, C. D., Isreal, J., McCarthy, G., Go-her, D. and Donchin. E. The use of event-related potentials
Inthe enhancement of system perforuance. Proceedings o! the 12th Annual HLASA.-University Conference on
Manual Control, University of Illinois, May, 1976, 124-134. (NASA TM X-73, 79,).

Wtckens, C. D., Isreal, J. and Donchin, E. The event related cortical potratial as an index of ask
.. orkload. Proceed inas of the 21st Annual Meeting of the Human Factors Riciety, San Francisco, Californis,
1977.

Wickens, C. D. and Kessel, C. The effects of participatory mode and task workload on the detection ot,
dynamic system failures. Proceedings of the 13th Annual NASA-University Conference on Manual Control,
Massachusetts Institute of Technology, June 15-17, 1977, 126-135.
Wierwille, W. W. and Gutmann, J. C. Comparison of primary and specondary task measures agia function of
simulated vehicle dynamics and driving conditions. Human Factor;, 1978, 20, 233-244.
 47

Werville, W. W., Ouimnn, J. C., Ricks, T. G., and Mhto, W. R. Secondary task measuraeent of workload as
a function of simaulated vehicle dynamics and driving conditions, human Factors, 1977, 19, 557-565.

Wierrille, V. W. and Villiges, I. C, Survey and analysis of operator workload assessment techniques.
Systametrics, Inc. ept. 8-78-101, September, 1976. (Navair Contract 100421-77-C-00631 AD-A059501).

Jildervanck, C., Holder, 0. and Nichon, J. A. Napping owntal load in car driving. R €W , 1978,
21, 225-229.

Wingert, J. V. Functiop Interlace modifications to analytic workload prediction. In K. D. Cross and

J. J. McGrath (ida.) Crew System Deslo. Santa Barbara, California: Anacapa Sciences, July, 1973.

Wisner. A Blectrophysiological measures for tasks of low energy azpanditure. In V. T. Singleton,

J. (• Fjc and D. WhitfI411 (•Dd.) Measurement of msanat weork. Londone Taylor and Francis. L973. 61-73.

Wol J. D. W)rkload evalua.Ion of control-display configurations for approvch to landing. Minneapolis,

1IUnjosota: kyreyve11, SAC Deport F0548-IR, August, 1977.

12

it

SII

F;
 49

TABLE I
Classification of Universal Operator behavior Dimension
(After Berliner, Angell, and Shearer, 1964)

Processes Activities Specific Behavior*

(.1.1 Detes te
.. Inpects
i. Searching for and receiving 1.1.3 Observes
information on1.1.4 Reads
S1.1.5 Recalv'-en
1. Perceptual proeeese4| 1.1.6 Scans
t1.1.7 Survey*
1.2 Identifying objects. actions,
events f"1.2.1
41.2.2 Discriminates
Identifies
Dr.m.3
Locates

f2.1.1 Ctegort"as
f2.12 ýalculates
[2:1.:3 toles
2.1 Information processing /2.1.4 Com Ites*
i2.1. 5 Intervolates
|2.1.6 Itemixeo
2.1.7 Tabulates
2. Mediational processes
)2.1.8 Translates

f2.2.2
:22 . An a~lyz e s
Calculates
2.2 Problem solving and I2.2.3 Chooses
decision-making 2.2.4 Compares
2.2.5 Computes
2.2.6 Estimate.
.2.7 Plans

3.1 Advises
3.2 _Avtsers
3.3 Communicates
3.4 Directs
3. Cormunication processes 3.5 Indicates
3.6 Informs
3.7 Instructs
3.8 Requests
3.9 Transmits
V4.1.1 Activates
4.1.2 Closes
4.1.3 Consects
4.1 Simple/Discrete 4.1. Disconnects
r.
S4.1.5 Joins
4.1.6 Moves
4.1.7 Presses
k4.1.8 sets
4. Motor processes
{42. Adjusts
t!.2:21 Aligns
4.2 Oxoplex/Continuoue (4.2.3 Regulates
4.2.4
44.2.5 Synchronizes
Tracks

Kil
K, S i
 49

TABLE 2

Classification of Workload MetbodoloLies Dimension

1.1 Rating S#-Ales

4 ±. Subjective On4nion
1.2 Interviews and Quaftionnaires

i v2.1 Task Analytic 2.1.1 Task Component, Time Sumation

C*. In format ion-Theoretic
S(!2.21 Sonladptive, Arithmetic/Logic
2. Spare 1lental Capacity 2.2.2 Nonadaptive, Tracking
2.2 Secoudary Task [2.2.3 Time Estiaation
I|2.2.4 Adaptive, Aritbmetic/Logic
V2.2.5 Adaptive, Tracking

*12.3 Occlusion

3.1 Single Measures

3. Priwiry Task 3.2 Multiple Measures

U.3 Math Modeling

4.1.1 FFF
4.1.2 GS5
4.1.3 f.
4.1.4 IM
44.1.5 U
44.1.6 ZcP
.1 Gingle Measures 4.1.7 By* and Eyelid Novsent
44.1.8 Pupillary Dilation
44.1.9 Mascle Tension, Tremor
44.1.10 Heart Rate, Hear Rate Variability,
Blood Pressure
4.1.11 Breathing Analysis
4. Physiological Measuress 4.1.12 iody
Fluid Analysis
r4.1.13 Handwriting Analysis

4.2 Combined Physio?.ogical

Measures

4.3 Speech Pattern Analysi3

r
50

TABLE 3

Applicability Matrix of Workload Methodologies Across

Universal Operator Behavior@

UNIVERSAL OPERATOR BEHAVIORS

•].•UqXD~LOPAOIB UVZL

4J
41

1 V,4 au u

l.'L 'Rating Scales 3 3 3 3 2 3 3

1..Zl Intervilews and Questio-,n-riea 3 3 3 3 2 3 3
2._1.1 Taik Component• Time Simation 3 3 3 . 3 3 3 3
2.L.2 Inforat ion-Theoretic 1 0 2 0 0 0 1
2.2.1 ,Nonadaptive. Arith,/Logic 3 3 3 2 2 3 3
2.2._.2 Nonadapt ive, Tracking 3 3 2 2 2 2 2
2.2,. 3 T4,,,e Kat~lmation 0 0 2 0 2 0_ 1
2.2.4 a:daptitve. Aurith./Loa:Lc 0 0 0 0 0 2 2
1 0 1 0 0 2 2
2.3 Occluion 1 1 1 0 1 01 0
i 3.. . Sinttpe
tinle cealure-Prisary
Neasur e-Prinary 1 1 1 2
3 1 2

3.3 )Kath.Nodelling 0 0 1 0 1 1
[.L.1 T C 1 1
3 1
_1
1 1 0 0
2.1.2 I tS 1 1 1 1 0 1 1
2..1 EG _ 1 1 1 1 0 2 2
2.2.4 T n 2 1 2 2 i 2 2
2.52 C3 0 0 1 0 0 0 0
2-.3c6 ., io I 1 1 1 _ 1 _ _
H.17 Eye anl Eyelid ovement 1 1 1 1 2 0 0
3. Pupillary Dilation 2 12 2 2 1 2 2
4.1.9 Mahsce Tenns Tr0or 1 1 1 1 1 2 2
4.1.43 EM Aayi 2lnwtn 1 2 2 1 2 2
4.1.10 Heart Rate, Heart ate 1 1 1 1 0 0 1
4.1.11
2 Variability.
reathing Analysis
Blood Pressure________ 1 1 1
_______ 1 1 1 1
4.1.12 Body
E luid Analyeis 1 1 1 2 10 2

4.*2 Combined Physiological Measure 1 1 1 1 1 0 1

4.3 Speech Patter' Analysisj 0 0 1 1 1 0 0

Weightings
0 - 4o research support or only negative support
1 - Limited research support; some conflicting data
2 - Limited research support; no conflicting data
423 - Well docCmented research support

3 S

-. !..s c n ..
 TABU 4
Worksheet for Guids In Selecting a Workload Assessment Methodology
for Aircrew Flight Test and Dweluation

UNIVUISAL OPERATOR DUHAVIORSS

DUAVIOR CM=C
_______2.__Tie'

1.1
1.2
latntL
t
t,/) OR WEIGTIU

scale&
Interview. and Questionnaires _______
I j
.t/
C1

___ 1: ___
_

2.1.1- Task Component, Time Summation___

2 * 1.2 Information-Tbeoretic ___

2.2. Noadative. Arith.ILozi____

2.2.3 timtion
2.2.4 Adaptive, Arith./Louic _____
2.2.5 Adaptive, Tracking ___ __
2.3 Occlusion ___

3.*1 Singa.e Measure-PrIslarly __

S3.2 Haitiple Measure-Primary_______

4.1.2 G81
4.1.3 W ______
4.1.4 DOG ___
4.1.6 2C_

4.1.7 Eye and Eyelidovemmt --

4 .18 Pupillery Dilation __

4.1.9 Muwcle Tension, Tremor
4.1.10 Heart Rate, Heart Riaet
Variability. Blood Pressure _

4.1.11 Bresthn Anallsis_____________

4.1.12 BDy Fluid Analrsis
4.1.13
4.2 Handvritim•
Combined Analysis; Measure
Physiolosical
_

I -

4.3 Speech Pattern Analysis_ __

;k
4 52
TABLE 5

Workshest for Guide in Selecting a Workload Aessmment Methodology

for Aircrew Flight Test and Evaluation
SS-3/IRDS Addition Evwmple

UNNEVUS
L OPERATOR BEHAVIORS

ilki
Ia
P
-0 4J4 0

I! 6-
C a

0 .0 4 U

1.1
1.2 Rating Scales
Intervews and _Questionnaires 66 15
15 66 15
15 88 66 99" 65
5 22
2.1.1 Task Component, Time Summation 6 15 6 15 12 6 9 9
6, 1
2.1.2 Information-Theoretic 2 0 4 0 0 0 3 9
2.2.1 Nonadaptive. Arith./log1c _6 5 6 10 8 6 9. 60 4
2.2.2 Noada Tra g 6 15 4 10 4 6o 53 5
S2.2.3S2.2.4 Time
Adapt Estimaton 0 00 40 00
0 08 0
4 3 - 15
10
2.2.5 AdO tive. TrEckInT 2 0 2 0 0 4 6 N4
S2.3 Occlusion 2 5 2 .5 0 2 0 16
3.1 Stinlegeasure-Prizsry 2 5 2 5 4 2 3 23
3.2 Iultieprle aeasure-Prtiaryi 2 5 2 10 0 2 9 27
3.3 Math Model-T hert 0 0 2 0 0 2 3 12
2. 13 O u2 15 2 5 4 0 0 18
2.1.2 G1 a2 ti5 i 2 195 2 3
4'.1.3
2.2.2 EKGndG 24 5
5 42 5
10 04 44 6
6 24
37
4.1.5 E tG 0 0 2 0 0 0 0 2
4.1.6 _CP 2 5 2 5 4 2 3 23
3.3.7 EMan elid Movement 2 5 2 5 8 0 0 22
4.1.8 Puillat Dilation 2 10 4 4 6 42
4.1.9 Muscle Tension, Tremor 2 5 2 5 4 4 6 28
4.1.10 Hea5 t Rate, Heart Rate 2 5 2 5 0 0 3 17
4.1.11
6 athng Analysis
VariabilityBod rsure1 2 5 2 5 4 2 3 23
4.1.12 Body alu!A Analovsis 2 5 2 10 4 2 6 31
4.1.83 andwpitr Analasis 2 5 4 10 8 0 0 29
4.2
1 •0bined
R Heart tee
hae easre 2 5 2 5 4 0 3 21
4.3 Speech Pattern Analysis 20 0 2 5 4 0 0 12

4i,
 53

TABLI 6
Feasibility of Workload Techniques for In-Flight Enviroments

CRITICAL CRITERIA

p4 0

a~l~

1.1
1.2
2.1.1
2.1.52
Rating Scales
Intervie .and Questionnaires
Task Component, Tise Sumation
S
S
S
_ _S
"
S
_ S

S
id
S
._ P

S
S
s
S
1S
s
S
2".2.1 .Eformat .- Theoretic
. onadapt.lve- -Arlth. /Loxc SS SS- S
8 S
S S.
P P5 S
8
2.2.2 N.onmdapO<.s Track:Lna
2.2.3 Tiau•pi •iatio- S S S S P S S
2.2.4 Adapt•ive, Arith./Lostic P S S p S P S
w ..2.5
2.3 Adapte,
LUcclusi.on Tracking ._ P S
S PP SP S
S P
P P
p

3.1 Skiacle Teasure-Primary S P; S S S S S

1.
3.2
3.3 Moth HN•delinz
1liple earure-Primary S S
S St
S S SS SS S
S
4..7 Ee ndEeld ovnt S P P_ P _ S P S
4.1.2 ,GsRi.b.l ..... off e s _ ss s _
ss ss P

4.1.3 EE i S S S S s S P

4.1.53 _an.riin S S S S S S
4.1.6 C ban Eyeli Movem:ent ...
4.1.7. Eye Ss S S
P S
P '" Ss rS P
s
4..1.8. Pupillary Dilt,, e P S P P Sn
4.1.9
4.1.4~ Muascle Tens~ion,
c Rate,
PEterHo~artTremor
Anyi S P
S l-S S _ SS tS SS 1 .S
______
.S
4.1.10" Hert Rate S S Pt

4.1.11 VariSbilaty,
Drea.thinit Ana.lysis Pressure
wlood d toS S S t S. S S P
4.1.12 Body.P lu Analysis
Pt
Analysis rlS SS
4.1.13 Handwriting P PS SS !P S SS Scn
S
4.2 Comibined Physiolos-ical Mea~ure S .S' 8 " S S P i
4_T.3 Speech Pattern Analysis ,S S S SS S

II
Weizhtftgt
S. Solvable without, difficulty; Problem doep not exrist.
S~P% Potential Problem; Dt/!ficulty vf-ll be encountered,
I

I
Ir

I.
I
I:

F
 WORKLOAD ASSESSMENT MITHODOLOGY DEVELOPMENT

by
Billy M. Crawford
Systems Research Branch
Human Engineering Division
Wright-Patterson AFB O 45433

The Workload Problem

During the development of advanced man-machine systems a number of important questions must be re-
solved. Many of them relate to human performance or manning requirements. for example: How much atten.-
tion is required by operator tasks? Which tasks can be Asesgned to a single operator? How long can an
operator ?erform his task effectively without a rest bre.zP." V.ow much learning or traiuing is necessary?
What is the mininum crew size for a system? How will time p-ure and other stresses affect task and
ultimately mission performance? All the foregoing questions relate to performance and workload.
The designer/planner, based on his appraisal of tha possible contingencies, typically attempts to

minimize the frequency, extent and seriousness of work overload situations. However, he can neither
personally nor vicariously, through others, rigorously assess the workload, or potential workload without
a standard metric for adequately defining and quantifying it. Even if he does identify particular periods
of potential workload excess, he dces not, except in extreme and obvious cases, have quantitative infor-
mation to assist in deciding which of the instances are the most critical and demanding and hence should,
within resources and technological limitations, be given priority consideration in design. Nor does he
have a criterion by which he can decide and demonstrate that the problem has been reasonably resolved.
In the development laboratories, alternative proposed designs or arrangements, or alternative proce-
dures, may be compared on the basis of speed. accuracy, or errors. However operationally significant
differences may not be revealed simply because the subjects are able to, and do, mater their resources
("try harder") and thus compensate for what would otherwise be real differences.
Work overload at the mental or "cognitive" level has been associated with increases in the United
States Air Force aircraft accident rate (Miholick, 1978). For example, during 1977 and 1978 "channelized
attention" or "distraction" were factors in 16 accidents involving the loss of 12 aircraft, 9 fatalities,
and a dollar loss of over 81 million dollars. "Task saturation" which results in intense concentration
on the task perceived to be most important at the expense of other critical performance requirements wa&
classified as "channelized attention." "Distraction" was used to refer to occasions in which an unexcepted
task causes attention to be diverted to coping with the cause of the unscheduled task load.

If we are to make safe, economical use of human and material resources it is necessary to determine
efficient crew compositions, appropriate assignments of duties and responsibilities to crew members, and
effective allocations of functions and tasks among men, machines and computers (including software). In
addition, it is necessary to identify the critical periods in a task cr mission during which the operator's
performance is particularly prone to degradation or failure because of work-overload stress. Further, it
is necessary to provide improved, valid and quantitative methods for assessing equipment and system design,
and procedural alternatives; and for mission planning and survivability/vulnerability analyses, to locate
and quantitatively define the most critical aud demanding task segments. In a parallel view, it is
necessary to identify and quantitatively define those periods, if any, of sub-optimal workload stress so
that the resources can be used elsewhere, or so that provisions can be made to preclude or alleviate
boredom, loss of "sharpness" or alertness, etc., the effects of which can carry over to anJ jeopardize
performance in subsequent tasks or mission periods. Due to the rapid advances in computer technology
and the more centralized role computers assume in advanced systems, emphasis probably should be on man-
computer interactions and information processing/decision-iaking functions which are not adequately
accounted for by conventional huann performance metrics, task analysis, time-and-wotion, and time-line
methods.

The principal objectives of a supportive workload research and development program should be (1)
establishment of a set of theoretically-consistent component functions descriptive of the performance
of crew members in relevant system tasks; (2) development of quantitative (mathematical) expressions of
relationships between input-output parameters for the component functions and appropriate combinations
thereof; (3) integration of the results of (1) and (2) above into a task analytic/computer modeling
methodology; and (4) validation of the analytic/predictive methodology in a system design, development
and test effort. Examplea of approaches and metods contributing to achievement of the above objectives
follow.

Adoption of a Workload Concept

Ryan (1947) addressed the problem of measuring the cost of sedentary, or "mental," work some 30 years
ago in his text on the psychology of production. His concept of effort, presented in the same context, is
similar to the concept of workload as it is used today. Ryan identified four possible meanings for effort:

(1) energy consumption,

(2) cost of work (e.g., fatigue, loss of health, dissatisfaction, etc.),
(3) aspects of psychological functioning which describe the "experience of the worker as he performs
his job, and
(4) the rate of performance of an individual in relation to the maximum possible rate of performance
under the given conditions.
 Ryan indicated his preference for the latter (fourth) meaning of effort probably because it required
that task performance be not only uaansured but also related to the capacity of the worker. The discussion
of topics which ensues is based upon the assumption that effective resolution of workload problems depends
upon the capability to measure, by a common metric, both task demands deriving from work situations and the
inherent capabilities of the worker to meet them.

Performance Theory Development and Application

'inorder to progress in an orderly, systematic manner, It is necessary to explain and relate per-
tinent facts in a logically consistent manner. Current human performance theo.-y can serve that function
in a workload assessment program. The primary goal of human performance theory is to analyze human
capabilities in a manner which will permit (1) identification and description of basic, component func-
tions and (2) quantification of the limits of capacity in each component function. Theories which treat
the human as princ--!.l.- 4, Nformation processor of limited capacity appe4r to be most appropriate.
Some of the rese.7.:.hich have been associated with the development of such a theory are
revealed by the to•.1- 4 pes" of theories:

(1) Sirgle Channe.i Theory (Welford, 1952; Broadbent, 1958). The human is strictly a "serial"
prt.pe-ssor.

(2) Undifferentiated Capacity Theory (Moray, 1967; Kahneman, 1973). The human behaves much like
a time-sharing computer with task interference strictly a function of total demand rate rather
than specific to the nature of the processing tasks competing for capacity.

(3) Limited Capacity Central Mechanism Theory (Posner and Keels, 1970). Some, but not all, pro-
cesses require the "central mechanism"; hence, parallel, as opposed to serial (Single Channel),
pr~cessing is sometimes possible.

An example of current theorizing based largely on the single channel concept is that of W. H.
Teichaer. For the past several years various U.S. Government agencies sponsored efforts of Teichner
to develop a general theory of human performance. The goal was a systematic approach to prediction of
human peformance as a function of task variables and environmental factors (Teichner and Olson, 1971).
Teichner drew heavily upon the available experimental psychology and physiology literature to identify
empirical relationships and develop models of simple tasks which could be combined into a more compre-
hensive model or theory or used in predicting performance in more complex tasks (Teichner, 1974).

Based on observations uf people engaged in a large variety of work situations, Teichner concluded
that the same general functions comprise the various human activities involved; hence, the feaiibility
of modelling any human activity in terms of a finite set of generic subtasks. Teichuer and Olson held
that tatks always involve a transfer of information from an initial input to a final output. In other
words, the human is a system which functions through a series of communication links and subtasks and
that system is the same whether flying an airplane or dialing a telephone. No matter Low the man-machine
system context varies, at a given level of human system analysis the only differences will be in the
activity or dngree of loading of the subtasks.

Teichner's theoretical approach is consistent with human engineering and system analysis tradition
in referring to man and machine as components of man-machine systems. Any operation on information
wi.thin a component, whether man or machine, Is called a "process" whereas transfers of I.1.formation
between components are called "tasks."

Although both the maximum complexity and maximum capacity of the human are constant according to
Teicbnerisn theory, system capacity may be varied in a number of ways. For example, since operations
may be performed by different combinations of available generic subtasks, it may be possible to replace
the limiting function in a serial process with a higher capacity subtask. Or, the system may be
redesigned for parallel processing at the limiting stage by allocating the function to another component,
e.g., a machine or another person. Assuming the human Is a single channel system, the maximm processing
rate can be no greater than the capacity of the lowest capacity stage in a sequence, of course.

In developing his performance theory, Teichi er bypassed the task taxonomy problem and went directly
to empirical relationships and principles uhich vould be used to predict dependent measures. The theory
builds upon Donders' Law which is based on data obtained from attempts to measure the physiological time
of mental processes associated with discrimination and choice in 1868 (Woodworth and Schlosberg, 1955).
Donders' Law simply states tiiat choice reaction-time (CRT) is composed of simple reaction time (a constant)
stimulus categorization time, and response selection time.
Teichner initially modified Donders' Law as follows: (1) Stimulus identification time was included
in simple reaction-time; (2) Stimulus code-to-response coda translation time (Tsor) was substituted for
the response selection component; (3) Stimulus code-to-stimulus code translation time (Tso-) was added to
account for tasks in which it was necessary to transform one stimulus code to another before selecting a
response; and (4) another component (c) was added to cover time required to select the motor program for
executing the response. The resulting equation was:

CRT - a + To-s + Ts.r + c

in which "a" includes both stimulus encoding time and neural transmission time.

Teichner adopted & response criterion model proposed by McGill (1963) and nrice (1968) in order to
Saccount for empirical evidence that the "a component" of the above equation depends upon stimulus
intensity and duration (Teichner and Krebs, 1972).
 57

Teichner proposed to use coding theory and information "etrics to quantify S-S translation. Two
examples of S-S translation are compression and classification. Compression in exemplified as follows:
Assume a four message, binary source encoded thus: 0001, 0010, 0100. and 1000 with equal probobilities
of occurrence. Compression could be achieved by recoding, e.g., 00, 01, 10. 11, with no change in
message probability. The average value of the original, or source code (L.), is 4 bit@ per message as
compared to 2 bit* per mesage for the recoded mssuages (Lc). In coding theory, the average compression
for a sequence of symbol@ is called the compression coefficient and is represented by the equation:
a - Lc/Ls. The value of stimulus compression is in its effect on the S-R translation process. Because
there is less information in the compressed message, the S-S translation should involve less time and
error. Obviously the lose resulting from the compression process must be less than the gain at the S-R
stags for it to be worthwhile.
The second form of S-S translation identified by Teichner is classification which results in a
reduction in toe number of messages S-S classification is exemplified as follows: Assume a message set
of four, e.g., F1, F2, 31 and B2. This message set may be sorted into F (fighter) and B (bomber), a case
of four-to-two sapping.

It can be seen that the cost effectiveness of S-S translations as described above is assessible in
teras of chanbes in the information transmission rate (R) achieved for CRT tasks. The cost effective-
neAs Index for c.vapression (CEc) is CEc - R/m. CEc is the rate of information processing per unit of
cfzmprecaLin. The cost effectiveness of reduction in messages through classification (CEr) is expressed
hs follo0: CEr = R/•c/Hs where Hs is the amount of information in the original message set and He is
the amount of information in the set after classification.

The same cost effectivensts concepts say be applied to the S-R translation process, in which case
the recoded message is a response and is defined by a response code. Again, the impact of reduction or
compression is expected to be greater speed and accuracy of response selection.

Teichner clearly distinguishes between response selection and response execution. It is assumed
that responoes are always defined symbolically by response codes. Only after the appropriate response
code has been matched with the stimulus code does the associated motor response b"gin. The ensuing
response execution may entail a series of effector selections whether the response modality is limb
movement, body movement, or speech. Execution time will depend on factors such as distance travelled,
amount and direction of force exerted, etc.

Teichner's thinking toward a complete theory, or model, of performance is represented by the flow
diagram in Figure 1. The "a" component of his equation for CRT derives from a combination of sensory
register of his equation for CRT derives from a combination of sensory register and scanner functioning.
The flow diagram shows that the response criterion applied by the scanner derives from long-term
stimulus memory (LTN-S) which also establishes operating levels for activating systems and scanner rate.
LTM-S also provides for selective tuning of the register so that thresholds of "energy cells" for expected
stimuli will be lower than for unexpected stimuli. Sensory register properties are derived from Hubel
and Wiesel (1962).

Teichner hypothesized that the human component obtains information transmission rates consistent
with system demands by making speed-accuracy tradeoffs in the following manner. At the input stage,
with experience at a task, an individual learns what stimuli to expect, how much stimulus evidence
is
required to respond, what sampling rate is required, and makes sensory register/scanner adjustments
consistent with task demands relative to speed and accuracy. Dependint upon the operations involved,
a range of speed-accuracy variations may be available at the S-S and S-R translation stages. And,
finally, at the output stage the response criterion say be adjusted upward or downward to favor either
accuracy or speed depending upon the information transmission demands of the system.

Hahituation is handled in a way consistent with Sokolov's (1963) neuronal model. When a novel
stimulus passes to the S-S translation stage and cannot be matched with a relevant event in LTh-S, the
responsive register cell is tuned toward increasingly high threshold levels on successive occasions.
When a stimulus event is detected by the scanner mechanism, a correspuading unit of short term memory
(STh) is activated for a duration of time (e.g., 30 seconds) during which comparison can be made with
LTN-S in support of the S-S translation. Teichner suggests that several available models are consistent
with the latter process (Norman, 1970; Saunders, Smith and Teichner, 1974).

The importance of Teichner's theorizing to workload assessment rests largely in its potential impact
on task analysis. Traditional human engineering task analyses provide an overwhelming amount of detail
almost totally unrelatable to available theoretical concepts and principles. Part of the difficulty is
attributable to the fact that the conceptual frames of reference tend toward anatomical rather than
functional task descriptions. Teichner's goal was to systematize the description of operator tasks and
performance at a generic level consistent with both the environment/performance literature and the
operational situation. An attempt to verify the applicability of a portion of Teichner'h theory for a
system simulation will be summarized in a later section.

Theory Testing via the Divided Attention Paradigm

Data, buch as that obtained by W. E. Hick (1952), relating reaction time to the amount of information
transmitted, and to the degree of stimfulus-response compatibility (Garvey and Knowles, 1954), caused the
idea that independent associative links exist between each stimulus and response to be replaced by the
concept of a mediating limited capacity central mechanism (or system). The single channel interpretation
of this system (Welford, 1952, and 3roadbent, 1958) holds that a signul entering the system dominates the
entire channel from the time it was selected until the response is initiated. Any other contending
signals are either filtered out or held in store and gated into the channel after the response to the
previous signal. Increase in response time for each unit of information transmitted provided measures
of the processing demands a signal places on the limited capacity system.
58

However, additional research, principally task interference studies, suggested v need to modify the
single channel concept. While the Pingle channel, or aerial processina, model requires that the time to
perform two tasks simultaneously should equal the total of the times required to perform each tas1• alone,
sometimes it is found to be much iess (Keels, 1967). This suggests that in such instances soue components
of the separate tasks may be processed in parallel and, hence, do uot require exclusive use of a single
channer. mechanism. Attempts to account for this apparent parallel, rather than serial, processing led to
the two alternate theories.

One of
ferences the alternatives
between tasks occurs isonly
thewhen
general, undifferentiated
the total capacity theozy
number of non-specific which holds
"processing units"that
is inter-
1xceeded
by the demand. That is, task interference is not specific to the peculiar nature of competing task
components, or operations, involved, but simply reflects an "overdraw" on the available pool of capacity
units. Moray (1967) modified this interpretation somewhat by hypothesizing a limited capacity processor,
similar to a time-sharing computer, which allocates from its undifferentiated processing capacity amounts
consistent with the demands of operations performed on the signal.

The second alternative to a single channel theory derives from the proposition that some, but not
all, operations performed by the human information processing system are channeled thr'ugh the limited
capacity central mechanism (Posner and Keels, 1970). Thus, onerations which do not require the mechanism
may proceed in parallel without ever interfering. While it has been suggested that the limited capacity
mechanism may be either a single channel or a parallel processing system which processes multiple signals
with reduced efficiency (Kerr, 1973), it may be that there are several limited capacity mechanisms each
of which is peculiar to a particular type of signal, sensory mode, or operation. It has been suggested
that the amount of interference between operations depends upon overlap between factors such as verbal
or spatial demands (frooks, 1967; Ailport, et al, 1972). Perhaps, after the fashion of Spearman's theory
of intelligence, there are central mechanisms peculiar to each of several "specific factors" whereas
operations of a "general" nature are processed in parallel. (Incidently, Teichner preferred a serial
processing model and was confident that he could account for any apparent contrndictions before his
theoretical development was complete.)

Divided attention effects produced by requiring subjects to attempt two tasks simultaneously provide
an excellent basis for evaluating hypotheses generated by any of the three variations of limited capacity
theory. This fact has been recognized by several theorists. The result has been a proliferation of
secondary tasks beyond the rather large number produced by engineering psychologists during the 1950's
and 1960's. During the latter era, numerous researchers tailored secondary tasks for compatibility with
primary tasks and used them to evaluate the efficiency of alternative procedural or man-machine interface
designs. Although the results were valuable to the specific applications, they made few contributions
to a basic understanding or quantification of human performance capabilities and limitations because of
the lack of standard methods and metrics. There were obvious practical reasons for that leficiency which
have been identified and discussed by Knowles (1963).

There is also some justification for using a variety of secondary tasks in exploring issues derived
from the limited capacity mechanism theories. However, the Sternberg task and associated model of in-
formation processing stages hold a great deal nf promise as a more or less standard approach to both
theory testing and reserve capacity measurement (Steinberg, 1969). In addition to its pouar in theory
testing and development, which has' been demonstrated by the late George Briggs and his associates,
primarily under sponsorship by the USAF Aerospace Medical Research Laboratory and the Office of Scientific
Research (Briggs, et al, 1969, 1970, 1972), the task prcvides a method for assessing reserve capacity for
a vatiety of workload situatfons. Although relatively simple and readily learned, the Sternberg task
facilitates manipulation and control of three key functions in information processing/decision making
tasks: (I) input, (2) central processing, and (3) output. Both input and output are readily quantified
in information metrics--a common measure to biologists/neurophysiologists, behavioral scientists, com-
munications and computer system engineers and, hence, potentially P boon to effective system engineering
including associated man-machine tradeoffs and functions allocation. Moreover, the Sternberg task is
amenable to variations in stimulus (e.g., visual, auditory, tactile) and response (manual, vocal) mode
making it adaptable to a variety of dual task situations.

The Sternberg task is a choice-reaction task which facilitates manipulation of the loading at
Stage 2 (Central Proceesing) while holding the requirements on the other stages constant. Stage 2
loading is varied by changing the number of "positive set" items (e.g., letters, digits, tones) the
uubject must maintain in memory. In performing the task, a subject listens, or watches, for a stimulus
cue, or memory "probe," while maintaining a readiness to respond via a response device "yes" or no
depending upon whether the cue "matches" or "does not match" an item stored in memory.

In applying the Sternb3rg task to the study of divided attention effects, the Sternberg task is
first administered alone to obtAin "baseline" data for 3 or more different "memory loads," e.g., 1, 2,
3 and 4 items. The resultant data (using correct responses only since incorrect responses are held to
a "negligible" level) is used to plot reaction time (on the ordinate) against memory load (on the
abscissa). A linear equaticn is fitted to this data plot to obtain a straight line with a particular
slope and y axls intercept value. Thus, the intercept reflects time required for Stage 1 and Stage 3.
The slope of the line reflects central processing time, i.e., Stage 2. Then, by requiring subjects to
perform the same Sternberg task simultaneously with a second task, which is treated as the primary or
priority task, one can acquire information relative to the nature and amount of workload imposed by
the second task. For example, if the slope of the equation for the Sternberg data plot changes between
the baseline and dual task conditions, the second task imposes significrnt demands on Stage 2 or central
processing. If the intercept changes, the demands of the second task occur at Stages 1 and/or 3. The
amount of change involved can be quantified in terms of the information metrics, bits and bits/sec., to
obtain an indirect Indicat:'on of workload associated with the task under study.

Ttie utility of the Sternberg task is readily apparent from a review of the research program pursued
by Briggs and his associarcs at Ohio State and New Mexico State Universities. Bri.gg's research centered
around efforts to isolate divided-attention effects within one or more of rhe four possible stages of
 tL.: •ith (1968) task paradigs: (1) encoding processes, which entail registering, sumpling and
preprocessing of stimulus information; (2) central processing (detailed analysis of sampled informatiou
r for stivulus identification and definition); (3) response decoding, and (4) response control and execu-
tion. The essence of the research progra results are, perhaps, sImarized most concisel7 by tracing
the progressive expansion of the function proposed by Sternberg (1909) for describing the relationship
between choice-reaction time (CRT) 1 and the size of the positive memory net. (The reader should recall.
that the Sternberg technique requires that a subject first semorite a set .f items of soie H4. Then, at
a later time, an item, or "probe: is presented and the subject respond& as to whether the item is. or is
• • not, a mumber ofof the
presentation the "probe" itemoruntil
mmorised, "positive: set. The
the response major dependent measure is the time (CRT) from
is executed.)

Sternberg expressed the f~inction so:

RT - a + b(K)
Data collected by Swanson and Briggs (1969) showed a logorittmic relationship between CRT and memory
load which led to the postulation that response time is a function of central processing uncertainty
(He), a metric from information or coemunication theory (Shannon, 1949). Hence, Sternberg's expression
was modified to read:

Subsequently, Swanson .nd Briggs (1969) demonstrated that the intercept constant (a) was linearly
to the aqount of informtton transmitted (Ht), a comunication theory metric of response accuracy;
Selutsd

thus, the exprescion became:

iT - c + d(Ht) + b(Lc)

An experiment by Briggs and Blaha (1969) suggested that b could be expressed in ters of the number
of displayed items to be classified (D) and the equation was modified again, thus:
RT - c + d(Ht) + e(Hlc) + f(HcD)

-Brigs and Swanson (1970) next varied the response load (R) in an experiment. The results showed
that it could be partialed out, thus quantifying still another component of performance and the expres-
was now: Ssion
RT -i + j(Ht) + h(R) + e(1c) + f(HcD)

By relating this resultant equation to the Smith information processing paradigm, Briggs (1972)
made estimates of the time required for specific functions of the human information processing system.
For example:

Preprocessing ;ime: 180-280 mac

Stimulus sampling rate: 6.5 bits/sec

Recoding (Teichner's s-s trarslation): 25 bits/sec

Transfer from long-term wm4ory to active memory: 39 bits/sec

Stimulus classification: 16 bits/sec

Response Decoding: 6 bits/sec

This is the type of quantitative information and generic classification scheme which ie needed to permit
the desired state-of-the-art advance in analytic/predictive methodology to effectively complement task
and time line analyses during system design. Of course, a great deal of theory development and testing
remains to be done.

In 1974, Biggs, Johnson and Shiner took a step toward integrating the Sternberg/Smith information
processing paradigm with fundauental decision-making research by using a Bayesian decision expression
to account for the sequence of decisions made by a subject in a classification task. Thus, the link
I
has been established between simple choice behavior and more complex decision processes to suggest
something of the potential for expanding and validating basic performance theory applicable to critical
comand-control-cozmunitation system design issues.

Workload Assessment as an Aid to Design

Queatio:.s concerning the impact of digital avionics for pilot workloaa have provided an opportunity
for preliminary tests of both performance theory and the divided attention paradigs in an applied setting
(Crawford, Pearson and Hoffman, 1978). The opportunity developed as follows:

The evolution of compact digital computers has ende possible the development of digital avionics
infcrmation systems. Such systems promise a number of advantages to both aircraft designers and users.
For example, when interfaced with multipurpose cathode ray tube displays and multifunction switches,

[. 1See
r oodorth and Schlosberg (1955) for a review of Donders classic research on simple and disjunctive
 Sf
digital computation and storage capaLilitien can be used to reduce the number of dedicated instruments
competing for cockpit panel area. Information which is not required by the pilot on a continuous or
frequent basis can be stored and presented ,n demand cither automatically, as related programmed mission
events transpire, or in response to manual control actions (Zipoy. Presselaar, Gargett, Wlyea and Hall,
1970). And with reduced demands for panel space, it will be easier to locate the multipurpoas controls
F and displays in prime reach and viewivi areas.

However, experienced pilots are troubled by the prospect of possible added activity--both mental and
physical--required to gain access to information which is normally on dedicated instruments. Should the
demand for such activities occur during peosioperator workload, the impact on mission success might not
be offset by the increased calculating piwer, speed, or accuracy afforded by the digital processor.
Hence, a study was planned to nyaluate the impact of multipurpose control/display tasks on the pilot's
r.eerve capacity. Of partucular interest was the question as the whether or not the maintenance of know-
lndge of procedures associated with multifunction keyboard operation reduced the operator's reserve
capacity for making choices or decisions such as might be required to handle contingency situations during
a mission. Another purpose of this study was to investigate the compatibility of keyboard operations with
continuous flight control tasks.

A computer-based simulator was used to present and score the task situations investigated (Brandt and
Wartluft, 1975). Of the three different tasks involved, two, flight control and comsunications/IFF
switching functions, represented actual tasks in aircraft systems. The third was a variation of the
Sternberg task which served as a test to measure cognitive reserve capacity under various primary task
conditions. All three tasks were implemented within a fixed-base cockpit simulator.

The front panel of the cockpit was equipped with three CRT-type displays. The center display was
used to present information concerning basic flight parameters in a moving tape format. The cockpit also
contained a throttle with afterburner switch (left side panel) and a center-mounted joystick control which
were used, in combination with the displayed flight information, to "fly" various maneuvers. Printed
computer outputs of simulator performance data included both mean absolute and root mean square error
relative to specified control values based on "fly to" instructions for altitude, heading, bank angle,
pitch, indicated airspeed, vertical velocity, angle-of-attack, and g-load.

Between the front instument panel and left side panel was a multifunction keyboard (MW). This
MFK, in combination with the CRT on the upper left of the frost panel and a numerical entry keyboard,
which was also located on the instrument panel (lower left), was used to simulate a multifunction inter-
face with digital avionics subsystems. Subsystems, functions and states were displayed on the CRT to
complement the feedback afforded by back-projected legends on the MFK push button faces.

The Sternberg task procedure used in this study was as follows: At the start of an experimental
session, the experimenter read to the subject a set of 3, 2, 4 or 6 letters of the alphabet. The subject
was asked to retain the set in memory during the succeeding block of trials. The tour sets used were as
ioilows: A, AlH, AHJQ and AHJQSX. (Such sets are referred to as "positive sets.") During the block of
trials the subject was presented (via a cassette tape player connected to his headset) a series of tast
stimuli or "probes" to which he was to make one of two responses: (1) "yes," the test stimulus matches
the positive set, or (2) "no," it does not match, ani, hence, is a member of a negative set. ihe negative
set included the 9 letters, 1, C, E, F, G, I, L, R and Y. Negative and positive stimuli occurred with
equal probability (.5). Letters within the two sets also occurred with equal likelihood. The average
inter-stimulus interval was 5.5 seconds and ranged from 3 te 7 seconds. "Yes" was indicated by the sub-
ject's pushing forward on a thu.mb switch on the joystick controller used for flight control; "no" was
indicated by moving the thumb switch backward, i.e., toward the cubject. Reaction times were scored
automatically to the nearest millisecond. If a subject did not respond wit-hin 2 seconds the trial was
scored "no response."

Central processing uncertainty (Hc) values for this study are: 1.00, 1.50, 2.00 and 2.31 bMts for
the 1-, 2-, 4-and 6-item vemory sets respectively. Because there is always 4 2-choice response, response
uncertainty (Hr) - 1.0 bit in each instance (Attneave, 1959).

Four male subjects were used in the study. They were paid volunteer university student, with an age
range of 20-24 years. During the experiment a nominal cash incentive system was implemented to encourage
performance. The amount of the incentive was based on the subject's relative standing in the group with
respect to task performance criteria for each session. For dual task conditions the incentive value was
weighted so as to emphasize priority for the flight control task when it was present. The incentive was
weighted in favor of the MFK task when it was paired with the Sternberg task.

Prior to the experiment proper each subject was trained on All three tasks. Training sessionb lasted
two hours and were scheduled 2-4 times per week. Each subject was trained until task performance measures
appeared to asymptote. Then each subject was tested under six different conditions: three single con-
ditions and three dual task conditions: Flight control, MFK and Sternberg choice-reactton task,alune;
and flight control plus MFK, flight control plus Sternberg task end MFK plus Sternberg task. When the
Sternberg task was combined with MFK, it occurred only during periods when the subject was awaiting
instruction for an MFK task of a 6iven difficulty level. This was consistent with the interest in mea-
suring cognitive loads associated with anticipation of MIV tasks rather than actual performance of them.
* The single task conditions preceded the dual task conditions for all subjects.

The four levels of MFK task difiiculty investigated were quantified in terms of the number of bits
of information transmitted via the keyboard in performing the tasks. The average value for each level was:
1-7 bits; II-11 bits, 111-17 bits and IV-26 bits.

The type of maneuver "flown" was the independent variable for the flight control task. Although
seven maneuvers were flown, preliminary anal)ses showed that not all maneuvers were discriminable in terms
*r of the weighted tracking error scores. Hence, the maneuvers were combined into two groups labelled "easy"
and "difficult." "Easy" maneuvers included straight and level flight and level turns. "Difficult"
 61

maneuvers were climbing and diving turns. The error scores (X) were comprised an follows: X - (0.01) A
altitude + (0.1) A airspeed for straight and level and stall, X - (0.01) A altitude + (0.1) A airspeed +
(1.1) A g-load for straight and level turns, and X - (0.005) A vertical velocity + (0.1I) A airspeed + (1.0)
A g-load for turning divas and climbs. The delta values represent average error, i.e., deviation from the
prescribed "fly to" value for the given flight parameter, per unit of time on the task. Altitude was mea-
sured in feet, airspeed in V.ots and vertical velocity in feet/minute. The flight parameter combinations
and associated weights for each maneuver type were based on pilot opinion and research findings summarised
in a separate report (Woodruff, 1972). MK performcnce on multifunction keyboard tasks was measured in
terms of task time and errors. The dependent measure for the Sternberg task was reaction time. Errors
14and failures to respond within two seconds were also recordeA.

A simple analysis of variance (repeated-meaaiures design) was applied to the ,|cores for the flight
control single task condition. The difference between easy and difficult conditions was statistically
significant (p < .05). The man and standard deriations for the easy condition were 1.09 and 0.17.
Correspovding values for the difficult condition were 5.11 and 1.51.

The effect of FK task difficulty proved significant statistically (p < .001). Mean task times
(seconds) and standard deviations (in parentheses) for the four d.fficuLAy levels were. 1-3.97 (0.32);
11-5.95. (0.53), 111-7.43 (0.68), IV-9.87 (0.83). The average rate of infkrmt1in transmission via
the 1FK system varied from 1.8 bits/sec. to 2.6 bits/sec. across the four levels of MPK task difficulty.

The method of least squares was used to fit a straight line to the Sternberg data. The result is
reflected by the following rEgression equation for the single task, or baseline, condition:

RT - 549 + llS(Hc)

Although mean flight control errcr was greater when the flight control task was combined with MIK
tasks, the differences were not statistically significant. Similarly, HFK task timev increased under dual
task conditions, but the differences were not statistically significant. Flight control error scores were
virtually identical for flight control alone as compared to flight control with the Sternberg task. The
St'.trnberg task had no statistically significant impact on 1K task time.

The method of least squares was used to fit linear equations to Sternberg response time data for each
dual task condition. This permits comparison of intercept and slope values with those obtained for the
Sternberg task baseline condition, for the purpose of localizing divided attention effects within the four
stage information processina model.

Preliminary analysis showed no significant diff-rences between levels of UFK task difficulty in terms
of slopes and intercepts. Hence, a single regression equation was derived for the combined UFK levels.
Equations for the resultant three dual task conditions are as follows:

Sternberg with MK "Rehearsal" RT - 617 + 1l8(He)

Sternberg with Easy Flight Control RT - 694 + 98 (Hc)
Sternberg with Difficult Flight Control RT - 855 + 31(Hc)

F-tests (Snedecor and Cockran, 1967) indicate that (1) slopes and intercepts for the flight control
conditions differ signific&ntly from those for the baseline condition, and (2) the intercept value varies
significantly between the baseline and NFK implicit rehearsal condition.

Itetarpreted in the traditional manner, the preceding results indicate that the effect of UK "impicit
rehearsal" is in the input or output stage of information procensing only. Following the empirical evi-
dence and logic of Briggs, et al (1972), the effect is probably in the input stage. The difference in
intercept values amounts to a 12% average increase in input-output time attributable to MNK "implicit
rehearsal."

Active flight control, on the other hand, appears to impact both input and central processing as
evidenced by differences form baseline in both intercept and slope values for the regression equation.
Moreover, there is an increase in input-output time (28% and 55% for easy and difficult flight control,
respectively) and an increase in central processing rate. The central processing rate for the baseline
condition is 8.47 bits/sec. as compared to 10.20 bits/sec. and 32.26 bits/sec. for the easy and difficult
flight control conditions respectively. This increase in central processing rate under the dual task
condition is consistent, with results obtained by Lyons and Briggs (Briggs, et al. 1972). It was attributed
to the subject's conducting fewer or less complete tests of the probe stimulus under the greater loading
conditions. This apparenL switch in mode of operation in the central processing stage may prove to be a
valuable aid to identification of significant workload changes.

The observed variations in Sternberg task response accuracy suggested the appropriateness of further
information analyses, i.e., calculation of the average amoUwt of information transmitted (which would
reflect all the data, including erroneous responses and no responses. These values for the baseline and
two levels of each dual task condition are presented below.

AVERAGE INFORMATI3N TRANSMITTED (Ht) IN BITS FOR STkRNBERG TASK

H

Condition 1.00 1.50 c 2.00 2.31

Baseline .85 .85 .88 .41

Easy MK .86 .85 .94 .42
Difficult UFK .84 .88 .86 .32
Easy Flight Control .82 .77 .79 .27
Difficult Flight Control .72 .72 .79 .26
- 62

These data clearly indicnte that the 6-item memory met (Hc w 2.31 bits) produced an overload situation
for every task condition.
Effective Uncertainty Reduction. Since perfect performance is represented by Ht - 1.00 bit in each
instance, the above tabled values were taken to represent percentage of the information reduction task
effectively accomplished by the subjects. An information reduction task is defined as one in which the
amount of uncertainty associated with the response is lass than that associated with the stimulus (Coombe,
Daves and Tversky, 1970). Thus, using the measures of central processing time, a set of "effective
uncertainty reduction rates" were derived and plotted graphically as shown in Figure 2. Note the consis-
increane in efficiency as Hc goes from 1.00 to 2.00 bits with the overload effect at Hc - 2.11 for all
conditions. Further study of Figure 1 suggests that cognitive reserve capacity is reduced by 20, 31, 45
and 54 percent by the four primary tasks (easy MFK "rehearsal," control), respectively.

With regard to the design issue addressed by the foregoing study, it appears that tasks imposed by
multifunction switch concept places demands on the operator which may detract from the value of digital
processing capabilities in avionics systems. The concept necessitates the concentration of uncertainty,
normally distributed among the various dedicated instrument control/display interfaces, at a single
interface. Hence, uncertainty which is normally removed via separate controls and displays for each
subsystem/function has to be eliminated via keyboard actions on each occasion that the operator interacts
with the multifunction system. Thus, while the digitally-based MFK system is relatively efficient in
terms of action and information transmission rates, the tasks are generally more complex and take longer
than corresponding ones for dedicated instruments.

The NFK flight control simulation and data appeared to provide a good opportunity for evaluating the
practicality of general functions incorporated by Teichnerian performance theory. One of the more complex
MFK task sequences was selected for that purpose. The task ioavolved the transmission of 40 bits of
information via 13 steps or key actions. Tiechner's theoretical components were then "sapped on" to the
NFK task sequence. Then a second laboratory simulation was generated by using cards with symbols on them
to model the same set of theoretical task components included in the MFK task sequence. The card-symbol
simulat~on was used to generate a set of performance data using students a. the Universit) of New Mexico.
Although this effort wes inly exploratory and has not been formerly documented, reasonably good agreement
between task time means and variances was obtained for the two sequences. Mean task time for the card
task was 8.3 seconds as compared to 8.7 for the WQK task.

As a further step toward integration of performance theory, part-task simulations of operator

workloads and system performance, a computer programmed model of the 4to-bit MMKtask components was
developed using Systems Analysis of Integrated Networks of Tasks. (SAINT will be discussed in more
detail in a subsequent section.) Close agreement was obtained betwe'n empirical data from the cockpit
simulator and SAINT modeling output. One hundred-sixty interactions of the computer model produced a
mean task time of 8.8 seconds.

Real-time simulations of operational tasks, as described above, are an Essential part of the theory
development and testing process which muse precede the achievement of an adequate analytic. descriptive
and predictive data base to effectively support workload allocation in man-machin%ý systems.

Physiological Correlates of Performance

Another line of research promising significant insights into the basis of human workload capabilities
and limitations at the neurophysiological level as well as providing intermediate workload assessment aids
involves the measurement of physiological correlates of performance. In 1934, Luckiesh atzd Moss, lighting
experts, reported data on the relationship between heart rate Rnd illumination level for a reading task.
The data showed decrements in mean heart rate as a function of taGk duration; moreover, the lower the
lighting level, the greater the decrement. Luckiesh and Moss, interpreted the finding as indicative of
the greater amount of effort required under low light level condftios. However, M. F. Bitterssn (1948)
in reviewing the lighting research literature complettly discredited this notion of Luckiesh and Moss in
the following words: ".... everything wm know about cardiovascular functioning would lead to quite the
opposite conclusion, i.e., that heart rate is directly rather than inversely related to the cost of work.
Heart rate is positively correlated with metabolic rate which we know to be a direct index of energy
expenditure, and Hadley (1941) has found a positive correlation between heart rate and muscular tensiek
which Dr. Luckiesh himself accepts as an index of exertion in visual work."

Whether Blitterman was correct or not in his crfticism of Luckiesh and Moss, it is interesting to
note that they might have had a basis for appeal in the resesrch of a physiologist, Darrow, who took an
apparently corroborative position in 1939--five years after Luckiesh and Moss published, but prior to
Bitterman's review.

Darrow (1939) reported data to support his postulation that both noxious stimuli and mental ectivity
involving "associative processes" are accompanied by cardiac acceleration in contrast to atteution to
sensory stimuli requiring "no extensive association of ideas" which is accompanied by cardiac deceleration.

Twety-six years later, Lacey (1965), having reviewed a large number of related experimental findings,
rephrased and expanded Darrow's postulation by suggesting that behavioral arousal, electrocardiacal
arousal, and autonomic arousal are different forms of arousal and that the associated activation processes
reflect the intended aim or goal of behavior as well as its intensive dimension. In elaborating, Lacey
noted that an increasing number of psychophysiological experiments demonstrated that different stimulus
situations reliably produce different patterns of somatic response. Listening to auditory stimuli, looking
at pictures, tapping telegraph keys, warm and cold stimuli--each condition produces a different 2attern
of somatic responses (Davis, ec al, 1955; Davis, 1957). To illustrate, receptinn of external stimuli,
with no motor response required, produces a heart rate decrease concomitant with the move "typical"
increase in other autonomic responses, e.g., palmer conductance (Lacey, 1959; Lacey, et al, 1963; Obrist,
1963).
 63

Without going into a detailed review of evidence cited by Lacey with regard to underlying physio-
logical mechanism and the complex nature of relationships between the cardiac response and coctical
activity, perhaps it will suffice for the purpose of this general discussion to use Lacey's findings as
an indi::ation of thk potential value of physiological correlates of behavior as a relatively unobtrusive,
objective technique for analyzing task performance at the cognitive level and obtaining guidanse with
respect to the stage, or stages, at which work overload occurs for ar ' given individual, or group of
individuals.

Lacey and associates began by presenting eight "stressor-situations" in different orders to three
samples of subjects. The situations could be ordered along a continuum in that some required only
attentive observations of the environment, e.g., looking at un intermitt'ntly presented light, while
others involved increasingly greater amounts of internal cognitive functioning--retrieval of information
from memory and problem solving activity, as in mental arithmetic. The results consistently showed that
sensory intake was associated with cardiac deceleration and restraint of systolic blood pressure whereas
tasks at the other end of the continuum (internalized cognitive processing) produced large `.ncrease in
heart rate and blood pressure. On the other hand, respiratory rate and palmar conductance showed the
non-specific or nondiscriminant, actuation pattern consistent with Cannon-based "arousal" or "activation
theory." Thus, depressor-decelerative processes are associated with facilitation of environmental intake;
pressor-accelerative processes with filtering out irrelevant stimuli which interfere with central cogni-
tive functioning. This finding was supported by Obrist (1963) using a different sample of subjects and
different stimulus situations. Confirmatory evidence was obtained from additional studies which shoied
(1) attention to visual and auditory stimuli to produce cariac deceleration while respiratory rate
increased, (2) "thinking" to produce cardiac acceleration, anC (3) the more "analytic" the child, the
greater the acceleration (Kagan and Rosman, 1964; Kagan and Lewis, 1965; Lewis, et al, 1965). Moreover,
in reaction time experiments, Lacey has found that the greater the cardiac deceleration in anticipation
of the stimulus, the faster the uotor response.

In sumary, Lacey concludes that different fractions of autonomic, electroencephalographic, and motor
response are mediated separately by mechanisms which are clearly dissociable although they may be closely
related. He suggests that the biological utility of the Aissocietion resides in the capability of the
different fractions of responsa to influence cortical and subcortical functioning different, sometimes
opposing, ways.
Kibler (1967), in an Aer-space Medical Research Laboratory study effort, sought to bridge the gap
between applications and laboratory research on the different cardiac response-stinulus situation
relation h'ips by means of a vigilance experiment. The resultant data showed a positive relationship
between tpe extent of stamulu-oraented cardiac decelerati ardiactespon efficiency during a i 1/2 hour
vigil. The study i.as regarded as a significant step coward developing an independent measure of alertness

during vigilance taskc. Subsequently, an unpublished pilot study by Crawford and Bachert, also of the
Aerospar- Medical Research Laboratory, showed a trend toward increased cardiac dRceleration, and reducid
sinne arLhythmia (the tendency of the Lormal heart rhythem toward irregularity), as a function of decreased
signal-to-noise ratios, produced by adding clutter to a simulated airborne digitized radar return display.

In the labozatory, KalabOek (1971) has found sign.f.ca2nt reduction in sinus arrhythnia as a function
of increases in the signal rate in a perceptual motor task. Kalsbeek (1968) also reported data indicative
of reduced arrhythmia as a function of increased task demands in a fiight control simulation.

Cardiac data obtained from Navy carrier pilots flying missions over Southeast Asia showed average
heart rates to be substantially higher during launch and recovery than during bcmb zuns (Plattner, 1967).
These results were interpreted to mean bombiAg uas c less demanding task than take-off and launch, which
was somewhat aurpris'ng to the researcher,4 although not necessarily to all pilots. It is conjectured
that analysis of the specific stimulus-situations involved in accordance with Lacey's theor4ticai position
might have reverscd the interpretation.

Some attempts to use cardiac response measurement, In combination with a battery of other physio-
logical correlates of performance, ,heve proven less than satitfactory. One possible explanation for
difficulties recognized in at least one such attempt is the ftilure to differentiate between actual
workload and performance, i.e., removal of fligLt instrument information produced a decrement in flight
control performance, wnici, was interpreted as an overload condition; iut it also teduced the information
load, which if e~fectively processed woulA havc resulted In improved performance. Careful, accurate data
collection and analysis is also essential to effective use of physiological data within theoretical
contexts as posed by Lacev.

Nevertheless, the evidence with regard to cardiac response "situational-specificity," is judged to

be sufficient to warrant further investigatien of the measure under carefully controlled conditions
employing the Sternberg task and information processing paradigm to aesess relationships betvpen increasing
denends cf the various "stages," stimulus input, central processair, etc., as well as the transformation
processes (classification, conservation, condensation, creation, . cc.) at different demand levels. The
ultimate potential advantages to this program are st least two-.oid: (1) Increased validity o: the per-
formance theory developed and (2) a relatively unobtrusive, objective workload assessment technique for
use during actual system operations and during system simulations to precisely identify crew functions
which isquire automated aiding via digital processing capabilities.

Evoked potential measurement appears to be another technique with reasonable promise for facilitating
performance theory and workload assessment developments. Instrumentation for obtaining average evoked
potentials involves the attachment of electrodes to appropriate areas of the scalp in the same manner as

required to produce an Ergo. The continuous electrical activity so obtained is onducted through an
-omputer. The resultant measure of the nonrandom activity ib the average evoked response (Childers and
Perry, 1969'.

k•a k•.fn..*s..~J.A.a.r...' .*.kM .~~h.n

r64

This response-averas;ing technique, which a; hances the signal-to-noise ratio, also accurately
identifies speciflc psychological variables with Louponents of the EEG. A stimulus initiates a series of
physiological processes related to both perception and preparation for an overt behavioral response.
Analysis of the electrical activity between stimulus and response can provide usetul information concerning
factors such as the timing, process speed and anatomical location of physiological events associated with
the psychological processes involved. Cognitive and motivational as well as timiua and response variables
may be included in the experimental situations achieved via this arrangement (Vaughan, 196C).

Theoretical issues related to the limiting central mechanism and serial vs. parallel processiug
appear to be most amenable to investigation via evoked potential methodology. Th' value of evoked
potential measures as an aid to acsessment of workload under operational or system simulation conditions
is yet to be established. However, Weissman (1969) in promoting the use of average evoked potentials for
assessing the level that the technique has no equivalent when it comes to minimizing interference with
the subject. Hence, evoked potential measurement must be considerei possibly as an unobtrusive method
for workload assessment under flight test or operational conditions.

It has been suggested that a complete battery of psychophysiological instruments might include the
measurement of heart rate, electrical activity of the brain, muscle activity, skin resistance, blood
pressure, sinus arrnythmia, average evoked potentials, urinalysis, parotid fluid, pupillary response,
metabolic rate, oxygen uptake and ventilatory rate (Gartner aad Murphy, 1976).

However, because of the prevalent interest in cognitive or information processing!decision making

activities, the EKC and EEG domains currently have the greatest appeal as primary sources of psychophysio-
logical data and continued exploratory development.

Analytic/Predictive Methodology

The final thrust of a comprehensive workload assessment development effort must include the incol-pk-
ration of the results of proiucts of the thrust areas into analytic and predictive methods. First th2
performance theory and quant;.tative functional relationships between human input-output parameters w4.ll
have to be reflected in task analytic procedures. The purpose of task and analysis is to provide the
basic building blocks for sulsequent human engineering analyses during system design and development.
Task analysis entails the sp,;cification of tasks to be accomplished by human operators including the
behavioral requirements of the tasks, kinds of discriminations to be made, decision making, motor responses,
etc. From the task analysis estimates of error rates, time line projections and personnel aptitude and
training requirements must b! made.

Task analytic methodology as it exists today represents little more than the crude beginn-ng made
some 25 or 30 years ago. COLtically needed research required to appropriately expand end validate
esential behavioral informat:Lon has not been forthcoming. Consequently, job analysea are exp.octed tu do
more than they possibly can. Although analysts continue to break work into smaller elements to produce
the expected documentation, :I.t is largely a reductionistic effort without sufficient regard to the rean-
ingfulness of the behavioral elements (Bryan and Regan, 1963).
It is suggested that emphasis should be upon implementation of system models (mathematical an,' com-
puter simulation models) as analytic/predictive tools during system design. It has been said that the
sign of maturity in systems analysir. will be the development of useful models (Shapero and Bates, 1959).
The SAINT methodology promises to be a useful vehicle in achieving the desired advance in systems
analysis (W -tman, Seifert and Duket, 1975). (SAINT was referenced briefly in the earlier discuusion of
simulation w,.ich was primarily concerned with real-time, man-in-the-loop simulations).

SAINT consists of a symbol set for modeling systems and a comruter profram for analyzing the models.
SAINT includes the conceptual framework for representing systems which include discrete task elements,
continuous state variables and interactions between thet.. SAINT is not a :,odel. It simply provides a
framework within which any quantitatively expressed model, or models, ..iay be deanribed and exercised.
And, since it was designed for addressing human performance, in particular, withia systeu contexts, it is
potentially an ideal vehicle for integrating generic behavioral functions such as are advocated within
human performance theory. The resultant computer models of systems concepts could, then, readily
evaluate the probability and source of system/task demands which exceed operator, or crew, workload
handling capability.

In applying SAINT, systems are represented as graphical networks of task-activities with which one
or more operators interact. Each task is described with respect to how its performance relates to other
tasks within the system of inte-est. The graphical analysis Is then input to the SAINT cmiputer prigram
for automated performance asse. .ent. Using Monte Carlo techniques, the SAINT program permits simulation
of probabilistic task performance and precedence relationships while collecting estimates of system
performance at the same time. Capabilitins are included for simulating continuous or discrete system
state variables and their response to discrete control task execution and for dynamic modification of both
operator and system characteristics as dictated by internal or external rtiulated "events" (Kuperman and
Seifert, 1975). Thus, this computer modeling technique permits fast time evaluation of human engineerino
design alternatives, and other human factors, e.g., skill level, training and motivation, within system
contexts. However, it is just as dependent on a valid scientific base as conventional task analytic
methodology.

Preliminary attempts have been made to apply SAINT to current USAF system design problems. For
example, a SAINT model of the cockpit simulator used to investigate multlfunction switching and multi-
purpose displays for the Digital Avionics Information System Advanced Development program was developed
(Kuperman and Seifert, 1975). Model networks were developed for both conventional dedicated avionics
subsystem iustruments and the multipurpose controls and displays. Exercise of the model provided
estimates of performance within the limzts of available empirical data. Conclusions of the investigators
included: (1) The SAINT simulation techniques are readily applicable to piedictive modeling of new
concepts of man/machine interaction. (2) The techniques are appropriate to the study of the theories of
human performance and to evaluation of experimental metrics for their implementation.

. . . . . .
 65

Allport, D. A., B. Antonia and P. Reynolds, "Co the Division of Attontion; A D.sproof of the Single
Channel Hypothesis," Ortly Journal Eperismital Psychology, 24, 225-235, 1972.

)i' terman, N. E., "Lighting and Viiual Ifficiancy," Illutinatina Entineariins, 906-931, 1948.
Briggs, G. I., "The Additiviy PrinciplA in Choice Reaction Time: A Functionalist Approach co Mental
Processem," In Topics in Learning and Pe,-ormance, Acadumic 7res", N. Y., 1972.

Briggs, G. E., A. M. .'ohason and D. Shinar, "CentrAl ProceosinZ Uncertainty as a Determinant of Choire
Heactier Time," Newry and Cognitior, Vol. 2, 417-425, 1374.

JIriggs, G. E., G. L. Peters ond R. P. Fishe., "On the Lwcna of tht. Divided Attention Effects," Perception
and Psychophysics, Vol. Il, No. 4, 315-320, 1972.

"brigga, C. 1. and J. Blaha, "Mmmory Retrieval and Central Comparison Times in lformation Prc'aessing,'
JournaJ of Experimenta1 Psychology, Vol. 86, 296-300, 19'0.

Briggs, G. E., R. P. Fisher, S. N. Greenberg, J. J. Lyons, G. L. Peters, and D. Shiner, "Kulti-Task

Time-Sharing_ ligrements," AM•L-TR-71-10.5, -PAFI, Ohio, 19/2.

Broadhent, 0. l., Perceptlon and Communication, Prgamwn, London, 1958,

Brooks, L. R., "The Suppression of Visualization by Reading," Ortly Journal of Experimental PEychology,
Vol. 19, 289-299, 1967.

Bryan, G. L. and J. J. Regan, "Training System Design," In H. P. Van Cott nnd R. C. Kinkade (me.) Human
EnAineering Cui~e to Equipment Des_!gn, Chapte-: 13, Revised Editioro, U.S. Government Printing Office,
WAshington DC, 1972.

Childers, D. C. and N. W. Parry, The Human Visual Evokied Response, Charles C. Thouas, Publisher,
Springfield, Tllinois, 1969.

Crawford, B. M., Y. H. Pearson and H. Holfman, "Multi-Function Svitching and Flitvht Control Workjload,"
AMYL-TR-78-19, In. Proceedi:•gs of Sixth Sym-osium on Psychology in the Department of Defense, 1'SAF
Academy, April 1978.

Crawford, B. N., V. H. Pearson, and '..6). Hofft , Multipurpose D.'glta3 _rwitchitg and Flight Control
Workload, AJRL-TR-78-43, U-PAFI, Ohio, '97e.

Crossman, E.R.F.W., "Entropy and Choice Time: The Effect of Frequency Unbalance in Chnice Responses,"
_rtly. Journal of Experimenwal Psychology, Voi. 5, 41-51, 2953.
Darrow, C. W., "Electrical and Circulatory Responses to Brief Sensory and Ideational Stimuli," Journal
oi Experimental Psychology, Vol. 12, 267-300, 1929.

Davis, R. C., "Response Patterns," Tramosactions. N. Y. Academy of Science, Vol. 19, 731-739, 1957.

Davis, R. C., A. N. Euckwald, and R. W. Yrankman, "Autonom; sad Huscular Responses and their Relation to
Stimple Stimuli," Psychological Monographs, Vol. 69, 1-71, 1955.
Gartner, W. B. aud M. R. Murphy, Pilot Workload and Fatigue: A Critical Survey of ConcepLs anc Assessment
Techniques, NASA TN D-8365, Ames Research Center, Moffett Field, Calif., November 1976.

Garvey, W. D. and W. B. Knrwles, "Response Time Patterns Associated with Vnrious Display-Control
Relationships," Journal of Experimental Psyc logy, Vol. 47, No. 5, 315-322, 1954.

Gerathewohl, S. J., E. L. Brown, J. E. Burke, K. A. Kimball, and S. P. Stackhouse, "In-flight Measurement

of Pilot Workload: A Panel Discusuion," Aviation, Space, and Environmental Medicine, 49 (6), 810-822,
1978.

Grice, C. R., "Stivulus Intensity and Response Evocation," PsychRoloscslRevew, Vol. 75, 359-373, 1968.

Hadley, J. N., "Some Relationships Between Electrical Signs of Central and Peripheral Activity: II.
During Mental Work," Journal of Experimertal Psychology, Vol 28, 53-62, 1941.

Hick, W. E., "On the Rate of Gain of Information," Qrtly. Journal of Exparirental Psychology, Vol. 4,
No. 11, 1952.

Hubel, D. H. and T. N. Wiesel, "Receptive Fields, Binocular Interaction and Functicnal Architecture in
the Cat's Visual Cortex, Journal of Physiology, Vol. 160, 106-154, 1962.

Kagan, J. and B. L. Rosman, "Cardiac and Respiratory Correlates of Attention and an Analytic Attitude,"
Journal of Experimental Child Psychology, Vol. 1, 50-63, 1964.

Kagan, J. and M. Lewis, "Studies of Attention in the Human Infant," Merrill-Palmer Qrtly., Vol. 11,
95-127, 1965.

Kahnexan, D., Attention and Effort, Englewood Cliffs, N. Y., Prentice-Hall, 1973.
66

Kalebeek, J. H.H. "Objective Measuremnt of Mental Workload Possible ApplicatLons to the Flying-Task,"
ProceedinAs of the 55th AGARD Conference. Problems of the Cockpit Environment, 1968.

Kalsbeek, J.W.H., "Sinus Arrhythmia and the Dual Task Method in Neosurlng Mental Load," In W. T.
Singleton, J. G. Fox, and D. Whitfield (Eds.), Measurement of Han at Work, pp 101-114, T*7lor and
Francis, London, 1971.

KeeI3, S. W., "Compatibility and Time-Charing in Serial Reaction Time," Journal of xperimental
Psychology, 75, 529-539, 1967.

Kerr, Beth, "Processing Demands during Meen.l Operations," Memory and Cognition, Vol. 1, No. 4,
4C1-412, 1973.

Kibler, A. W., The Relationship Between Stimulus-Oriented Cha•Aes in Heart Rate and Detection
Efficiency in a Vigilance Task, AMRL-TR-67-233, H-PAID, Ohio, February 1968.
Kleimer, E. T. and P. F. Mueller, Jr., The Rate of Handling Information: Key Pressing Response to Light
Patterns, USAF Human Factors Operations Research Laboratory, Report No. 34, 1953.

Knowles, W. B., "Operator Loading Tasks," H'maan Factors, 5, 155-161, 1963.

Kuperman, G. G. and D. J. Seifert, "Development of a Computer Simulation Model for Evaluating DAIS
Display Concepts," Proceedings Human Factors Society 19th Annual Meeting, pp 347-353, Dallas, Texas,
October 1975.

Lacey, J. I., "Psychophysiological Approaches to the Evaltation of Psychotherapeutic Process and

Outcome," In E. A. Rubinstein and M. B. Parloff (Eds.), Research in Psychotherapy, Washington DC,
iamerican Psychological Association, 1959.

Lacey, J. I., J. Kagen, B. C. Lacey and H. A. Moss, "The Visceral Level: Situational Determinants and
Behavioral Correlates of Autonomic Response Patterns," In P. H. Knapp (Ed.), Expressions of Emocions
in Hen. International Universities Press, 1963.

Lewis, M., J. Kagan, H. Campbell, and J. Kalafat, "The Cardiac Response as a Correlate of Attention in
Inftnts," Child Development, 1965.

Luckiebh, M. and F. K. Moss, "The Effect of Visual Effort upon the Heart-Rate," Jourual of General
Psychology, Vol. 31, 131-139, 1935.

McCormick, E. J., duaan Factors Engtneerin: (2nd Ed.), KcGraw-Hill, N. Y., 1964.
Mc~ill, W. J., "Stochastic Latency Mechanism," in R. D. Luce, R. R. Bush, and E. Galander (Edo.),
Handbook of Mathematical Psychology, Vol. 1, New York. Wiley, 1963.

Midholick, Major, "Safety Treads for Commanders Attention: First Things First," in T SrieflO,
rG
AFRP 11-1, The Inspector General, 1978.

Moray, N., "Where is Capacity Limited? A Survey and a Model," Acta Psychological, 27, 84-92, 1967.

Norman, D. A. (Ed.), Models of Human Nemory, New Y¢rk: Academic Press, 1970.

Obrist, P. A., "Cardiovascular Differentiation of Sensory Stimuli," Psychosomatic Medicine, Vol. 25,
450-469, 1963.

Pew, R. h., "Human liformation Processing Concepts for Systems Engineers," In R. E. Macho! (Ed.),
System Engineering hwndbook, McGraw-Hill, N. Y., 1965.

Pierce, J. R. and J. E. Karlin, Bell System Technical Journal, Vol. 36, 497, 1957.

Plattner, C. N., "Heart Strain Greater in Landing on Carrier," In Aviation Week and Space Te-hnology,
March 13, 1967.

Posner, N. I. and S. W. Keele, "Time and Space as Measures of Mental Operations," Invited Address,
Division 3, American Psychological Association, Sep. 1970.

Quastler, H. and V. J. Wulff, Control Systems Laboratory, Report No. 62, University of Illinois, 1955.

Ryan, T. A., Work and Effort: The Psychology of Production, The Ronald Press Co., N.Y., 1947.

Saunders, R. S., H. G. Smith, and W. H. Teichner, Models of Short-Term Menory: A Critical Review,
TR 74-1, Department of Psychology, New Mexico State University, Las Cruces, New Mexico, February 1974.

Shapero, A. and C. Bates, Jr., A Method for Performing Human Engineering Analysis of Weapons Systems,
WADC TR 59-784, WPAPB, Ohio, 1959.

Smith, E. E., "Choice-Reaction Time: An Analysis of the Major Theoretical Positions," Pasychlogical
Bulletin, 69, 77-110, 1968.

.okolov, E. N., Perception and the Conditioned Reflex, Trans. S. W. Waydenfold, Oxford: Pergamon Press,
1963.
 Starnberg, S. 'lbe Discovery of Procepning Stages- Kxtenoa.in of Donder's Ke':hod, In W. G. Xosner (Ed.),
SAttention and Performance II. Acta PsychologLical, 34),276-315, _S9.

Steruberg, S., "High-Speed Scanning in Husav Newory," Science, 153, 652-654, 1966,

Sternberg, S., hemory-Scanning: Meatal Procepaces Revealed by Reactlon Time Experimer.ts, Aaerican
Scientist, 57, 42,-457, 1969.

SvanaoA. J. H. nvl G. 1. Brig~a, "lnformai.ion Processing as a Function of Speed verryes Accuracy,"

Journal of Experimental Psyr-oloa, Vol. 81, NJ. 2, 223-229, 1969.

Teichner, W. E., H-=uAn Performance Simulation, Annual kteport, AFOSR Contract Nr3.F44620-76-C-0013,
New Mexico State Uriversity, 1976.

Teichner, V. il., Quantitative Mode1s fol predIctins Human Visual/Perceptual/Notor Perfurmance, Final.
Technical Report 74-3, Contract 0111 NO0lI-7G-A-O147-0002, New Mexico State Univo-rsity, Las Cruees,
Nay Mexico, October 1974.

Telchoer, W. R. and D. E. Olson, "A Preliminary Theory of the Effects of Task and Environmental Factors
tn Human Performance," Human Factors, Vol. 13, No. 4, August 1971.

Teichner, W. H. and M. J. Krebs, "Laws of Simple Visual Reaction Time," Psychological Review, Vol. 79,
344-358, 1972.

Teichner, W. F. and M. J. Krebs, "Laws of Visual Choice-ReactSLon Time," Psychological Review, Vol. 81,
No. 1, 75-98, 1974.

Vaughan, H. G., Jr., "The Relationship of Brain Activity tc Scalp Recordings of Event-Related Votencials,"
in E. Donchin and D. Lindsley (Fds.0, Average Evoked Potentials: Methods, Results, and Eval~ations,
NASA SF-191, Proceedings of Conference at San Francisco, Calif., September 1968, pp 45-94.

Weisszan, N. W., Preface, In E. Donchin and U. Lindeley (Eds.) Average Evoked Potentials: Methods,
Results, and Evaluations, NASA SP-191, Proceedings of Conference at San FrLancirco, Calif.- Sept 1968.

Welford, A. T-, "The Measurement of Sensory-Motor Performance: Survey and Appraisal of Tgelve Yers of
Progtews," E•;ir'coulcs, Vol. 3, 139-230, 1960.

Welford, A. T., "The Psych.ological Refractory Period and the Timidig of High Sneed Performance: A Review
ard a Theory," British Journal of Psychology, Vol. 34, 2-19, 1952.

Woodworth, k :. and H. Schlosberg, Experimental Psychology, Henry Holt and Co., Inc., 1965.

Wortman, D. B., S. D. Uuket, and D. J. Seifert, "GAlNT Simulation of a Remotely Piloted Vehicle Drone
Control 'acility, Proceedings. Humsn Factors Society, 19th Anxual Meating, pp 342-346, Dallas, Tests,
October 1975.

I.
 69

QUANTITATIVE MILITARY WORKLOAD ANALYSIS

by

Richard A. Albanese
USAF School of Aerospace Medicine
Brooks AFB, Texas, 78235, USA
F- INTSqODUCTION

A central goal of a military workload analyst is to understand the determinants of mission success
in a mllitnry setting. The emphasis is on the human determinants of mission success with particular
consideration to how the human uses the system he is given to accomplish the mission at hand. In quanti-
tative workload analysis the final goal in many instances is to provide various numerical measures of
mission performance. For example, when examining a bombing mission, a workload analyst using mathematical
models right attempt to estimate the probability that bombs would land on target. Specifically, he might
attempt a statement such as: "The estimated circular error probable is 250 feet given the present work-
load conditions." Other measures he might estimate include summary statistics such as anticipated loss
rates against specific enemy defensive configurations, and rates of overall success against enemy targets,
and these summary statistics will be of particular interest belay.
A workload analyst studies the system under corsideration to determine its capabilities and, when
appropriate, he designs system changes or modifications with a view to improving system performance. The
main purpose of this paper is to suggest that the wozkload analyst attempt to evaluate his pýoposed design
modifications within the framework of a quantitativa or -=-quantitativecost/benefit tradeoff. This is
particularly .rprepriate when the analyst hns developed relevant metrics describing system performance
both with and without the system modification.
A workload analyst can suggest a wide variety of system changes ranging from hardware modifications
to changes in syptem operating procedures. Whatever the changes sugSesta.d, a workload study in the mili-
tary setting can be represented by a cost/benefit table as shown in Figure 1.
In Figure 1, the basic or unmodified system has effectiveness e, vulnerability v, and cost per system
c. Subsequent discussion will provide definitions of e and v. System modificatious can Improve the
"nffect'-veness of a system from the point of view of making the system more capable of inflicting losses on
the enemy. However, improving the fighting effectiveness of the system can increase or decrease the sys-
tem's vulnerability, just as decreasing a system's vulnerability can either decrease or increase the
system's fightiug effectiveness. In Figure 1, a,, v 1 , and cl are the effectiveness, vulnerability and
system's cost of the basic military system with system modification #1. The symbols e2, v 2 , and c 2 are
used in a similar manner for the system with modification #2.
Perhaps most readers will agree that composing the table in Figure 1 is a step forward, but, of
course, still remiaining to the question of how to use the assembled data for actual decision making.
Should an investment of money be made, and if so could modification #1 or modification 02 be purchased,
or should one simply recommend that more elements of the basic system be procured? Analytical scenario
modeling can be a decision aid in this circumstance, and this will be described in the following section.
The method or methods whereby a quantitative tradeoff table such as that shown in Figure 1 can be developed
will be described briefly in the section of this paper following the next concerning analytical scenario
modeling.
ANALYTICAL SCENARIO MODELING

In this section, system mission effectiveness, e, and system mission vulnerability, v, will be
defined in the context of analytical scenario modeling. For the purpose of illustrating the usefulness
of analytical scenario modeling, a simple example from among a class of combat models called Lanchester
models, will be employed. This class of models was developed by Lanchester, an aeronautical engineer, in
about 1914, and is extremely simple in conception and approach (1, 2).
Csnaider two opposing forces, Blue force versus Red force. The rate of attrition of the Blue force
shouli be proportional to the number of Red systems available, that is
dB/dt - XR
X R Eq. 1.

where B is the numter of Blue force systems or elements, R is the number of Red force elements, and r is
the constant of proportionality which reflects Red's ability to reduce Blue force. Similarly, the equiv-
alent differential equation for the attrition of the Red force is where b is a\constant of proportionality
dR/dt - - b X3 Eq. 2.

which measures Blue force's ability to reduce the Red force. If the Blue force is identified as the
analyst's side, the proportionality constant b can be identified with system effectiveness e, and,
similarly, the proportionalit, constant r can be identified with Blue force system vulnerability v. Thus
the following equations obtain.

dB/dt-- v X R Eq. 3.

and
dR/dt - - e X B Eq. 4.
and these equations provide quantitative definitions of effectiveness and vulnerability. These equations
are easily solved to provide B and R as functions uf time t. These solutions are shown below for the
interesting but highly simplified case where e and v ore constants.
 70

B(t) - (1/2r) {(/e B, + /v Ro)exp(-/ev t) +

(Ve Bo - 4v Ro)exp(+/ev t)) Eq. 5.

4
R(5) = (1/24) {(6 B, + 4VRo)exp(- v t) +
(4 3ý - 4 B,)exp(+r-ev t)} Eq. 6.

H• In these equations, Be and Ro are the sizes of the Blue force and the Red force, respectively, at time
L t - 0 at the start of combat (prior to any losses). These last two equations describe force attrition
during a battle. Far more complex attrition models are often developed to study 'orce adequacy and
tactics. The suggestion made here is that such attrition models or analytical scenario models be adapted
and employed in workload tradeoff analyses. The concept is to compare design alternatives against pre-
dicted combat outcomes, and to choose that system modification which optimizes desired outcomes. This
concept will be illustrated in the following by using equations #5 and #6.

A natural military goal is to reduce the enemy while minimizing one's own losses. This military
goal can serve as an outcome metric which can discriminate between differing system modifications. Other
outcome metrics can be defined such as minimizing one's own losses while reducing the enemy in the shortest
time possible. However, for the purposes of the present illustration the simpler metric of minimizing
losses alone will be employed.

Examining
the Red unit if equations #5 and
the quantity re #6,
Bo isit greater
can be noted limediately
than the quantity that
rv Ro Blue unit( /e
(since will
R ultimately
- e Bo) is' domirnate
tbn a
negative quantity in equation #6). If re B. is greater than v Re. R will be zero at time t - t* where

t- - (l/2Vev)ln((,e Bo + /4 Ro)/(v/e Bo - 4vo)) Eq. 7.

and, using this critical time, maximum Blue force losses can be calculated using the following formula:

Losseab - Bo - ((eo 2 - VE)le)h Eq. 8.

Similarly, if /v Ro is greater than re Bo Red force will ultimately dominate Blue force, and B will be
zero at time t - tB having lost all Bo systems.

These last equations will now be employed to accomplish an example tradeoff analysis. A hypothetical
cost/benefit table is shown in Figure 2. In this table, the fact that 4oB - 4 . 250 - 500 is less than
v Ro - x 400 - 565, certainly motivates the Blue analyst to recommend changes. Modification #1 allows
Blue to defeat Red while sustaining loss of 177 Blue elements. The cost of modification #1 is 37.50
million dollars which is a sum which would allow procurement of 37 additional uumodified systems. Sinct.
Sx 287 is greater than /2 x 400, the Blue force augmented by 37 elements, would defeat the Red force,
but in so doing the Blue force would sustain a loss of 201 systems. Thus, modification #1 would be pre-
ferred over the equivalent Blue force augmentation.

Now consider modification #2. This modification allows Blue force to win with a loss of 142 units.
The cost of modification #2 is 125 million dollars. With this money, 125 additional unmodified systems
can be procured to form a force of 375 fighting elements. With this size force, Blue force defeats Red
force while sustaining losses of only 92 systems indicating that equivalent augmentation of the unmodified
force would be preferable to purchasing modification #2. More complex mathematical models would allow
consideration of purchases of various combinations of modifications #1 and #2. Putting these more compli-
cated situations aside, and simply using what has been computed above, it can be concluded that if 125
million dollars were available, augmentation of the basic force should be accomplished without modifying
the individual elements of the force. However, if 37.5 million dollars are available for use, modification
#1 will minimize losses.

It has thus been illustrated how analytical scenario models cun be used in workload analysis tradeoff
studies. These models can help workload analysts define their earned return on investment and can help
with decisions concerning modification alternatives. Admittedly perhaps the simplest scenario model has
been employed here to illustrate scenario model usefulness. It is anticipated that real-world decisions
would employ simulations which are far more complex and extremely well tested. Nonetheless, from the
above very simple example, the workload analyst should be prepared to realize that in some instances it
may be preferable to procure more of an unmodifiee system than to proceed to a modified system.

CONSTRUCTION OF QUANTITATIVE TRADEOFF TABLES

In this section, the construction of quantitative tradeoff tables ns shown in Figures #1 and #2, will
be briefly discussed. These tables can be constructed using three different types of data sets. A data
set of type #1 consists of data deried on the military systems of interest and including the precise
effectiveness and vulnerability figures needAed to complete the tradeoff table. Type #1 data sets are
rarely encountered in practice. These data sets can be developed from records of actual combat or they
can be developed from records of realistic practice or training encounters where different systems are
employed or compared. This data type, when it is available, provides the best and most direct data for
tradeoff studies.

A data set of type #2 consists of data derived from the actual military systems of interest in the
tradeoff study; however, the performance measures available from these systems are not the desired effec-
tiveness and vulnerability measures. Often in this setting the available data are indirect measures of
mission performance, or measures of human operator workload stress during mission performance, from which
the likelihood of mission failure can be inferred. For example, when the concern is with a bombing mis-
sion, instead of "btaining the numbers of enemy targets destroyed per unit time by the competing systems,
this data type might provide circular error probable figurer from which enemy target destruction would
have to be inferred. Still more indirect data concerning the bomber performance would be data that
 71

related to airerew stress during the performance of trial missions. Such measures sre, for example,
[ voice stress measurements, galvanic Pkin responses, cortisol secretion and the like. It is clear that
with da ta set@ of type #2, the workload analyst faces n problem o2 extrapolating from the available
performance measures to measures that are mote relevant to a military decision in a tradeoff setting.
•! •A data oý,t of type #3 consists of data derivPA from esytme which are not thn military systems of

interest or concern in the tradeoff deliberations, but are other military systems currently in the
inventory, or are, as is oiten the case, laboratory simulatiors of the real systems under consideration.
Thus data sets of type #3 also pose serious extrapolation proble--. In thim case, the extrapolation
problem is one of relating data from one system to the rele,,ant performance measure applicable to another
Systesm.

As discussed above, data sets of type #3 and type #3 rpr,,ire that the workload analyst extrapolate
between measures, or between military systems, or both. extrapolation can be done via experimentation
or via the use of experimentation coupled with the application of mAthematical models. The use of mathe-
maiical models in ml.itary wiorkload analysis has been outlined in a previous publication wherein a coarse
classification of availabli modeling techniques is provided.

The aboqe discussion has suggested that military workload analyses proceed in the setting of quanti-
tative or esai-quantitative tradeoff analysis. This setting is already quite familiar to the hardware
engineer, but may be a novel eetting for the huiman factors workload specialist. The term semi-quantitative
analysiu is employed to recognize the fact that it will not always be possible to precisely quantitate
effectiveness and vulnerabil-tty as well as one would wish.

The methods de3cribed in this report rely heavily on mathematical modeling techniques. This is seen
in the suggestion to employ analytieLl scenario modeling in the tradeoff study, and is also seen in the
suggestion to employ mathematical radels in the construction of the tradeoff table from data sets that are
act directly applicable. While mathematical models can be extremely usa'f'_1 and cost effective in appli-
cation, they must be used with sober caution. Mathematical models are best employed with an attitude
which conolders the mathematical models, not as a replacement for traditional methods, but as an adjunct
to comonly employed methods of analysis and deliberation. Mathematical models should in no way displace
the direct use of experience and the direct consideration of empirical data. Rather, mathematical models
should be used to enhance ant highlight the utility of available data sources. The analyst's dictum "never
believe your mathematical model" is a wise rule which is simply a statement of caution intending to remind
the analyst that mathematical models are as fallible as any other human-contrived decision aid.

CONCLUSION

This report has discussed a method of tradeoff analysis as applied to workload analysis in the military
environment. It is suggested that workload studies be performed in a tradeoff setting which allows the
analyst to estimate the return on investment he has earned through his proposed system modifications.
The methodologies described employ mathematical modeling techniques, and it is reinforced that these
techniques are an adjunct t i, and not a replacement of, more traditional methods of workload analysis.

REFERENCES

1. Lancheaster, Frederick William. Mathematics in Warfare. IN: The World of Mathematics, Vol. 4,
Simon and Schuster, New York, np. 2138-2157.

2. Perla, Peter P. Approximation Techniques and Optimal Decision Making for Stochastic Lanchester
Models. Technical Report #137. Department of Statistics, Carnegie-Mellon Univeiaity, January ,.978.

3. Albanese, Richard A. Mathematical Analysis aad Computer Simulation in Military Mission Workload
Assessement. AGARD CP-216, 1978.

J
 - - - !73

: • VISUAL PEIROIMANCE: A METHOD TO ASSESS WORKLOAD

IN THE FLIGHT ENVIRONM(ENT

by

R. Simons, M. Sanders, and K. Kimball

"United StatesArmy Aeromedical Research Laboratory
Fort Rucker, Alabama, 36362, USA

INTRODUCTION

Operator workload for the task of vehicle manipulation perhaps could be defined as the sum of sen-
sory inputs, psychomotor responses, and cognitive processes. Sensory inputs to the operator are util-
ized to direct control manipulation, obtain feedback as to degree of effectiveness of the control
movemenus, and to monitor system status. This input workload is combined with the psychomotcr workload
requi-:ed to move the vehicle controls as dictated from the sensory inputs and feedback modes. No-e
simply stated, workload measurements can be derived by objectively measuring the input and/or output of
the operator.

The ability to manipulate an aircraft, as well as a tank, car, or any other vehicle, is directly
related to inputs or cues the operator receives from the environment. Of these perceptual invitst
(tactile, visual, auditory, etc.) required to fly an aircraft, visual cues are considerel vital. R.
Hartan has even estimated that vehicle operators acquire over 90% of their required iuformation
visually. Processing and integrating these visual cues allow the pilot Zo detect the aircraft's rela-
tive stability, ground references, and provide feedback from his control functions. ydrin7, flight
conducted under instrument meteorological conditions (INC), lack of cues from the envtrotf.ent outside
the aircraft requires the pilot to obtain the necessary visual information from instrument displays.
As a consequence, there exists the need, independent of visual conditions, to determine what cues or
visual workload are required to achieve maximum pilot efficiency with miminal fatigue-iaduced errors
and safe mission accomplishment.

A great variety of apparatus and techniques have been developed for the study of visual performance/
workload (2, 3, 4). One of the earlier devices was a smoked-drum Kymograph attached to the sclera of tht
eyeb&ll via fine wire and barbed hooks. During the 1930's, electrooculography (BOGI techniques were
developed which utilized electrodes placed around the eyes of the facial structure to monitor differ-
ential voltages as the eyeball was rotated (5).

The earliest documented technique for measuring the vital performance of pilots was to simply
record pictures of the operator's face while he scanued the instruments (6). In.provements of this
method were accomplished by arranging mirrors on the instrument panel and photographing the total
arrangement. Documentation of eye movement was obtained by means of a camera tAo-mnted behind the
pilot. Ouring analysis a photo interpreter scanned the film to determine which mirror reflected the
eye of the pilot at various times during the flight (7).

This technique was further refined by Mackworth (8. His approach was to mount a lightweight
moving picture camera beside the operator's head along with a series of mirrors which reflected a dot
representing the eye's motion. This dot was superimposed on photographs of the scene directly in front
of center line of the head. More recently this same "corneal reflection" technique has been utilized
by the US Army Aeromedical Research Laboratory in the study of Army pilot visual performance during
helicopter flight (9, 10).

The corneal reflection technique is possible because of the smooth spherical front surface of the
cornea. An incident beam of light can be partially reflected forming a bright spot or "highlight" on
the cornea. The angle of the reflected light depends upon the anglq between the incident light ray and
a plane tangent to the reflecting surface. Since the cornea forms an eccentric bulge on tho nearly
spherical eyeball, the angle of this tangential plane on the cornea at any one point changea1 a. the eye
rotates around ita center during eye movement. As a revult, the position of the highlight tallows the
direction of movement of the cornea. The reflected beam is easily photographed on film. By mounting
a camera lens on subject's head slightly above and between his eyes, the subject's normal visual field
can be recorded and the highlight can be superimposed on the scene to give a constant eye reference to
the eye's highlight, the area of visual concentration and the percentage of time for eye stabilization
during any flight maneuver can be recorded.

Past research has demonstrated two major advantages of the corneal reflection technique for study-
ing eye movement. First, the method is convenient for large scale testing of subjects In that it
requires minimal training. Second, these studies have reported no significant interference with normal
eye movement (11, 12). This laboratory utilizes motion picture film to record the visual performance
data. Figure 1 is a picture af the octilomotor lense and peripheral equipment. The total methodology
is outlined in USAARL Report No. 77-4 (13).

Investigations which have been devised to collect data related to visual performance can be divided
into three categories: (1) subjective opinions of visual performance, (2) objective visual performance
data during fixed wing flight, and (3) objective data during helicopter fl 3ht. Studies by Siegel and
MacPherson (14), Clark and Intano (15), Simmons, et al. (16) have analyzed the opinions of aviators as
to which instruments they felt were utilized to fly selected maneuvers. However, these findings do not
agree with the research results of Frezell, et al. (10), Sanders (12), and Simons, et al. (13). These
investigators have reported a very poor agreement between subjective data and actual pilot visual per-
formance. Additional studies by Milton, Jones, and Fitts (6), Fitts, at al. (7), and Diamond (17) have
utilized equipment to obtain objective visual performance data of aviators during flight maneuvers in
several fixed wing aircraft. Although these investigations provided useful information as to visual
performance during fixed wing flight, data obtained during this work cannot easily be generalized to
rotary wing flight because of the extreme aerodynamic differences between airplanes and helicopters.
74

Sunkes, et a1. (18), Sterm and Byuum (19), fresell, et al. (9, 10) have recorded visual performance
in helicopters during selected visual flight rules (V7R) flights. Additionally, two reports (20, 21)
investigated a number of maneuvers utilizing both the interview technique as well as ir.flight recordings
of visual performance of two aviators under instrument flight rules (IIR) conditions. These efforts
have provided some needed information an to the frequen:y, duration, and sequence of fixations during
helicopter operations.
Although these studies have provided useful information for the visual performance data base much
investigation remains to be accomplished before a reliable visual performance/workload model can be
established fov safe helicopter flight. The pupose of this report is to ettempt lo combine the visual
performance investigations being performed at the US Army Aeromedical Research Laboratory into one mode
for predicting visual workload.
MN1OR0

Several meacuremente of visual performance derived from data collected via the corneal reflection
techn~ique eontribute to the total relationship of v' ual workload. In simple terms, oculomotor activity
can be divided into two categories: (1) movement of the eye during which minimal information gathering
occurs and, (2) fixation, a period of relatively no movement during which information transfer ie felt to
be the greatest (1). The movement activity is defined as the visual link value or the visual, path
traveled from one area of interest to another. On the other hand, the visual nonmovement term, visual
fixation, is defined as stationary eye movement within a designated area for at least 100 milliseconds.
Other visual terms which could be included are the total number of areas that are concentrated on (or
fixated), the length of time of each fixation (or dwell time), and the frequency that areas of interest
are fixated.

If one assumes that the major input mode is the fixation period, two possibilities exist. Visual
workload could be a function of the time required for information to be transferret duriug fixation; or,
workload could be related to the frequency of visits to an area of interest. Since from a previous
investigation (22) neither term was found to adequately describe visual activity independently, both
comprise this input mode workload and should be combined. Thus, a formula utilizing these two terms
would reflect the worktoad cost of all areas that were fixated by an operator during vehicle manipulation.

This formula would appear as: CFa - (T/ET+N/EN)/2. CFa represents the "cost factor" of an area of
interest. "T" is lapse time spent fixated on the area divided by total time (ET) while "N" is the
frequency of fixations of the area divided., by the total number of fixations (EN). If these two values
are divided by 2, the CF is in percentage of workload. If the CF values of several areas of interest
lend themselves to being combined into common zones of interest, the CF values are simply sumed to.-
gether (Cls + CFa5 + CFs 2 + CFa3 +..+...+).

Based on our experience, the visual inputs required to manipulate an aircraft can be divided into
three broad categories: (1) basic vehicle control, (2) barrier avoidance, and (3) navigational tracking.
The first requirement takes precedence over the latter two. Urler this category of basic vehicle control,
visu.1 workload can be further separated into three major zones of comon areas visual interest. Again,
the highest priority zone contains visual cues which provide information relating the basic vehicle
stability about its three major axes of pitch, yaw, and roll.

The second zone of coommn areas of visual interest include the input information which supports the
first areas biut provides for more precise vehicle control. Information such as vehicle speed, altitude,
and rates of acceleration would be provided from this zone.

The last zone would be comprised of vehicle status information. These cues would provide operator
visual feaiback cs to the operational condition of the vehicle. Examples of such types of information
woul- be provided from engine oil temperature, fuel pressure, or electrical gauges. As long as there
were no malfunction of the vehicle as annunciated br one of these instruments, this zone of visual inputs
vould have the lowest priority of being monitored.

To summarize, the CF t•aory provides a method of combining numerous blocks of visual data to provide
a more concise picture of input workload of vehicle operators. 1he CF value computed for Zone 1 should
be an indicator of basic workload required to perform the task successfully. Zone 2 will also provide
supportive data of the basic workload based on time available after Zone 1 requiresents are met.

It should, however, be quickly pointed out that maximum visual performance of an area or zone of
areas could indicate high visual workload. On the other hand, this same performance could reflect a
high percentage of nonworkload (free time) in which the particular zone was fixated because it was
centrally located. This could be demonstrated by similar visual workload in the central viewing field
of a boat operator on a large lake and a helicopter pilot during nap-of-the-earth maneuvers. Howev'er.
by establishing conditions which provide measurements of the baseline for the maximum time utilized and
the minimum time required to maintain vehicle stebility, these "free time" periods can be estimated. An
example of this can be reviewed in USAARL Report No. 78-6, Visual Performance/Workload of Helicopter
Pilots During Instrument Flight (22).
The remainder of this report all deal with the data base which the US Army Aeromedical Research
Laboratory has obtained during helicopter and fixed wing maneuvers in attempting to establish a visual
workload model. These types of data not only provide the needed information to test the CF theory, but
also will provide Information to improve and refine the theory to provide operational answers for safer
military airborne operations.

APkPLICATION

Initially, a study was designed to investigate the visual performance of helicopter pilots during actual
flights under instrument flight conditions (IFR) (22). This study was unique because the aviators Ttere
~~ k
 75

forced by the test conditions to receive any and all visual cues to manipulate the aircraft from the
instrument panel. This limited visual field allowed investigators to analyze which cues were fixated and
derive what Information was required by the pilots. During VFR this extraction of visual performance date
would be very difficult because of the lack of precise deffnitions as to the quality of possible VI• cues.

Visual performance via the corneal reflection technique was collected from two groups of subject
pilots. Subject groups were categorized on the basis of flight experience, with one group having over
2,000 more flight hours than the other. All subjects flew the same instrument flight profile comprised
of eight basic maneuvers. The results of the study are summarized by Figure 2. IQS identifies the pilots
with the most flight experienced while SQA represents the low time pilots. Z1 , Z2 and Z4 designate the
"three zones of instruments following the previously aiscussed method of classification. Table 1 is the
listing of those instrument* comprising each zone.

Since Zone 1 is the most critical indicator of visuaYl wor!-load. the data reflect that the experienced
pilots had more workload to complete the missior than did the less experienced pilots. This could be
further interpreted to mean that: (1) the IQA could spend less time in this flight environment before
becoming fatigued, or (2) the IQA would most likely make more flight errors soover than the SQA pilots.
These results appear to contradict the comon philouophy that experienced pilots should have been the
better combat-prepared pilots. Therefore, the data were re-examained more closely for other possible
explanations. In attempting to astablish other gioup differences, it was concluded that although the IQA
group did have the most totas flight time, they were all currently holding job positions as initrument
flight instructors. For this reason they, in fact, had less current "hands-on" experience than the SQA
group who were all recent graduates of flight school and therefore had just completsd a very concentrated
block of '"ands-on" flight experience.

To further test this line of thought, a single subject was selected who currently had 2,500 hours of
flight experience but who had not flown for the past three years (23). His initial flight test results
(NQA) are reflected by Figure 3. The results indicate a significantly higher level for his visual work-
load in Zone 1 to perform the same mission as the previous subjects. ThIs subject was then given 14 hours
of refresher training by the laboratory's a&structor pilot. Figure 4 presents the results of his lost
flight (NQA) on tl,*same profile compared again to the initial SQA subject group. It is apparent that
his worklcad to pr=rfora the mission has been reduced to a similar level as that of the SQA group. These
resulta would seen to indicate that utilizing the CF method of calculating visual workload aided in
identifying differI.ng v'sua. woikload as a function of aviator's current proficiency.

This same method was again aiessed during a second investigation which compared the visual workload
associatoý with flight of a fixed wing aircra't, during instrument conditions, compared to the original
rotary wing instrrment flights (24). AD-21 fixed wing aircraft was flown over the same flight profile
as in the helicopter instrument flight. Two subject groups were again utilized. However, for this
investigation, the firec group were current instruccor pilots (ICA) which compared to the IQA group in
the helicopter repnrt. The second ptoup consisted of noncurrent 13-21 pilots. (NCA) who had not flown the
U-21 for at least 3 years prior to the test flight. The purpose of this investigation was twofold in
that it allovud a comeArison of visual workload as a function of vehicle stability (i.e., rotary wing
versus fixed wing *i,:craft) while further testing the currency versus experience question.

Figure 5 repiresents comparison of the two U-21 subject groups. The results indicate, as have past
findings, that the noncurrent aviators (NCA) experienced more visual workload than did the current aviators
(ICA). However, a confounding variable was that the NCA subjects were all current in the UH-1 helicopters.
Because of this variable, the level of difference between subject gre, ps is perhaps not as significant as
would us anticipated. Nevertheless, the CF visual workload theory wgs effectively utilized to indicate
the visual workload associated with aviator current proficiency.

Th. ICA subjscts of the U-21 stud,, ere then compared to the IQA aviators from the helicorter instru-
ment report. A representation of this coaparison is referenced by Figure 6. Again, if the CF theory is
an iodtcatin of cost or workload associated with the manipulation of a vehicle, the results would demon-
strate that the UH-1H helicopter requires more .visual workload on the primary Zone 1 instruments than did
the U-21 fixed wing aircraft. These findings would be predicted by subjective data and the relative
rating of the stability of the visual data and the relative ratings of the stability of the two vehicles
by other test agencies. However, the implications of the visual data are that 11 the helicopter stability
iao
Improved, then the crew could remain on station or in combat longer before becoming fatigued.

rhis same method of testing could be implemented to test future generation of helicopters to determine
relative stability. If such aircraft did impose less visual input work to manipulate, they would provide
a better platform for combat utilization,

To further expand the line of thought that the CF theory could reflect in some part the visual workload
aoeoiated with the stability of the vehicle, a study has been completed. This investigation compared two
groups of subjects with qualifications similar to the original SQA and IQA groups of the helicopter study
(25). The two groups were; however, tested in an UH-1 flight simulator which was developed for the US
Army to duplicate the flight, engine, and system characteristics of the UH-l helicopter.

The results are oumnarized in Figure 7. The conclusions that can be drawn from these results are
that the UH-1 simulator does, iz general, have the same visual workload pattern as the UH-1 helicopter.
However, because the visual workload in Zone 1 is higher for the simulator, the vehicle is less stable
than the UH-1 helicopter. An expansion of the CF of Zone 1 can be seen in Figure S. The three instru-
Rments that comprise this zone are indicated by Al for the artifical horizon, YXI for radio magnetic
compass, and T-B for turn and bank indicator. From the major difference of the two vehicles as eaes. on
the workload of the MI, the indication would be that the U13-1 simulator is less stable mainly on the
yaw axis. In addition, the inter-group subject dOfferences in the simulator reflected the same results
as had been reported in the UH-1 helicopter study.
76

CONCLUSIONS

In s Tary, this paper has attempted to addrers a method of assessing workload requirements imposed
on one of five possible operator sensory channels. Sluc, the demand of this visual input channel is
estimated, as previously stated, to be 90Z of the total input demands for vehicle manipulation, any theory
which allows even a partial but precive description of the workload could aid future hardware decigna,
tzaining, and miss&on delineation. These data will further be useful in determining an approach to reduce
operator fatigue in the flight environment.

The current CF theory. although not the final answer, allows a more concise picture of visual work-
load than the classical methods which normally consist of the peruatation of seemingly unrelated visual
data points. The appl.-ation section of this report demonstrated how the US Army Aeromedical Research
Laboratory has collected and is continuing to expand a data base describing pilot visual performance in
the military environment. Such data are considered invaluable to expand and test the current CF theory
as well as providing an objective method to be utilised in answering current operational questions and
problems. The examples were brief descriptions of studies which are already published in their entirety
or are in the process of being completed. The implications from the results sugg•et that the CF theory
is a valuabie tool in testing and determining what the visual workload level should be for combat profi-
cient pilots, how long pilots with varying degrees of proficiency could be expected to fly In the combat
environment, and aircraft design requirements (such as stability), to reduce Lhe onset of fatigue-induced
errors. Additionally, the CF theory can be utilized to test and determine varying mission related work-
load, as wall as the workload required by special equipment such as the night vision goggles, navigation
equipment, and experimental flight displays.

The ability to measure visual input workload and/or psychomotor control is recognized as an invaluable
tool required to validate instrument panel design, develop training and proficiency requirements and, in
general, provide a more effective helicopter cystem for mission accomplishment.

TABLE 1

INSTRUMENT CLUSTERS WITHIN EACH ZONE

ZONE I 1. Attitude Indicator AH

2. Radio Magnetic Compass RMT
3. Turn and Slip Indicator T&B

ZONE II 1. Altimeter ALT

2. Airspeed Indicator AS
3. Vertical Velocity Indicator VSI

ZONE III 1. Aircraft Monitoring Gauges TORQ, RPM, ELEC,

OIL, FUEL
2. Special Navigatiou Instru- OBS
mentation
3. All Other Visual Areas REST

REFERENCES

1. Senders, J. W., Fisher, D. F., and Monty, R. A. (Eds.) Sye movements and the higher psychological
functions. Hillsdale, New Jersey: Lawrence Erlbaum Associates, 1978.

2. Hall, R. J., and Cusack, B. L. The measurement of eye behavior: Critical and selected reviews of
voluntary eye movement and blinking (H]L DK 18-72). Aberdeen Proving Ground, Maryland: Human
Engineering Laboratory, July 1972.

3. Klein, R. H. and Jex, H. R. An eye-point-of-regard system used in scanning and display research.
Paper presented at the SPIS 15th Annual Technical Symposium, Anaheim, California, September 1970.

4. Monty, R. A. An advanced eye-movement measuring and recording system, American Psychologist, 1975,
30(3), 331-335.

5. Mowror, 0. H., Ruch, R. C., and Miller, N. N. The cornea-retinal potential difference as the basis
of the galva,'ometric method of record in8 eye movements. American Journal of Psychology, 1936, 114,
A23.

6. Milton, J. L., Jones, R. E., and Fitts, P. M. Eye fixation. of aircraft pilots: Frequency,
duration, and sequence of fixations when flying the USAF instrument low approach systems (ILAS) (USAF
Tech. Rpt. No. 5839). Dayton, Ohio: Wright-Patterson Air Force Base, October 1949.

7. Fitts, Paul N., Jones, R. E., and Milton, J. L. Eye fixations of aircraft pilots: Frequenry
duration, and sequence fixations when flying Air Force ground controlled approach system (GCA) (USA"
Tech. ipt. No. 5967). Dayton, Ohio: Wright-Patterson AFB, November 1949.
 , !, 77

8. Mackworth, N. H. and Thomas, E. L. Head-mounted eye marker caner.-. Optical Society of American
Journal, 1963, 52, 713-716.

9. Frezell, T. L., Hofmann, N. A. and Oliver, R. E. Aviation visual ;erformance in the UH-iHl. Study I
(USAARL Rpt. No. 74-7). Fort Rucker, Alabama: US Amy Aeromedical Research Laboratory, October
1973.

10. Freasll, T. L, Hofmann, M. A., Snow, A. C., Jr. and Mcdlutt, R. P. Aviator visual performance ia the
UH-iH, Study 11 Aeromedical Research Laboratory, March 1975.

11. Young, L. P. Measuring eye movements. American Journal of Medical Electronics, 1963, 300-307.

12. Sanders, X. G. Personal communication based on data being prepared for publication. November 1976.

13. Simons, R. R., Kimball, K. A. and Dias, J. J. Measurement of aviator visual performance and workload
during helicopter operations (USAARL Rpt. No. 77-4; Fort Rucker, Alabama: US Army Asromedical
Reseuu.cL Laboratory, December 1976.

14. Siegel, A. and MacPherson, D. Pilot opinion of the optlmm arrangement of primary flight instruments
in Naval aircraft (HADA-AC-6910). Warminster, PA: Naval Air Development Center, September 1969.

15. Clark, W. and Itano, G.. Helicopter display improvment study (IFC-TH-75-1). Randolph AFB, TX: USAF
Instrument Flight Center, May 1975.

16. Simons, R., Hofmann, M. and Lees, 1. Pilot opinion oZ flight displays and monitoring gauges in
the UH-1 helicopter (USAARL Rpt. No. 76-18). Fort Rucker, AL: US Army Aeromedical Research
Laboratory, April 1976.

17. Diamond, S. Time, space and sterosccpic vision: Visual flight safety considerations at supersonic
speeds. Aerospace Medicine, 1970, 41, 300-305.

18. Sunke*, J. A., Patera, K. E., and Howell, W. D. A study of helicopter pilots' eye movements during
visual flight conditions (Task Assignment No. 59-205-10). Atlantic City, N. J.: Test and Experimental
Division, September, 1960.

19. Stern, J. A., and Bynum, J. A. Analysis of visatal search activity in skilled and novice helicopter
pilots. Aerospace Medicine, 1970, 41, 300-305.

20. Barnes, J. A. Tactical utility helicopter information transfer study (HEL 7147-72). Aberdeen
Proving Grounds, ND: Human Engineering Laboratory, April, 1972.
21. Barnes, J. A. Analysis of pilots' eye movements durlg helicopter flight (HEL TH 11-72). Aberdeen
Proving Grounds, ND: Human Engineering Laboratory, April, 1972.

22. Simons, R. R., Lees, 1. A., and Kimball, K. A. Visual performance/workload of helicopter pilots
during instrument flight (USAARL Rpt. No. 78-6). Fort tucker, Alabama: US Army Aeromedical Research
Laboratory, January, 1978.

23. Simons, R. R. Visual workload as a function proficiency (Draft USAARL Rpt.). Fo-.t Rucker, AL:
US Army Aeromedical Research Laboratory. Report in preparation, October, 1978.

24. Simions, R. R., and Kimball, D. A. Aviator visual performance: A comparative study of fixed and
rotary win& aircraft (USAAJ. Rpt.) Fort RLcker, ALý US Army Aeromedical Research Laboratory. Report
in prepai ation, October, 1978.

25. Simmons, . R., Lees, 1. A., and Kimball, K. A. Aviator visual performance: A comparative study
of a helicopter simulator and the UK-1 helicopter. Paper presented at the NATO/AGAIR Aerospace
Specialists Meeting, Fort Rucker, AL, May 1978. (To be published ia the Conference Proceedings.)
 OWN-
11, 1 P

78

Figure 1. Visual Recording Equipment

ICA

CF

.74

-3-<
--

12IC

.14 - ý = -- - ----- - - - - - - - -

Figure 2. Graph of CF/Zone

Is---'
 79

Too

4 747A

010 CLIME CRISli DRISCII C1.1 Inc DISCINVIMO LIVIL TUAN ILI
IUIN. lII'N

Figure 3. Graph of CF/Zoite

-uA -1

"-S.---1

40

3' 7

,I.-
tic CtIms CEUSIs DSICINI CLIMBING WORDINING hIsL WIBN HII
TURN TURN

Figure 4. Graph of CF/Zone

 800

90 , - ' ' rn - ., -r
n .ICA~r.
----r

I00 i

NCA-

90
B0

S, ,

-0 20 - 1

'2
1o

~00- ItO CLIM CEUSI DISCINI CtItMINO DISCINOINO LEVIL 115

MANIUVII U TURN TURN

Figure 5. graph of CF/Zone-Fixed Wing Aircraft

34 -------------------- ------------------

U-21----
S.

[NDNIU a 1%ll It

r 4

liD €IIMO CIW

III IC011N CUIMBING I*IICINDINO L|IL It

VIU SUM
MA N UItN 1U.tN

Figure 6. Zone/CF for IZQ/Aircraft

e.b

|.
 - - - - - - - - - - - - - - - - - - - - - - - -

I0 U"H.I;F"
!09

-5
AN

-4

ITO CLIMB CRUISE IJS

DESCENT ITURNL

Figure 7. Graph of CF/ZoneI

 83

HANDLING QUALITIES, WORKLOAD AND HEART RATE

by

Alan H. Roscoe
Royal Aircraft Establishment
BedforO England

Introduction

The important and close relationship between aircraft handling qualities and pilot workload has bee'
underlined several authors. Twelve years ago Westbrook and his colleagues (1) stated: "To a pilot
the multipie stresses of flight, his workload, are swunerized under his judgement of the handling
qualities." Today, there in an increasing tendency for the pilot to be less of an active controller and
more of a supervisor, but even so, this statement - especially when applied to short term workload -
still holds true. This changing role of the pilot has led to a wider interpretation of the term handling
qualities. Cooper (2) remarked: "Because of preoccupation with manual control in the past, it has been
a general practice to associate handling qualities primarily with aircraft stability and control
characteristics. Actually,, handling qualities encompasses not only the aircraft stability and control
but the total of the pilot-aircraft interface features as well." Most people now interpret the
term handling qualities in this way and it is convenient to do so in this papbr.

Unfortunately, no such agreLment exists about the interpretation of the term workload. It is,
therefore, important for authors to make clear taei'• own interpretation of the term. In this paper, the
basic idea of pilot workload is considered to be effort-related, as distinct from task-or performance-
related concepts. A suitable definition is that given by Cooper and Harter (3): "the integrated physical
and mental effort required to perform a specified piloting task." The idea of workload as effort is one
with which most pilots would agree (4); and it is consistent with the measurement of heart rate as a
means of assessing workload.

Assessing handling qualities and the associated workload is an important .jart of flight evaluation,
whether of control and stability or of guidance systems, and various assessment methods are used by test
pilots and engineers for this purpose. Measurement of performance, which is un important and essential
part of control and guidance evaluation, may be used to estimate changes in both handling qualities and
workload (5) (6). Unfortunately, changes in handling and workload are not always refected by changes in
performance. In 1956, Duddy (7) highligitted the difficulty of estimating the extent of improved stability
in a directionally unstable fighter when fitted with a yaw damper as &iming accuracy was not improved.
As Spyker et. al. (8) have observed: "An evaluation procedure which relies exclusively on perforna-ce
measures is inadequate. That is, a pilot with one configuration may work twice as hard as he does with
another, yet achieve equal performance with both." This ability of pilots to "compensate" is referred
to by Cooper and Harper (3) in their Handling Qualities Rating Scale.

Another method of assessing handling qualities and levels of workload, especially during landing
approaches, is by measuring control activity. Morrison and Stimely (9) quantified pitch activity and
used the results to augment pilot's subjective impressions of workload during noise abatement approaches.
Barber et. al. (10) suaunAcd fnrce inputs from elevator, aileron, and rudder to give a workload factor
during the evaluation of genet-. %viation aircraft handling qualities. Nevertheless, these authors
accepted that using force inputo L. -lve a worklosd factor "... has some deficiences."

Objective techniques, especially if they involve precise measurement, are particularly attractive
to engineers. However, by far the most used techniques for evaluating handling qualities and workload
are subjective. These techniques, which vary from simple comments by pilots to complicated -uestionnslres
and rating scales have, for the most part, benn developed for assessing aircraft handling rather than
pilot workload. A well known aud accepted handling qualities rating scale is that of Cooper and Harper
(3), which refers to workload by asking the question: "Is adequate performance attainable from tolerable
workload?"

Clearly, workload levels for a given task are related to the aircraft's handling characteristics,
but a valid rating for the latter may not always give a reliable estimate of workload. Experienced test
pilots may be quite adept at using opinion rating scales but occasionally it seems difficult to separate
assessments of workload from those of handling qualities, leading to anomalies and ambiguities. Westbrook
and his colleagues (1) commented that: "If a reliable method were available to obtain a measure of
workload or stress, it is undeniably true that many of the anomalies in handling qualities data could be
explained."

Several investigators have recorded physiological variaible from pilots in real and simulated flight
as a means of estimating levels of stress and workload. This paper, by describing two current flight
trials and by referring briefly to previous studies, examines the relationship between pilot's heart rate
and subjective assessments of handling qualities and workload.

Materials and Methods

All the subjects referred to in the following examples were qualified test pilots who were experi-
enced and current on aircraft type. Most of the flight trials invo:&ved either the take-off or the
approach and landing and so the task was well defined and realistically demanding. Performance was
closely monitored by on-board intrumentation and by airfield sited kinetheodolites. Whenever possible
flight trials were designed in such a way that experimental variables could be compared during the same
sortie. In this way the affects of weather, learning and other irrelevant influences were miniaised.

Various aircraft, ranging from pure research to representative civil and military types, were used
 84

Pilot assessments of handling qualities and workload were made by using the Cooper-Harper scale,
by straight-forward comments, or by a questionnaire designed for a particular trial. In most cases the
pilot recorded his comments or gave a rating while in the aircraft; questionnaires were completed aftc
:i landing. Latterly, a formal workload rating ca!z, based an the Cooper-Harper handling scale, has bee,-
constructed and Is currently being evaluated.

At Bedford, pilot's heart rates are obtained by recording the EGG signal in analogue form with the
"R"
wave being used to trigger a cardiotachometer.The resulting beat-to-beat rate is then plotted
against time for initial examination and analysis (Fig. 1). Subsequently, mean heart rates for
consecutive 30 sec epochs and mean values for a particular flight phase or sub-phase are used to compare
levels of workload.
Flight Trials

I 1. "Ski-jump"' Ramp take-offs.

The first example is of a trial to ass-ess the advantages of using an inclined ramp to improve the
take-off performance of shipborue H S Harrier VTOL combat aircraft (11). The aircraft is accelerated on
ski-jump ramp from a short run (50-100m) with nozzles rotated rearwards. At the top of
to
the the
ramp, and on shaped
the point of becoming airborne, the nozzles are rotated downwards to a pre-set angle.
After reaching conventional flying spced the nozzles are rotated back to the aft position.
One of the Harriers used in the trial, a two-seat version (T3), is equipped for telemetering heart
rate fromn both cockpits. A portable ERG recorder (Oxford Instruments) is used to monitor pilot's heart
rate from the other (s:ngle seat) aircraft.
&o Two or three ramp take-off s are normally carried out on each aortae with the take-off weight and
distance being varied between runs. The ramp was set at an angle of 6 for the first series of take-offs
and at 9° for the second series. IncremeTnttl increases in ramp angle are planned for future stages of
the trial.
Performance measurements are essential in this trial but assessments of handling and workload,
especially for the critical period of accelerating along the ramp dbring the partially jet-borne flight
phase, are also important. Pilot's heart rates together with subjective ratings of handling qualities,
using the Cooper-Harper scale, and of workload, using a special ten-point rating scale have been obtained
for a large number of take-offs.
Three test pilots, flying the T4, proved the bulk of the heart rate data and subjective ratings.
Take-offs with and without autostabilisation, during simulated and real night conditions, and in
crosswind. up to 10K wers evaluated.
Table 1 gives overall pilot ratings, and mean heart rates for 40 sec epochs centred on ,he time of
nozzle rotation. This epoch includes a period of from 10 to 15 sec before releasing the brakes, a time
when pilots carry out a final check of instruments and configuration.

TABLE 1

60 HARRIER RAMP TAKE-OFFS ("NORMAL" CONDITIONS). MEAN HdEAT RATES (40s),

HANDLING RATINGS (COOPER-HARPER) AND WORKLOAD RATTNGS.

Pilot n - Heart Rate bpa HQ Rating WL Rating

A 15 110.9 4 4
B 11 119.3 4 4
C 5 93.6 3 4

Fig. 1 is a typical beat-to-beae heart rate plot for the handling pilot during a ramp take-off flowtu
from the front seat. Close examination shows that his heart rate increased some 10--20 sec befnre going
to take-off power prior to releasing the brakes. Pilot comments confirm that the workload Jue-reases
rapidly at this time and remainds high until conventional glight some 30 to 40 sec later. Overall
assessment of handling qualities and workload, after the 6 stage of the trial, were favorable and ramp
launches were considered to be easier than normal runway short take-offs (STOs).

Comparison of mean 60a heart rate shows no difference between types of take-off if the epoch iocludes
the 15-209 before rolling. Howev-.r, if the epoch starts when brakes are released the runway STO heart
rates tend to be 2-3 bpm higher (Table II). The finding agreen with ratings for handling qualities and
for workload. The influence ,f ground effect, during runway STOs, seems to caise a deterioration in
handling with a consequent Increase in workload.

Neither night take-offs, both simulated and real, nor crosswinds up to 10K caused any difficulty and
resulting heart rate values and ratings were similar to those for 'normal' ramp launches. Take-offs in
the unft bilized mode tended to result in higher ratings with marginally increased heart rates. Results
og the 9 ramp evaluation showed ratings of handling qualities and woraload to be similar to those for
6 . Heart rates, which were slightly lower, agreed witD pilot opinion that 9 ramzp take-ofes were no
more gifficult and could well be easier than those at 6 . Pilots commented on a smoother ride along
the 9- ramp.
 85

F IEt 1. HS H I E A H R O G 1

LE II
TA1I I P
Pi l4t 1 2 1

A- "la Ac ue -- i -- i I

N"i -sijm-Tk-f

•'FIGURE 1. HS
A HARRIER.
S~
BEAT-TO-BEAT
110.8.• 10.HEART RATE AND NOZZLE
109.
TABLE II
1108ANGLE RAMP TAKE-OFF i
COMPARISON OF HARRIER STO MEAN HEART RATES (608)

• • 60
6 Ramp Runway

iPilot 1 2 1 2

SA i110.8 108. 9 109.3 110.8

S(n - 15) (n - 21)

B 116.5 114.8 115.9 117.3

(n-li) (n - 6)

C 90.4 89.6 90.8 91.2

(n -5) (n -18)

I Epoch from 15 - 20s before releasing brakes.

2 Epoch from releasing brakes.

2. Direct Lift Control.

A modified SAC 1-11 is currently being used to evaluate the benefits of using Ilrect Lift Control
(DLC) to improve handling and perf'rmance during the approach and landing. DLC should enable the aircraft
to be flown more precisely on the glide slope and alsai result in better all round landing performance
with less touchdown scatter. Workload during this pha•e should be reduced and the ability to cope with
turbulence improved. More direct control of descent rave should prove beneficial during steep gradient
approaches and improve safety during the flare.

In addition to monitoring aircraft performance, pilot assessment of handling and workload, and
measu'ement of pilot's heart rates, are important trial requirements.

The DLC system fitted to the 1-11 uses the four wing spoilers to generate lift changes; these are
. controlled by electrical sensing of control colu'ii pitch movements. A wash-out circuit is inserted into
the system to provide the pilot with relatively normal acceleration responses to pitch inputs.

The first batch of flying (phases one and two) was concerned mainly with op..imising the control
characteristics and giving the piloe' .ome experience of tew handling techniques. Control was rated
batter after the DLC was ma 'I"nes octer ctill after a lag was incorported in the DLC signal frow
the control column. Pilot j monitored on only a few sorties during this stage of the
trial.
 ~86

The main flAght investigation (phase three) was aimed at evaluating DLC during both conventional 30
approaches and 6 steep gradient approaches. Each sortie Included a batch of basic-aircraft runs for
comparison with DLC.

For most sorties pilots were briefed to concentrate on a precise position at 50 ft and glide slope
tracking was not appreciably better. But on the occajions when pilots were briefed0 o maintain precise
Slide slope track:Lng, performance improved. This was particularly evident on the 6 approach gradient.
Landing performance was definitely better when off steep approaches but not noticeably different when
from 30 approaches. The ability of DLC to quickly arrest descent helped to produce more accurate and
smoother landings, but if the flare was started too early there was a tendency for pilot induced oscil-
lations (PiOs) to occur.

Overall pilot assessments indicated that the aircraft's handling qualities in the landing configura-
tion were improved with DLC. Workload during the approach and in the flare was thought to be generally
lower, but especially so fcr the 6 glide slope. Heart rate responses appeared to agree with pilot
opinion although, at first there were a small number of discrepancies. These were resolved after discus-
sion with the pilots. For example, when PIOs occurred heart rate responses for the flare epoch were much
higher; the overall effect was to result in mean values for this manoeuvre which were similar whether DLC
was used or not.

Perhaps it is not surprising that pilot ratings for the flare, both of handling and of workload,
varied considerably accordivg to whether PiOs were present or not!

Fig. 2 shows mean heart rate values for 30 and 60 approaches, with and without DLC, flown in similar
weather conditions by one of the three project test pilots. These results show an obvious trend in favor
of DLC but the only appreciable reduction in heart rate is during the glide slope interception and early
part of the 60 approach.

HEART HEART
RATE RATE
bps bpe

I1O PILOT A lio PILOT A

100 -too

90 O90

70 TO1
70To

Sso SECOND EPOCHS o0 SECOND I.POCHS

L. (15) 3' SASIC AIRCRAFT e,. e (0) Go SASIC AIRCRAFT APPROACHES

.- --. (13) 3 DLC -- .,- (5) Go DLC APPROACHES

FIGURE 2. BAC 1-11. MEAN 30s HEART RATE VALUES FOR DLC AND BASIC-AIRCRAFT

Results for one of the other two pilots were similar to those illustrated, again showing a definite
trend in favor of DLC. Mean heart rate values for the third pilot did not differ appreciabty between DLC
and basic ailcraft approaches. In fact, because PIO seemed to disturb this pilot more mean rate for the
flare from 6 approaches wda slightly higher with DL•. He asseused handling qualities overall as being
improved with DLC, he was unsure about workload on 3 slopes but felt it was reduced on 60 approaches.

Results from subsequent flignt (phase four), following minor changes to the system, have confirmed
the benefits of DLC. However, because pilots were briefed to fly more accurate glide slopes, heart rate
F. values were not noticeably reduced, but performance was improved. A sortie of 30 and 60 approaches flown
in turbulence provided the opportunity to demonstrate the advantages of DLC. This was confirmed by the
markedly lower heart rates for DLC approaches when compared with basic - aircraft approaches.

It was hoped to carry out sufficient flying to allow statistical analysis of results, but the number
of sorties has been limited and only trends were establisnhed.

Comments

The two trials described above are both typical examples where assessment of handling qualities and
related workload are important features. However, they differ in some respects. The concept of the
'ski-jump' ramp is aimed at increasing the maxim-m take-off weight of ship-borne vectored thrust combat
Laircraft thereby impruving their overall tactical performance. This being the primary objecttve.
Handling qualities and iorkload are of secondary Importance; Jt is only necessary to eiisure they aie not
increased beyond a level which might Jeopardise the take-off.
 87

DLC, on the other hand, is aimud primarily at improving safety during the approach and landing.
Therefore, assessment of handling and workload assumes much greater importance. Of course, Performance,
because of its relationship to safety, is also important.

In both examples there is generally good agreement between handling qualities, workload, and heart
rate. The few anomalies that have occurred, especially in the early stages of the trials, have been
resolved by detailed discussion with the pilots. For example, a high heart rate and high workload rating
but low rating for handling qualities during the first ramp take-off by ole test pilot was due to
excessive anticipation. The pilot rated the workload as 7 (out of 10) and the handling as Coope•>Harper
3. His 4 0 second mean heart rate was 156 bpm. Afterwards he reported: "with the benefit of hindsight,
I realize that I was much more keyed up than I need have been, and I expect that my workloed will be very
much less on subsequent launches as I gain experience: His subsequent workload ratings avera3ed 4 with
autostabs on and 5 with autostabs off and the corresponding hangling assessments were, similarly, Cooper-
Harper 4 and 5. The overall mean heart rate level for eleven 6 take-offs was 119 bps. The high heart
rate generated by the first ramp take-off is typical of the increased arousal experienced by test pilots
a bout to carry out a novel flight task. Roscoe (12) has suggested that experimental test pilots frequeuat]
overestimate the level of difficulty for the first run of an untried or unusual manoeuvre or Lask.

Occasional heart rate measurements of two pilots during the preliminary phases of the DLC trial
resulted in appreciably lower values for the new system compared with the basic-aircraft. These pilots
weie also most enthusiastic in their early comments on the system. It was, therefore, something of a
disappointment to find emaller decreases in heart rate when it was routinely monitored during phase three.
Subsequent discussion revealed that in flight trial proper, pilots had changed their strategy and flew
more precisely than ir. the previous stage. Improvements in handling were apparently being used to increase
performance although this was not always measurable. Occasional distrepancits between heart rate end
pilot opinion were cauted by failure to record the fact that PIOs occurred during the flare.

Previoiis Studies

For some nine years, at RAE Bedford, pilot's heart rates have been monitored during various flight
trials as part of a long term study of workload and ctress. Evaluation of handling qualities was a
primary requirement of many of these trials and it is interesting, and perhaps profitable, to refer
briefly to some earlier ones.

Autostablisation systems should lessen the effects of poor stability and control and thus lead Co
an overall improvement in handline. The VTOL research Short SCi was an erample of, a Telatively unstable
aircraft and pilots comparing the stabilised with the unstabilised configuration invariably commented on
the marked improvement in handling qualities of the former. It is interesting to compare heart rate
responses for one pilot flying two similar 6 min sorties consisting of a vortical take-off to 30m (100 ft),
small accelerations and decelerations, ending iv•a vertical landing. The Yirst flight was stabilised
and resulted in a mean heart rate of 109.6 bpu. the second flight, which was unstabilised, resulted in
autostabilisation, but similar heart rate comparisons for other test piloto did not show any differencea.
Detailed discussion with pilots revealed that most of them suspected the integrity of the autostabilisation
system aud transferred the spare effort made available by improved handling to monitoring the system itself.

Even though handling qualities and workload zre closely related, it does not necessarily follow
that an improvement in handling will invariably lemd to a reduction in workload. It may sometimes be
perferable, especially for well motivated pilots, to Improve Performance and maintain the same level of
workload. If performance is monitored such improvement will be obvious and this alone will indicate the
defree of benefit gained by .mproved handling. Such improvement is evident in some of the DLC flying
referred to earlier. Pilots also make use of additional spare effort or capacity to increase oonitoring
or to carry out other covert tasks which are not immediately obvious.

Gerathewohl (13) made the point that subjective ratings of handling qualities ".... as accurate as
they may be in regard to control desirability or difficulty, do not contribute t' workload determination,
since they are only loosely connected to task demands and pilot response.'" Certainly, as in the above
example, a pure handling qualWiis scale may not give an accurate estimate of worhload. It is clear that
subjective assessments of workload must be derived from rating scales specifically designed for the purpose.

Subjective assessments, in general, as sometimes unreliable, for example, it iv knevn that they are
susceptible to both inter- and intra-subject inconsistency. In particular, subjective ratings for what
may be minimal changes in handling characteristics can be misleading by suggesting the existence of larger
differences, especially if the ratings were obtained under diiferent conditions. Such anomalies ma, be
due to a poorly designed rating scale, because the test pilot has varied his assessment strategy, or
because of undetected changes in flight conditions.

This problem is typified in the trial of a powerful rudder autostabiliser fitted to the BAC 221 slender
delta supersonic research aircraft. Lateral directional handling characteristics during the landing
approach were assessed by three test pilots. The task consisted of a "side-step" minoeuvre at a height
of 75m (250ft) placing the aircraft to one side of the centre line. An "S" turn was necessary to realign
the aircraft with the runway, thereby testing the effectiveness of the system. Different autostabiliser
settings were evaluated and, as it wa. possible to vary these in flight, the associated handling charac-
teristics were compared under similar conditions. The Cooper-Harper rating scale was used for this
purpose. The pilot's heart rate was monitored on several flights so that mean values for each approach
could be compared. It was also possible to examine the relationship between levels of heart rate and
ratings of handling qualities.

Except for the extreme autostabiliser settings and for "Luo autostab" approaches when ratings and
heert rates were appreciably higher, results were disappointing. Heart rate values were inconsistent,
tempting, therefore, to conclude that the differences between the various autostabiliser setting* were
inconsequental and that heart rate mmasurement correctly interpreted this fact.

Disagreements between workload assessment and heart rate, which have been rare, have tended to occur
during relatively undemanding tasks when changes have "ean minimal or unimportant.

Naert Nmrt
1610

110. 110-
Turbvia"~ sI)
Turssms (n.I'

/T/K.
mo .. s. s!

90 ~9 s9
__ __ __0 _ __ _ .3
Laod TO

FIGURE 3. VC-10. MEAN 30s HIART RA;E XALUES FOR NORMAL CONDITIONS IAND
SEVERE TURBULENCE. a. 3 , b . 5 /3 . APPROACHES AND LANDINGS

Different weather conditions can influence handling to varying extents, and turbulence, in particular, 4
causes increased workload by degrading stability and control. This is especially noticeable during a
flight task where accurate tracking is required. Fig. 3 compares the heart rate responses of a pilot
flying two different types of approaches and landings in severe turbulence with mean rates !or similar
approaches flown in relativellt smooth conditions. This a-ample is from a flight trial of noise abatement
approaches using a BAC VC - 10, (14). It can be seen that there are marked increases in heart rate for
both Syp~s of approach though the increase is marginaliy greater for the earlier and steeper section of
thq 5 /3 two-segment profile when compared with tne 3 gradient. These findings agreed closely with the
pilot's assessment of the changed handling qualities and workload. He considered that turbulence increased
the workload more for the two-seiment than for the conventional approach, especially during the acquisition
and early part of the glide slope

Pilots occasionally reveal sou'e degree of bias towards or against a particular experimental flight
condition which, based on falacious reasoning, may affect their judgement and result in misleading
subjective ratings.

In the early stages of a series of flight trials to evaluate various types of noise abatement
approaches initial pilot opinions of 7h°/3 0 two-stage flare approaches, in a HS Andover, were unfavorable
Pilots felt instinctively that transiting !rom a 7½° slope to one of 30 at a height of 200 ft would be
too demanding and they rated the workload quite high. However, from the beginn•ng of the trial heart
Yate responses for this approach profile were similar to those for convential 3 approaches. Careful
,hought and subjective re-analysis, by the two test pilots who flew early sorties, led to a review of
their original subjective easessment of workload. Subsequently, these pilots, together with other
participating pilots, terided to prefer the 7V1/3° approaches to the 30. They considered that improved
handling on the steep segment was an important factor in mairtaining a reasonable level of workload.
Heart rate and workload assessments showed good agreement (15).

In this instance both the task and the handling qualities are changed but the net result is that
workload and heart rate are unchanged.

The pvevious trials were concerned mainly with stability and control which tc.many people used to
be the accepted interpretation of the term handling qualities. But, as stated earlier, evaluation of
guidance systems is aiso relevant; indeed a large proportion of test flying at Bedford is directed to
evaluating approach guidance displays and systems.

A typical trial, which took place in 1971, was to evaluate an airborne visual approach indicator
(VASIO presented as a HUD. For the purpcse of the experiment only omnidirectional runway edge lights
and green threshold lights were used. All other lights were astinguished and moonless nights were
selected for the trial sorties. Two pilots alternated as experimental pilot (P1) and co-pilot (P2) for
four sorties giving a total of 32 approaches. Each run ended in an overshoot at 30M (100 ft). Four
difference runway& were used in ordeo to reduce any element of learning, though, as it happened, varying
weather conditions eliminated this effect. The aircraft used for this stage of the trial was an HS
Comet 22.
 89

Glide slope performance was considerably batter with the HUD-VASI and the heart rate of the handling
pilot was reduced. Compared with no-aid approaches, the overall decrease was 4.2 bpm for one and 6.8 bps
for the other pilot; whereas there were only negligible differences in heart rate when acting as co-pilot.
Both pilots were keen to point out that workload was significantly reduced by the HUD and they anticipated
larger decrasee in their heart rates but agreed that the benefit of the aid would have been greater had
the experimental approaches ended in a landing rather than in an overshoot.

Improvement in performance without any evidence of increased workload was, in itself, adequate proof
of the advantages of the HUD-VASI. Nevertheless, the trial scientists were delighted to have evidence of
a reduced workload as well. Unfortunately, because of the severe and widely differing weather, the wide
variatious In heart rate between the sorties precluded statistical significances.
Thase studies, which have used examples of data obtained during operational test flying, demonstrate
the viseof heart rite as an indicator of workload. Such data have proved to be of great value in the
overall study of pilot workload, but it has to be admitted that the direct value of heart rate measure-
ment in evaluating handling qualities is still not clear. Nevertheless, pilots and engineers associated
with these trials consider monitoring heart rate to be a worthwhile adjunct to those techniques commonly
used in flight evaluation. It should be noted that these examples have been confined to trials where the
pilot has been handling the controls. Heart rate changes for handling pilots have, for the most part,
proved to be reliable indicators of important changes in workloadi when the task has been realistically
demanding. But in other trials at Bedford, where the pilot has been in a monitoring role his heart rate
responses did not appear to reflect changes in workload with anything like the same reliability. This
difference in heart rate sensitivity, between the pilot In the control loop and the pilot outside the
loop, is important.

Discussion

A large number of reports on aircraft handling trials refer to related levels of pilot workload.
However, it is patently obvious that in most instances assessing handling characteristics was the primary
objective whereas estimating workload was very much a secondary aim. This approa..la is usually adequate
and leads to realistic estimations of workload, but it is apparent that sometimes a pilot's main concern
with handling has adversely affected his ability to assess workload. Ellis (16), in pointing out that
it is important that ratings for workload and handling qualities are not confused, wrote: '"henpilots
are asked to make a formal assessment of workload as a primary measure, it should be absolutely certain
that workluad is the ultimate aim of the exercise." Ellis also observed: '"orkload is always important
in handling qualities investigations and so pilots should be encouraged to coment on it and rate it but
workload should not be allowed to usurp the place of the handling qualities rating where the latter il
the more appropriate measure."

It is clearly an advantage to use a specially constructed rating scale for assessing workload during
flight testing. Unfortunately, such scales suffer from the same problems and attract the same criticisms
as do pilot opinion scales for assessing handling qualities. By using some other method of estimating
workload it may be possible to augment pilot opinion and, perhaps occasionally, resolve anomalous findings.
I Physiological variabl2s, which have been recordad by many research workers during studies of pilot workload
and stress, may be used for this purpoce. The literature, though, contains few reports where the relation-
ship between handling qualities, workload, and physiological responses has been studied in detail.

In 1962 Roman and Lamb (17), in discussing the results of seasuring heart rate in flight, observed
that: "Pulse rates correlate well with the pilot's estimates of the difficulties connecting with
handling the aircraft during any one phase of flight." Rowen (18) printed out that high heart rates
recorded from the pilot of the M2 lifting body were associated with the poor lift/drag characteristics,
which made particularly heavy demands on pilot skill. By measuring pilot's heart rates, Billings et. el.
(1) demonstrated that helicopters fitted with hydraulic boost systems were significantly less demanding
to fly. Hasbrook and his co-workers (20) used heart rate measurement to augment pilot oplLnion during the
flight evaluation of a new instrument display. They were able to show that the ncw displey, which
reduced panel space by 25Z, was an acceptable alternative to the conventional display.

These examples from the literature demonstrate a relationship between handling characteristics and
workload as indicatsd by heart rate. But what is the extent of this relationship? Is it reliable and
consistent? Can it be usefully employed in flight evaluation? Is monitoring pilot's heart rate during
test flying a practicable exercise?

The technique of monitoring heart rate is relatively simple, it does nut intrude into the flight
task nor does it compromise flight safety. It is readily accepted by pilots. In fact, Bedford test
pilots have co-operated to the extent of applying their own electrodea and preparing their monitoring
equipment for flight on many occasions. The resulting heart rate data are often studied with interest
by the pilots who find them helpful in recalling vzrious aspects of the sortie.

Heart rate does not give absolute values of workload and so in order to obtain meaningful results
it is necessary to use it as a comparative measure. It is worth noting that pilot rating scales, though
appearing to give absolute values, because they are subjective, are really scales of comparison as well
(21).

The examples of flight trials presented in the previou., section relied on comparison with another
experimental condition or with some form of datum; anJ wherever possible the comparison was made during
the same sorties. The trials by Billings ard his colleagues (19), and by Hasbrook et. al. (20), referred
to above, were similarly Lhmed on co-iparison.

To compare the changes In workload caLsed by different handling characteristics it is necessary to

ensure that otiher aspects, such as the flight task Itself, remain constant. This is sometimes difficult
to -ahieve. For ezample, different flap settings on the approach may change handling but may also require
90

Subjective assessments of handling and workload appear to be more consistaet when they demand a
high level of piloting skill than when they require little effort. Likewise, heart rate measurements
are more consistent and reliable if the experimental flight task if realistically demanding. Data from
the Bedford studies show that test pilots have, for the most part, given estimates of workload which
agreed well with their heart rate levels. But the agreement was better when heart rates, and presumably
workload levels, were higher; and as might be expected, anomalies tended to occur more often when workload
and heart rates were low. It is interesting to note, though, that from measuremnt of heart rate and
finger tremor, Nicholson and his co-workers (22) concluded that: "...... high workload associated with
difficult approaches and landings rendered the pilot's subjective asbessment more variable."

Results from workload studies made in real flight are generally more reliable than those made in
simulated flight and this must be particularly so in studies of handling - related workload. Unfortunately,
it is difficult to set up well controlled flight experiments and so there is a strong temptation to resort
to using laboratoriea and simulators for workload investigations. Protagonists of these techniques point
to the undoubted value of research simulators in assessing handling qualities. But laboratory and
simulator experiments tend to restrict the number of input parameters to which the pilot is assumed to
respond. in real life the pilot is faced with a wide range of input information - much of it redundant,
but all liable to have some effect on his behavior and hence his workload.

As noted in the HUD trial, the inability to control such variablee as weather, and the mall number
of experimental sorties - limited by the high cost of flying aeroplanes - often results in differences
in heart rate which are not statistically significant. Nonetheless, trends, especially if they support
pilot opinion, may be quite adequate, but even if they conflict, heart rate data can be most valuable
in attracting attention to possible ambiguities. Further examination and discussion with the pilot may
then reveal previously undetected factors. Beat-to-beat heart rate is particularly useful ia identifying
short term changes in workload which may not be obvious to a pilot making an overall assessment.

Unfortunately, most of the flight trials at Bedford did not use numerical rating scales for assessing
workload and some trials did not use them for assessing handling. These omissions, together with the
limited number of experimental sorties, has precluded any opportunity for statistical analysis of the
relationship between handling qualities, workload and heart rate. A flight trial designed to examine
this relationship more closely is currently underway. Three test pilots, flying three different aircraft,
compare handling characteristics during various demanding tasks. These include the approach and landing,
low-level high speed flight, and formation flytng. Pilots use the Cooper-Harper scale for rating handlung
qualities and a 10 point scale (based on Cooper-Harper) for rating workload. Heart rate is recorded on
all sorties.

It is hoped that this investigation will result in enough data to permit some degree of statistical
analysis. It is worth noting, though, a point made by HcGregor (23) who stated: "One of the criticisms
of numerical pilot rating scales as opposed to adjectival scales is that statistical games will be played
winh numbers that are not statistically meaningful." He continued: "If statistical indices are used they
must be adequately enough defined to enable the reader to assess their validity and sufficient data
presented to allow a check to be made of the results." With this in Rind, it is not intended to attempt
to identify any mathematical relationship between assessments of handling or workload, based on rating
scales, and heart rate. The individuality of pilots sakes this virtually an impossible task anyway.
Summary and Conclusions

There is obviously a distinct advantage in augmenting a pilot's subjective assessments of handling

qualtties and workload. This paper presents practical examples where pilot's hearc rates have been used
to augment their opinions of handling and workload during various flight trials. These studies provide
good evidence to show that, in general, this technique gives reagonably good indications of the workload
generated by particular handling qualities.

The technique of monitoring heart rate is simple, it is accepted by pilots, and it is compatible
with test flying. To improve reliability and consistence the flight task should be realistically demanding
and require the pilot to be in the handling loop. Comparisons between experimental conditions, or with
some form of datum, give more maouingful results; wherever possible comparisons should be made during the
same sorties. Raw data in the form of beat-to-beat heart rate are invaluable for revealing rapid and
short duration changes in handling qualities which affect workload.

In this way, potentially misleading results can be identified in good time, thereby drawing attention
to the need for further investigation. Anomalous findings may be resolved by examination of heart rate
data and by discussion with the pilot.

The author has not made any attc•-it to satisfy strictly scientific criteria, the primary objective
being to draw attention to the value of using heart rat* as a flight test procedure. But in addition,
it is hoped to stimulate thought and dis 'ission so that it may be possible to redu..e soma of the anomalies
found in handling qualities and workload,, referred to by Westbrook et. al. (1).

REFERENCES
1. Westbrook, C. B., Anderson, R. 0. and Pietrzak, P. E. Handling qualities and pilot workload,
Conference Proceedings No 14 Assessment of skill and performance in flying, ACARD, Paris, 1966.

2. Cooper, G. E. The pilot-aircraft interface, Symp Vehicle Technology for civil aviation, seventies
and beyond, NASA SP - 292, Washington DC, 19)1.

3. Cooper, G. and Harper, R. P. The use of rilot rating in the evaluation of aircraft handling
qualities, NASA Tech Note TNeD - 5Y53, Washington DC, 1969.
 91

4. 1llic, C. A. and Roscoe, A. H. Pilot workload: A survey (In preparation).

5. Brickson, C. A. Pilot landing performance under high workload conditions, Conference Proceedings
No 146, AGARD, Parris, 1974.

6. Lees, N. A., Kimball, Z. A. and Stone, L. N. The assessment of rotary wing aviator precision
performance during extended helicopter flights, Converence Proceedings No 217 Studies on pl'ot
workload, AGARD, Parris, 1978.

7. Duddy, R. R. The quantitative evaluation of an aircraft control system, AGARD Report No 29,
NATO, Paris, 1956.

8. Spyker, D. A. at. al. Development of techniques for measuring pilot workload, NASA CR-188,
Washington DC, 19717.

9. Morrison, J. A. and Stiaely, R. L. An operational look at the two-segment approach, AIAA Paper
No 74-979, !974.

10. Barber, M. R. et. al. An evaluation of the handling qualities of seven general-avlation aircraft
Technical Note, NASA TN D-3726, Washington DC, 1966.

11. Coleman, H. J. Harrier 'Ski Jump' launch studied, Aviation Week and Space Technology 105,
pp. 17-18. December 1976.

12. Roscoe, A. H. Stress and workload in pilots, Aviation Space Environ Ned (In press).

13. Gerathewohl, S. J. Definition and measurement of perciptual and mental workload in aircrew and
operators of Air Force weapon systems, A status report, AGARD CP-181 Higher Mental Functioning
in Operational Environments, AGARD, Paris, 1976.

14. Roscoe, A. H. Pilot workload during steep gradient approach, Conference Proceedings No 212 Aircraft
operational experience and Its impact on safety and survivability, AGARD, Paris, 1976.

15. Roscoe, A. H. Heart rate monitoring of pilots during steep gradient approaches, Aviation Space
Environ Mod 46, pp. 1410-1415, 1975.

16. Ellis G. Pilot opinion measures, Assessing pilot workload, A. H. Roscoe, ed., ACARDograph AG-233,
AGARD Paris 1978.

17. Roman J. A. and Lamb. L. E. Electrocoodio~raph in flight, Aerospace Med 33, pp. 527-544, 1962.

18. Rowen, B. Biomedical monitoring of the X-15 program Air Force Flight Test Centre Edward AFB,
TH - 61-4, 1961.

19. Billings, C. 1. at. al. Physiological cost of piloting rotary wing, Aerospace Ned 41,
pp. 250-258, 1170.

20. Pilot performance and heart rate during in-flight use of a compact instrument display, FAA Office
of Aviation Medicine, Report No FAA-AM-75-12, Washington DC, 1975.

21. Pilot rating techniques for the estimation and evaluation of handling qualities, AFFDL-TR--68-76
Wright Patterson AFB O, 1968.

22. Nicholson, A. N. et. al. Activity of the nervous system during the letdown, approach and landing,
A story of short duration high workload, Aerospace Med 41, pp. 436-446, 1970.

23. McGregor, D. M. Lead discussion, Conference Proceedings No 106.19 Handling Qualities Criterior,
AGARD, Paris, 1971.
 (• 93

BRAIN WAVES AND THE ENHAZCD4DIT OF PILOT PERFOR1XANCE

by

G. H. LAWRENCE, Ph.D.
Office of Naval Research
Physiology Program (Code 441)
800 North Quincy Street - Room 433
Arlington, Virginia 22217

The use of brainwaves (BEC) for the enhancIent of the performance of aircraft pilots is an idea
which requires, for its development, the integration of two previously independent linrs of -search
endeavor: human performance assessment and central nervous system neurophysiology. A human performance
research paradtgm specifically relevant to the study of pilot performance, in the context of which the
use of brain waves my feasibly be studied, will be discussed later. Attention is now directed to the
state of the art of brain wave research and brain-behavior relationships, specifically those aspects which
are considered to be feasibly and usefully applicable for potential use in simulated aircraft crew stations
or eventually in a reel-world environment.

BASIC IB81AICCH

Two basic types of paradigms have been employed in studies of brain waves and perforaance. In the
first case, spontaneous, ongoing IEG is nonitored, and frequency and amplitude for some time period
(usually preceeding, during, and/or succeeding some experimental treatment) are related to various aspects
of performance. Usually this sort of treatment has included use of an intervening variable called, most
frequently, activation or arousal; Davies and Parasurasan (1977) have identified four separate types of
experimentation within this rubric. The first type of study "has attempted to discover whether decrements
in (signal) detection rate, or sometimes increments in detection latency, are paralleled by corresponding
changes in one or more pyachophysiological measures. *" Another approach has attempted to identify
psychophysiological (not only EEG, of course) processes and events which discriminate between periods
preceding successful as opposed to unsuccessful attempts to detect a signal. A third has involved varying
environmental parameters presumed to affect arousal, thereby causing an ultimate effect upon performance
if in fact arousal and performauce are related. Frequently these studies have led to the observation of
performance changes without concomitant variation in physiological measures of arousal; as Davies and
Parasuraman have puti . . a dissociation of performance indices and physiological measures occurs."
A fourth approach has attempted to predict individual differences in level and quality of performance
froja baseline scores on physiologinal measures, through the so-called arousal hypothesis of vigilance.
Generally, this hypothesis posits an inverted U relationship between arousal and performance; at low
levels of arousal errors of omission (e.g., missed detections of targets) occur, and at high levels the
well known detrimental performance effects of stress and high anxiety are seen. Arousal is measured
independently, usually via physiological events.
From his 1970 review, however, O'Hanlon is led to the conclusion (q%.oted by Davies and Parasuraman
1977) that "No reliable physiological index of &!tartness has been accepted, although several promising
ones have been proposed. No physiological variables have been found that are as sensit-ve to task and
environmental effects as is performanze. No underlying process has been so clearly defined as to permit
rational control of cerebral vigilance..' Davies and Parasuraman agree with his discourgaging assessment,
resultof both number performance
from inadequate measurement and from
the
and lack of an
suggest thatextant deficiencies
task taxonomy
methodological to make sense the rnormous of &xwperimental situations which
have been used. Beck in his 1975 paper considers that brain waves cannot feasibily be brought under
stimulus control, and therefore does not include EEC studies (as constrasted with evoked potential studies)
in his review.

Nevertheless, a few semi-consistencies have been observed across experiments and laboratories,
although virtually no general conclusion can be put forward which does not admit of some exception or can
be considered invulnerable to challenge. Certainly decrements in detection performance, rate, and or
latency are usually seen to be accompanied by EEC changes as fatigue develops over the course of lengchy
experimental sessions, and it is clear that the probability of failure to respond to transitory sig.aals
altogether increases considerably under conditions of lowered arousal - i.e., when the subject is bored
or sleepy. Further, several studies have indicated that individuals with greater baseline/ability for
GSR (though not, oo far, for EEC) do better indetection situations and generally are better able to
maintain a state of vigilance. Variation in the complexity of visual stimuli, memory task requirements,
and differential Lemispheric activation via varying stimulus modalities have all been shown to affect EEC
measures in relatively stable wiys. Brain waves have been reliably shown to vary with behavioral sleep
events in a relatively stable manner, and consistently over the time course of a normal nights sleep,
and thus probably reflect variation in arousal (albeit at the very low end of the scale). Gales (1977)
points out that ". . . high alpha and beta frequencies are more sensitive to discrete changes in
stimulation than are lower alpha frequencies and theta activities. . ." i.adicating that the relationship
of EEC events to arousal is more easily studied in alert states. Gales' statement is made in summary,
and follows a passage in his paper where he acknowledges that arousal is not a unitary state which has
straight-forward and systematic relationships with measures of behavior or of subjective report. He goes
on to say that changes in theta and lower ranges of alpha reflect other, presumably Pon-task-relevant,
effects.

Other attempts to relate ongoing EEC experimentally to vigilance, or to other behavior, especially
unenr conditions of fatigue have been made. Consistent with data from other situations, it is typically
fo•nd that changes in fatigue and arousal can be inferred from brain wave activity (power shifts fror.
higher toward lower frequencies, i.e., from beta toward theta and delta)

Prepared for the Environmental Physiology Program, Office of Naval Resaarch, USA
 94

The work of O'lHanlon and deatty (1977) supports the general form of the arousal hypothesis of
vigilance, showing that percc-.tage of alpha and theta increases and that beta decreases were related to
variation in performance on a simulated radar watching task. As Beatty and O'hanlor point out, alpha
may either increase or decrease with arousal; concurrent changes in theta and beta must be taken into
account (i.e., are frequencies increasing or decreasing?) before sense can be made of the variation in
alpha.

Finally, there is 3ome indication (e.g., Dimond 1977) that hemispheric differences say relate to
arousal and vigilance performance. Split brain studies suggest that the two hemispheres way have
different vigilance systems. Perhaps, as suggested by Jerison (1977), the left hemi ihere deals with
selective attention and the right with continuous attention.

It is probably a fair statement that there is not much promise of new and exciting use of ongoing
EEC for the enhancement of pilot performance at the moment. There appears to be sufficient consistency
in the literature so that some confidence may be felt in the use of changes in brain wave power across
frequencies to infer rather general stats changes. One can tell when a subject is getting drowsy, has
gone to sleep (and hrief bursts of sleep frequently appear under sustained task performance requirements,
especially when some degree of sleep deprivation exists), or to a lesser degree of certainty, Is si;-.4l
inattentive. This kind of information is not without interest and use, but its lack of information
specificity and very low data rate lead to thA conclusion that the Instrumentation and data processing
requirements to collect and act upon it would not likely pay off in greatly enhanced performance, though
its potential for monitoring organismic state is obvious.

The event-related potential (ERP) is another matter. The ERP is an EEC response evoked by a specified
stimulus and usually averaged over a group of trials. A series of positive and negtative deflections
L is observed, usually conceptually and empirically divided into two categories. The earlier components,
those occurring in the first 100 ms or so subsequent to the stimulus, are referred to as exogenous-they
reflect characteristics intrinsic to the stimulus event itself, such as louJness, brightness, intensity,
or other psychophysical attributes. This activity is considered to represent the processing of sensory
information. The later components, up to perhaps 600 ma beyond the stimulus, are considered to be
endogenous, reflecting cognitive processes and attributes of the stimulus deriving not from its physical
properties but rather from its task-relevant context (e.g., whether it is to be counted or ignored, its
suprisingness, its information value, etc.). It is ;hec* latter components, reflecting as they seem to
aspects of performance potentially applicable to cockpit or crew station situations which are of primary
interest. The following discussion of these later ERP components and their studied relationships is
largely derived from comprehensive and thorough reviews by Donchin et al (1977) and Beck (1975), and to
a lesser degree from the recent chanter by John and Schwartz (1978). The catalog of endogenous components
offered by Donchin et al (1977) is worth presenting in full:
N200. This component is elicited uhenever a rare or unexpected
event occurs. It is of rarticular interest because it car, be elicited
by stimuli that are in the periphery of the subject's attention. Unlike
the other endogenous components it appears sensitive to the modality of
the stieulus. ITe positive-gotug return of this component is sometimes
labeled P3a.

P300. This robust endogenous component is reliably recorded in

association with task relevant, rare stimuli. We apply this laiel to
a component whose latency may range from 275 to at least 600 msec.
It is characterized by its scalp distribution as it tends to be larger
in tlve central and parietal electrodes. It is further characterized
by a -ery specific response to experimental manipulations.

Slow Wave. This is a slow potential sh-.ft that, as far as is

known, is affected by the same variables that affect P300, except
that it has a different scalp distribution. Whereas P300 appears
largely as a positive-going potential peaking on the parietal
scalp, the Slow Wave is positive-going at the parietal electrodes
and negative-going in frontal electrodes. As the Slow Wave is so
closely associated with P300, it will not be discussed further in
this report.

The Contingent Ne, _ive Variation (CNV). This term is normally

used to describe the sl~w negative shift of potential that occurs
the warned foreperiod preceding a motor or mental task. It begins
approximately 400 msec after the warning stimulus and, normally,
terminates after the imperarive stimulus, that is, the stimulus
demanding a response or decision by the subject.

The Readiness Potential (RP). As the CKV, this is an event-

preceding negative shift. It is distinct in the sense that it
appears prior to self-paced voluntary responses. Its occurrence
is independent of the presence of an eliciting, or command, stimulus.

As John and Schwartz (1978) stated, these endogenous componeats have been studied in connection with
arousal, attention, selective attention, emotional valence, assessment of novelty, time estimation,
uncertainty, detection of targets, differential identification of stimuli indepedent of size and shape,
and the semantic classification of linguistic symbols.

Some of the more potentially useful and applicable findings that have occurred with some consistency
are:
"7n 95

o it is clear that ERP component amplitude is related to attention For example, P300 (whose
latency actually varies from 250 to 600 me) is rather significantly amplified by the perception of a
meaningful or suprising stimulus (or by the absence of a stimulus, when one is expected); in other words,
by the resolution of uncertainty. This basic finding has been otserved in situations involving reaction
time, signal detection, rignal confirmation, pattern completion, motor set, as well as other experimental
paradl-se. The findiiti. that P300 occurrence follows omission of an expected stimulus seems especially
interesting, demonstrating clearly that this component reflects cognitive rather than sensory processing.
The potentials evoked by missing stimuli are indistinguishable in confirmation, from those evoked when
sensory stimuli are in fact presented. Results of studies utilizing the omitted stimulus paradigm allow
4
the inference that an nternal model of sequences of stimulus events is formed, and that the P300 is
evidence of a mismatch between this model and the observed (non) event which unexpectedly does not occur.

o in general, the P300 is enhanced only when etimius inormation is being actively processed and is
uniquely associated with the occurrence of a signal and its correct detection. (Beck 1975). It occurs
subsequent to stimuli in any sensory modality. EvWdence exists that amplitudds and latencies vary over
the scalp, cnd appear to interact with different task analysis requirements for cognitive processing.

o there is some indication that auditory ERP's vary interhemisperically as a result of a linr.istic
task; left-side responses are larger when a task requires linguistic analysis but not when the same stimuli
are compared non-linguistically.

o John's (1978) review of P300 variation relating to semantics and logic, and varying v-e- the scalp
in amplitude and latency, lead him to the conclusion that:

these articles (e.g., Thatcher, R. W. 1976) provide compelling

evidence that the featureL of the ERP are influenced dramatically
by subtle semantic relations such as whether stimili are synonymous
or antonymous, whether they are bilingual equivalents or not, or
whether they are logically true or not. The changing latency and/or
anatomical locus of the LPC differences, observed as these information
parameters shift within an overall match-mismatch paradigm of otherwise
conztant design, provide perhaps the most convincing evidence that these
ERP features relate to cognitive mental processes rather than the
nonspecific or physical factors in these experiments.

The CNV has also proved to he a fertile source of research efforts attempting to relate aspects of
its occurrence to stimulus attributes. The slow negative shift 3f brain potential occurring between a
warning and an action stimulus which is referred to as CNV has attracted the attention and interest of a
nimber of investigators, resulting in a body of literature summarized by Beck (1975) into groups of studies
which interpret the CNV as reflteting expectancy, motivation, conation, or attention. Beck asserta the
overall implications that (a) the CKV is not a single process, and (b) that its nature aod cerebral
topography are dependent upon the state of the organism and .he task imposed. He points out that CNV
magnitude has been seen to relate to the uncertainty, intensity, and amount of information in the actIon
stimulus, the interstinmulus interval, concentration, and anxiety.

Rosenweig and Leiman (1968) provide a useful smmary of the major research approaches to the study
of brain-behaviox relationships:

Interest in the relation between electrical indicants of neural

function and dimensions of behavior has taken several fcrms. First,
considerable research has been dirzcted toward the examination of the
neural representation of r•nsory inputs, p-:,iculary the relationship
between neural coding and generalizations derived from the data of
psychophysical experiments. A detailed specification of the transfer
operations at relay nuclei in particular sensory systems has been noted
by several investigators. Secord, various invectigators have sought to
relate electrical parameters of neural functloning to more elaborat'd
behavioral phenomena--dimensions such as attention and learning. This
type of experimental inquiry has b"en direct to anal,,es of mechanisms of
irtegration and to relating regional slectrophysiological differences to
particular attributes ot behavior. Uzpsrlmental stratagims have been
diverse, although most studies can be considered as either (a) exhsuseive
analysis of the synaptic mechanisms of discrete functional systems which
may form a basis for the analysis of ,iore complex operations mediated by
these systems; or (b) determination oY the behavioral. correlcesa of
particular intrinsic wave processes, e.g., alpha rhythm or hippocamspl
theta rhythm; or (c) assessments of the induced slectrographic states
produced by special behavioral mu|,ipulations e.g., evoked activ!ty and
reaction time. (pp. 69-70)

EF. has several attsibutes which result in its popularity for use in situations where monitoring is
required; it occurs continuously and spontaneously, and it is readily available via relstively inexpensive
and simple, instrummntstton. Spontaneous, ongoing EEC can be automatically recorded and analysed Ptth
relative ease, and has served as an Indicator of many disparate varieties of states ranging from the
clincial (e.g., death, cchisophrenia, anxiety) to pertcroance (e.g,, arousal, perception, processing
efficiency). As Mirtk:' (1969) has pointed out:

. . . large masses of data may be processed quickly; the data go obtained

may be more easily subjected to statistical evaluation. Methods range
from total integrated energy (more properly, voltage) through base-line
crossing, frequency analysis, and analysis of slope chauges into
amplitude measures. (p. 325)
96

Evoked electrical potentials differ from ongoing EEC by occurrir.g in close temporal proximity to the
stimuli by which they are elicited, by their relative consistency of shape, and (usually) by much smaller
amplitudes than background brain wave activity.

Sen-Jacobsen (1971) has monitored EEC (among other physiological indices) from pilots under stress
in real-life missions, and has shown that previous quality of pilot performance can be used to predict
occurrence of delta-theta (2 - 8 hA) activity under stress. Sem-Jaccbsen attributes many aircraft
accidents io pilot task overload which elicits a freezing under conditions of extreme stress, reflected
in a showing and eventual flattening of brain wave activity.

A POSSIBLE SYNTHESIS

A recent paper by Cooper at al (1977) presents an interesting demonstration of cortical potentials,

large enough to stand out from background EEC and be recognized easily without averaging, which occur as
a result of changes in a visual display consisting of a video recording of a model landscape, comprising
an aerial view of countryside with residential buiidings, fields, hedgerows and roads in the foreground,
and with mountains and sky in the distant background. Various vehicles (vans, trucks, or cars) were
introduced and moved through the scene at varuous infrequent, unpredictable times. The task of the
subje'-tq was to (,i)detect and then (b) identify by type, each vehicle which entered the display screen.
Typically, abouL a second before the detection renponse was made, S's eye movements indicated a sudden
shift to the interesting for the present consideration) "a large, discrete, well-defined, positive-going
potential (occurred) in the EEC." This brain the range of this latency varied from 0.1 to 2.0 seconds,
which rules out a simple analogy to the P300 paradigm with detection the eliciting stimulus. There was
no such wave, incidentally, associated with the other half of S's task, the identification of the type of
vehicle.

As the authorp poInt out, since

it appears that the positivity occurs when the observer sees one
of the ,alass of events that he has been told to detect.

Both the role of thece potentials, ti.-ir distribution, and their

positive polarity suggest that they might have coon origins with
the P300 component of the cortical evoked potential which occurs
characteristically during discrimation and decision-making tasks.

A less well-defined brain event observed in this experiment was

a slow increase of negativity that started before the detection

positivity . . . (which) can start to increase while the eyes
are scanning other parts of the display, and it decreases after
caompleLion of the motor tasks indicating detection and recognition.

Although the possible relationship of this observation to the C(V previously described seems a tempting
area for speculation, the authors refrain from its pursuit. It is, however, a most appropriate area for
research, and An elucidation of the relationship of these phenomena with P300 and CNV right open the door
to the application of the rathor considerabln body of knrowledge which has already been gathered about
these brain wae~ evens in laboratory, settings to situations much mcre veridical than the typical lrboratory
problem now utilized.

THE WORK,,AD PARADIGU

Perhaps wore progress has been made toward the 'tilization of b-ain wave information for the enhance-
aunt oi pilot perforwa-ce in the area of monftorini; and a'seaesient of worklowd than in any other area.
This concept has been pu-suea effect-ivcly by Dotchin and his cnilsajuss (Wichans, Israel, and Donchin 1977,
e.g.), in the context of a large-scelt effort which is cento-sd ii Douchln's Cognitive Psychophyeiology
Laboratory at the University of Illinois and is a& ned at the devalc4mont of very closely coupled man.
computer syatems. Wickens (1978) provides a brief but thorough and clear dec.cviptiun of the purposes of
the workload measure and several hints as to the potentitl operational signIticance of information deri,,ed
therefrom-

in the op line aleptivQ context, V. may be assert.ed that atiintelligent

computer-based adaptive eystem, in order to optimally deploy the resovcu's
of huran and computer, should be peovided with a real-time, updatable
estimate ur model ,if the state (availability aid allocation) of the operator's
attentional rasourcen, so that adaptive procedures may be initiated. The
characteristics of Zhis stats estimate and their potential ase to the adaptive
decision maker are as follots.
(1) The available resources must be sufficimnt to meet the demands
imposed by all taska which challenge the operator at any time; the
characteristic cP task workload or reserve capacity. If the momentary
workload demands become sufficiently grest, adaptive siding procedures
can be implemented to temporarily unburden the human operator. These
may take the form of initiating alternate strategies of computer
processing or information display, implementing automatic control systems,
or of calling for extra manual assistance where such is available.

(2). Even if Zhe resources are adequate, the attention must be

allocated propeily to the critical tasks, display or sources of
information, so that important sources are not ignored: the
charucteristic of attention allocation. The dist~nction between
 S~97

workload and allocation is crucial. It is self-evident that adequate

capacity inadequately deployed may lead to non-optimal performance.
Thus if resources are allocated incorrectly along channels of incoming
information or, if critical sources of information are being ignored,
warning or cueing signals might be provided to redirect the proper
distribution of attention as determined by some preset establishment
of priorities. Alternatively, system characteristics might be alerted
with derived knowledge of the distribution of operator attention. For
example, computer resources might be allocated to monitor for, or
process signals along, information channels that are inferred to be
non-attended.

The research basic paradigm which is at present envisioned for the evaluation of workload measures
and their incorporation into a computer-aided aircraft control system involves an operator working on one
or more tasks which can be varied in complexity and difficulty. A flash of light, or similarly transient
auditory stimulus, is introduced; this stimulus muot be counted or otherwise used in the performance of
a secondary (or teritary) task, and the latencies and am)litudes of the p000's elicited by these so-called
probes (i.e., tests of attention, reserve processing capacity, etc.) are ubzd as indicators of operator
workload; the controlling computer then allocates task responsibility between man and machine primarily
on the basis of this information, thereby opti.mizing overall system performance. Too little work for the
operator will result in boredom and performance deterioration, so it may be w•se to require of the man
some work which theoretically is most efficiently and competently undertaken by machine. Too much work
for the operator, particularly under especially stressful conditions (e.g., landing in bad weather, combat,
maneuver in crowded airspace) may result in dangerous overload and, again, deterioration of performance;
so too there are times when a function which is theoretically best handled by a human should (or even
must) be assigned to the machine. The overall guiding principal is, obviously, to use knowledge ot the
operator's reserve processing capacity and level of performance in the light of current task demands.

Wickens (1978) asks,

"what represents an appropriate control, non-adaptive system against

which to compare adaptive system performance? Should this control be one
that functions with maximum aiding or minimum, or an intermediate level?
Should it perform under best-case or worse-case environments? The latter
refer to those conditions that are not typical, of real world operations,
occurring only very infrequently, but with potentially disastrous
consequences? What performance index should be used to evaluate and
compare such systems?

As a specific hypothetical example, consider t'ie following: An

adaptive multi-loop control system is de'eloped which will shift a&L
axis to automatic autopilot control, when concurrent task workload is
Inferred to excede some criterion. A tracking performance index of
this system cau clearly be measured. However, against what should
it be compireo? Tna performance of a maximum-aiding (all autopilot)
system--to which it will almost inevitably be inferior, or to that of
a minimum aiding (all manual) system to which it will probably be
4
super .or? To complicate matters, this superiority relationship might
bierev,=rced by Introducing "worse case" environments: e.g., levels
of disturbance which the autopilot is incapable of handling, or auto-
pilot system failures which, it has been argued, are less easily de~acted
under nutopilot that 1i-the.,loop dituaticns (Young 1969). Under these
circumstances, what arbitrary costs Phould be assigned to latencies in
failure detection, as opposed to errors in tracking performance, when
deriving an ultimate performance index? These again, are questions whose
answers may lie beyond the scope of the conference, but which inevitably
arise when adaptive uses of workload measures are considered.

Some of the experiments which have provided the basis for the development of his paradigm, and which
for the most 2art emanate from the Cognitive Psychophysiology Laboratory at the University of Illinoi3,
will now be briefly describtd. As Donchin (1976) has stated in general,

specific components of the ERP's have been shown to be manifestations

of such cognitive events as the preparation tG respond, the preparation
to intake and -Orocess information, the registration of a surrrri-ing
event, or the procescing of task-relevant information. Our studies
during the past two years have indicated that the ERP is a particularly
valuable source of information about the operator, information which
either is not otherwise available, or which can only be obtained
throegh poeatxally disruptive direct questioning. Briefly stated, the
r kRP appears to be a sensitive index of the strategies adopted by
individuals to cope with their assigned tasks, the manner in which they
allocate the±r resources and distribite their information processing
capacicies.

EXPECTANCIES

A good example of a large group of experiments on the way in which S'3 mental set or expectations
match the information conveyed by the stimulus is Squires et al (1976), which demonstrates changes in
P3U0 amplitude and conformation changa "as a function of the prior probability of the stimulus and the
specific sequence of the preceding stimuli." In other words, for sviy given L.evel of frequency of an
eve-at (the intrinsic probability of the event) the suprisingness of any single orcurrence of it is
98

significantly affected by the value of its imediate predecessors. Under conditions of relatively low
workload a previous sequence of stimuli up to about five is taken into account; when the workload is
high, the number of previous scimuli which are apparently taken into account by S decreases. This slope,
which can be reliably shown to be a function of task demands, is interpreted as an index of the reserve
processing capacity of S's short-term memory buffer, and therefore is considered a useful means for the
assessment of workload (Duncan-johnson and Donchin 1977). It is interesting to not- that this effect has
been demonstrated experimentally to hold over a range of intrinsic probabilities of .10 to .90 (Duncan-
Johnson and Donchin 1977), that the interval between probes i- not of critical importance (Donchin,
McCarthy and Kutas 1977), and that there are some modality differences (auditory 's visual) in (1) the
way workload afiects changes in P300 amplitude and (2) in the operation of the sequential effect model
described above. Opecifically non-target visual stimuli (i.e., visual stimuli to which S need not respond)
do not, apparently !ollow the model for P300 amplitude (Squires et al 1977). Incidentally, consideration
of the use of the seo1uence effect described ibove should be tempered by the realization that this effect
disappears when the S knows the intrinsic probabilities in the situation. Perhaps an implication of this,
and it might be a useful one, would be that in a situation of known probability parameters, a rare and
important stimulus which occurred several times in a row would not rapidly become habituated. The
boundaries of this situation, i.e., the length of the epoch (in which the postulated rare stimulus occurred
frequently) which would be necessary before S decided that this event had now become frequent, at least
temporarily (as opposed to perceiving, simply an unusual frequency of a truly rare stimulus), might be
interesting as an experiment. How rapidly does a rare or unusual stimulus habituate and what factors
affect the rate?

In gene--l, the P300 relates to a stimulus through changes in latency-which reflect the speed with
which the stimulus is recognized-and amplitude, which reflect the informational value and task relevance
of the stimulus. Informational value, or feedback, is usually operationally represented in these
experiments by stimulus rarity; task relevance, through instructional set. In general, greater task
relevance increases amplitude and greater attentional demands (e.g., greater display complexity) increase
latency. The stimulus must have at least some task relevance in order to elicit a P300 at all. If it is
an unusual or rare %low probability) stimulus, and therefore a surprising stimulus (the degree of which
may be affected by preceding stimulus patterns, knowledge of intrinsic probabilities, and, no doubt,
other circumstances), it will elicit a relatively large P300.

intelligence, motivation, e.g. - any or all) which are to be allocated among various demands at any given
time. A computer-based performance enhancing system should monitor the current resource allocation, the
additional leftover utilizable capacity, and be able to program optimal sharing of task responsibility
between itself and the human operator. Usually the primary task involves tracking, and its difficulty
can be readily manipulated (e.g., by varying the number of dimensions); the secondary task involves,
usually, counting the less frequent member of a pair of tones which vary in pitch.

Experiments utilizing this paradigm "indicate that while first order ERP's are relatively insensitive
to momentary fluctuations in tracking difficulty, they clearly discriminate between low levels of tracking
demand (no tracking vs. one dimensional tracking). Higher levels of ýIemand (one vs. two dimensions) are
differentiated by the extent of sequential processing of the stimulus series, a measure that is similarly
revealed in the ERP's." (Donchin 1976). Taken together, these data provide the basis for a tentative
measure of workload (and, moe importantly, the obverse: reserve resource capacity) based upon amplitude
of P300. This model has been useful, for example, in experiments (1) showing that a secondary task
requiring an occasional button-press interferes with performance on a (difficult) tracking task and (2)
studying the effect when a third task is added to ongoing primary and secondary tesks. In the former
c(qe, if whatever action is taken as a result of the button-push is taken instead when P300 occurs,
hypothesizes Donchin (1976), performance on the tracking task would show less interference effect. In
the latter case, the slope of the sequential effect (i.e., the number o4 pr,.vious stimuli which affect
P300 amplitude to the auditory probes comprising the secondary task) determine the likelihood that the
third task can be introduced without deterioration of performance on tasks I and II. It should be noted
that performance on tasks I and II are not enhanced in this model (though it is likely that intormetion
derived therefrom could be used to optimize all-task performance by changing parameters when P300
indicates low processing capacity); it is the overall system effectiveness which benefits. In other words,
the third task could be introduced only Tlhen the operator is cognitively ready and able to handle it.
Another experimental paradigm of interest is the comparison of (1) thi relationship of P300 latency
to reaction time in speed-demand situations with (2) this relationship under an accuracy-demand instruc-
tional set. Under speed conditions response generally precedes P300; under accuracy conditions, reaction
follows P300. Experimental results (Kutas, McCarthy and Donchin 1977) show that there is a high probabil-
ity that if an error has been made under speed conditions, reaction time is less P300 latency in a paradigm
where S is required to count the rare stimuli, and that making use of this knowledge can enhance perfor-
mance under the speed condition to the level attained under the accuracy condition - at no sacrifice in
speed.

Thze are other experimental methodological ingenuiries which appear interestingly relevant. The
Cooper et al study which seemo to elicit brain events which look like P300 in quasi-veridical settings
has already been discussed, and would appear to offer the means to take into account simultaneouely
infornation to be gained from on-going monitoring of background EEG with naturally-occurring ERP's.
Another idea which may be promising is Regan's (1977) work on steady-state ERP's. Here a transient change
of intensity or some other Important parameter of a sensory stimulus is repeated to elicit a series of
ERP's which, through Fourier transform analysis, can be used to monitor cognitive response to the stimulus.
Finally, the finding by Kutas (dissertation in progress) that hemispheric CNV amplitude indicates intended
choice of hand with which a response is to be made (in a left-right discrimination task) might be
eventually put to good operational purpose.

PROBLEMS AND APPLICATIONS

"Itmay be useful at this point to group the sorts of pilot problems which may lend themselves to
 99

brain wave enhancement, and to speculate briefly as to some of the ways in which the experimental findings
'described above may, eventually, apply.

WORKLOAD ALLOCATION

Workload allocation rnflecting s•ýcurate continuous assessmenL of operator state and processing
capacity probably would function for pilot performance enhancement most frequently in normal flight, to
provide the earlier mentioned theorottically optimum mix between h'ran and machine control of aircraft
function. Donchin (!976) describes an intended experiment which All test the workload allocation model
described previously and which could serve as a prototype for an aleraticnal aircraft operation:

The function of an autopilot is to replace the huiwn as an active element

in the flight control loop. The reduction in pilot work.-load gained
when the aircraft functions in an autopilot mode is however purchased at
a cost. The cost is represented by the increasing likelihood of a
malfunction in the control system with the greate'r number of mechanical
elements In the loop, the loss of flexibility o' the system in responding
to unexpected enviroimental inputs, and the deciease in speed and accuracy
with which the pilot can resume command If such events occur (e.g., the
appearance of an input requiring an evasive zaneuver or implementation of
some other control strategy).

Because there is a cost associated with autopilot control based upon

actuarial data of component reliability and on the likelihood of external
event occurrence, such control should not be in effect al all times, but
only when the external (non-control) demanrs of the operator reach a
sufficiently high level that the cost of remaining in the loop, as
determined by excessive work-load, exceeds that of autopilot function.
This experiment will explore the implementation of an ERP-based adaptive
elgorithm that will shift control from human to autopilot, according
to inferences based upon total processing demand and the distribution of
processing resources.

Subjects will fly a predetermined simulator flight path, perturbed by

an intermediate level of turbulence. External demands will be imposed
in the form of a discrete decision making task, stimulating target
identification and classification. Subjects will be instructed to
monitor a display for discrete items of alphanumeric information and
identify and classify this information into one of four categories
via a manual response.

The subject's cognitive state will be monitored by three kinds of ERP

inducing stimuli. The requirement to process a Bernoulli series will
generate a measure of overall work-load. ERPs elicited by the
appearance of the peripheral targets will be analvted to infer the
degree of processing of these targets. Finally, probe stimuli delivered
along the channels conveying flight path information will elicit ERPs
that will be employed to make inferences concerning processing or
neglect of the axes of tracking control.

An adaptive algorithm will monitor the ERPs and make the decision to
implement autopilot aiding according to the foloiwing decision ri.ie.
If work-load is Inferred to be high and and tracking axes uegLu:eA. thc
autopilot will be implemented. Otherwise the pilot will remain .n f;h
control loop. Once the pilot is out of the loop, the decision to
de-activate the autopilot will be made if the level of work-ioad ard
peripheral target frequency both drop below predetermined critets,.
Performance of the adaptive system will be based upon a joint measure
of target identification performance (speed and accuracy), and tracking
error, integrated over both the manually controlled and autopilot
deviations. From this index will be subtracted a fixed cost per minute
of the time spent in the autopilot mode. This performance index will
be comared with that achieved in a regular non-adaptive session of the
same task in which naturally the autopilot cost term will be zero.

Computer-controlled workload allocation could also function in an important beneficial manner under
acutely derinding and stressful conditions, applying the same basic algirithm in a slightly different way.

WARNINGS

The use of brain waves for the automated enhancement of warning, effectiveness could occur in two
ways. A computer could sense some deficit in an operator's state of being, or potential deficit
(anticipating a crisis), or the computer could observe, perhaps, a lack of attention to a wnrning display
or other performance (as opposed to state) deficit, and take action to further stimulate the human
oper3tor. In the former general Arcumstance the system monitor would make inferences about Operator
state, probably from a set of physiological information channels; in the latter, it would make inferences
about observed deficits in operator performance, probably from assessment of ERP's or their lack, in
respon, to warning signal displays treated functionally for this purpose as probes. The potential for
use of brain wave indicators of dangerous operator state or behavior is doubtless apparent: presence
of theta can predict drowsiness and deteriozation oi performance, and of course the sleeping state can
be readily detected. The detection of undesirable levels of arousal (inappropriately high or low) or
 100

emotional states can probably be enhanced through other physiological or behavioral channels. Atten'.ion
to a display could he assessed via Regan's (1977) steady-state ERP technique mentioned earlier. Donchin
(1976) holds out the promise of use of ERP's to distinguish non-response to a warning resulting from a
purposeful decision to ignore it, from accidental non-recognition; this way the computer-based system can
refrain from repetition or intensification of info .. ion which the operator has already processed and to
whirh he presumably will respond.

SENSING EVENTS AND SAVING TIME

The computer could also certairly determine the occurrence of an event like target acquisition (the
Cooper et al paradigm, e.g.) from occulography, pupillometry, and brain waves more rapidly and less
disruptively than this information can be made known by a human obrerver employing a gross skeletal
response such as pushing a button. In some cases it may even provide a more accuratc judgement of the
task-relevant event than could the button push (i.e., the p300 latency - reaction time relationship under
speed conditions described earlier). in any event, there are circumstances wherein a few milleseconds
could provide an important advantage. Use of information about the laterality of the readiness potential
to (a) infer the imminence of a motor response, (b) the hand with which it will be made, and (c) to
execute the implied command (e.g., fire a weapon, change course and/or speed, transmit a message) or
(d) to bring a control device to the desired hand (to avoid reaching), might significantly affect
performance efficiency - especially if the small single savings in time and effort were to be accumulated
over rapidly successive events in a continuing, recycling context of swift decision making and response.

FARTHER ALONG

At some point in the development of more highly interactive, closely-coupled man-machine interfaces,
a serious effort should be made to develop the capacity for real time thought commands. In this mode
the computer would sense specific wishes and needs (and evaluations of the adeqoacy of its own moment-to-
moment performance in meeting these needs) on the part of thn .pirator. Ultimately the ability to infer,
accurately, sequential chunks of complex iuformation would be nedded, u'-ilizing electrical representations
of verbal or non-verbal cognitive activity. Chapman's efforts (1977) to locate ERP's related to specific
words in multi-dimensional semantic space via the semantic differential technique may be a promising
step toward this end.

A more proximal goal would be the development of machine ability to sense such general intangibles
as operator uncertainty (and therefore the need for more information, or at least a need to maintain
decision options), and approval or disapproval. Sensing approval or disapproval (for want of a better
descriptor) would provide instantaneous qualitative feedback to the machine somewhat in the way "warm-
cold" feedback is provided to the blind searcher in children's games - or, perhaps more appropriately,
in the way varying intensities of temperature guide a missile toward a heat source. Ability to assess
these variables continuously and sensitively could provide the basis for very fine control of machine
by man, perhaps even allowing the creation of an artifically intelligent servomechanism so closely
responsive in real time to its operator's cognitions that it could serve virtually as a functional
extension of the operator's own nervous system.

The ultfmate aim for this type of man-machine system in general, and a goal at least as well suited
to enhancement of aircraft pilot performance as to any other military application, is the utilization
of the human operator for those purposes for which he is uniquely qualified: as a complex pattern
recognizer and decision maker - the pure strategist or tactician. The computer would, in real time and
functionally as if an organic part of the operator, undertake such activities as storing, organizing,
and retrieving data base information as it is acquired or needed, performing other data handling functions,
and carrying out decisions once it can accurately determine that they have been made.

CURRENT PROBLEMS AND NEEDED RESEARCH

Aside from the obvious, which is to increase the certainty with which inferences from P300 -an be
made and to refine the methodology of making use of them, it is possible to describe several areas where
current research techniques need to be made more powerful, and new methods which have been identified
and require further development.
Virtually total reliance on P300 for access to cognitive events is too limiting, and there are
several ways in which more brain information might be made available. One line of attack into this
problem area would be to !ek to understand the events underlying other components. Vidal (1977), for
example, has made potentially interesting use of some of the early exogenous components to guide a cursor.
Also, there is some indication that P300 bandwidth might be increased by independent probing of isolated
sectors of each retinal field. Further, Donchin (1976) has suggested some conditions under which N100
(an exogenous component) and N200 might yield operationally useful information on attentional, perceptual,
and processing events. Another line of investigation would be to open up new sources of information.
For example, using multiple arrays of eleýtrode3, latency, and amplitude differences arising from different
sites mlght reflect distinct cognitive activities. Further, regional variations in latency and amplitude
of the same component might be related in stable ways to various cognitive activities. Development of
functions representing, say, ratios of P300 amplitudes at various locations combined with zoncurrent sets
of latency difterences, or other secondary treatments of multiple recordings of the some event, might
yield fine discriminations among processing stages or other relevant aspeuts. Also, the development of
•:• magnetoencephalography
impinging sources). would allow access to subsurface activity (as well as providing physically non-

Even better and more reliable single-trial identification of brain wave events of interest is needed.
~
6'-While the development of Donchin and his colleagues (e.g., Squire & Donchin 1976) of such a capability,
using a sliding template and stepwise discrimination analysis has made feasible the real time use of ERP's
for vehicle control, much more needs to be known about such issues as the sLability of individual templates
for recognition of ERP's, optimum strategies for updating these templates, and, of course, increasing the
accuracy with which such recognitions are made. Present capabilil.ies seem remarkably good, but if life-
 i I01

or-death actions are to be taken on the basis of them, either accuracy must be improved or a fail-safe
procedures developed. For allocation of workload under routine or even somewhat demanding conditionhs
present levels of accuracy seeu adequate.

REFERENCES

Beck, 2. C., "Electrophysiology and Behavior." Ann. Rev. Psychol. 1975, 26. Palo Alto: Annual Reviews Inc.

Chapman, R. M., Blsgden, H. R., Chapman, J. A., and McCrary, J. W., "Semantic Meaning of Words and
Ave.a•.d Evoked Potentials." In: LANGUAGES AND HEMISPHERIC SPECIALIZATION IN hkN: CEREBRAL ERP's.
Prog. •lin Newrophysio., 3, (Ed.) J. E. Desmedt (Karger:Basel, 1977)

Davies, Dr. and Parasuraman, R., "Cortical Evoked Potential and Vigilance: A Decision Theory Analysis."
In VIGILANCE: THEORY, OPERATIONAL PERFORMANCE AND PHYSIOLOGICAL CORRELATES. Mackie, R. R. (Ed.) New York:
Plenum 1977.

Dimond, S., "Vigilance and Split-Brain Research." In VIGILANCE: THEORY, OPERATIONAL PERFORMANCE, AND
PHYSICLOGICAL CORRELATES. Mackie, R. R. (Ed.) New York: Plenum 1"77.

Donchin, 1. 1976. Personal Cnmuunication

Donchin, E., McCarthy, G., and Kutas, M., "Electroencephalographic Inves "gations of Hemispheric
Specialization." In: LANCUAGIA AND HEMISPHERIC SPECIALIZATION IN MAN: CEREBRAL ERP's. Prog. Clin.
Neurophysio., 3, .Ed.) J. E. Deamedt (Karger: Basel 1977).

Donchin, E., Ritter, W., and McCallum, W. C., "Cognitive Psychophysiology: The Endogenous Components of
-the ERP." In CALLAWAY, K., Teunting, P., and Koslow, S. (Eds.) In press (Academic Press) 1978.

Duncan-Johnson, C. and Donchin, E., "On Quantifying Surprise: The Variation of Event-Related Potentials
with Subjective Probability." Psychophysiology, 1977, 14, 456-467.

Gale, A., "Some EEG Correlates of Sustained Attention." IN VIGILANCE: THEORtY, OPERATIONAL PERFORMANCE,
AND PHYSIOLOGICAL CORRELATES. Mackie, R. R. (Ed.) New York: Plenum 1977.

Jersion, S., "Vigilance: Biology, Psychology, Theory and Practice." In VIGILANC•E: THEORY, OPERATIONAL
PERFORMANCE, AKD PHYSIOLOGICAL CORRELATES. Mackie, R. R. (Ed.) New York: rlr nnum 1977.
John, E. R. and Schwartz, E. L., "The Neurophysiclogy of Information Processi.nq and Cognition." An.
Rev. Psychol. 1978, 29, Palo Alto: Annual Revies Inc.

lutes, M. and Donchin, E., "Studies of Squeezing: Handedncss, Responding Hand, Responded Force, and
Asymuetry of Readiness Potential." Science, 1974, 186, 545-548.

Kutas, M., McCarthy, C., and Doiiumn, E., "Augner.ting Mental Chronometry: The P300 as a Measure %f
Stimulus Evaluation Time," Science, 1977, 197, 792-795.

Mirsky, A. F., "Neuropsychological Bases of Schizophrenia." Ann. Rev. Psychol. 1969, 19. Palo Alto:
Annual Reviews Inc.

O'Hanlon, J. F. and Beatty, J., "Concurrence of EEG and Performance Changes during a Simulated Radar
Watch aa-d Some Implications for the Arousal Theory of Vigilance." In VIGILANCE: THEORY, OPERATIONAL
PERFORMANCE, AND PHYSIOLOGICAL CORRELATES. Mackle, R. R. (Ed.) New York: Plenum 1977.
Rosenzweig, W. R. and Leiman, A. A., "Brain Functions." Ann. Rev. Psychol. 1968, 18. Palo Alto:
Annual Reviews Inc.

Regan, D., "Steady-State Evoked Potentials." .1.Opt. Soc. Am., 1977, 67, 1475-1489.

Sea-Jacobsen, C. W., "Physiological Aspects of Aircraft Aczideut Investigation.'" Aerospace Med., 3.971,
4_2, 199-204.

Squires, K. C. and Donchin, E., "Beyond Averaging. The Use of Discriminant Functions to Recognize Event
Related Potentials Elicited by Singie Auditcry Stimuli." Elect. & Clin. Neurophysiol. 1976, 41, 449-4i9.

Squires, N. K., Donchin, E., Squires, K. C., and Grossberg, S., eisensory Stimulation: Inferring
Decision-related Processes from the ?300 Component." J. Exp. Psycbel. Hum. Pert. and Perf. 1977, 3,
299-315.
Squires, K. C., Wickens, C., Squires, N. C., and Donchin, E., "The Effect of Stimulus Sequence of the
Waveform of the Cortical Event-related Potential." Science, 1976, 193. 1142-117T6.

Wickens, C. D., Position Paper, NATO Conference on Workload. In press, 1978.

Wickens, C. D., !sreal, J., and Donchin, E., "The Event Related Cortical Potential as an Index of Task
Workload." Proc: 1977 Ann. Mtng., Hum. Fec. Soc.

Vidal, J., "Real-Time Detection of Brain Events in EEG." Proc. of the I.E.E.E. 1977, 65, 633-641.
 $a

PON.!ONT

BI

INT6RNALBASLN

BIOLIFFCGLTCAL
AEqFFCULT FLG

1 2 13AELN
li BIS

w I
2.
Figur rae fefctv irai euto
EASYe~ nStibr a
 103
PUPILLONETRIC AMIIODS OF WORKLOAD gVALUATION:
PRESENT STATUS AND 1fUTURE POSSIPILITIES

by

Jackson Beatty, Ph.D.

Departmant of Psychology
University of Califonis at Lo" Angeles
Los Angeles, California 90024

INTRODUCTION

The assessment of :ilot workload is a special case of the measurement of Information-processing load,
the aggregated demands rlaced upon an individual in the performance of a particular cognitive task or
function. Three general approaches have been employed in the measurement of Information-processing load.
The first is that of subjective estimation. Subjective estimates are involved when workload is estimated
from the task engineer's opinion as to the prouable magpitude of processing load, an opinion that may be
based on previous experionce or an analytic theory. However, subjective estimates of workload by the user
or participar.. are the most common form of workload measurement in aircraft design. Both types of
subjective ratings have srious Weknesses.
The second major method of measuring processing load employs behavioral measurement. Here the notion
is thct the information-processing capacity of an individesl is limited so that the workload imposed by
one task can be estimated by the degree co which it interferes with the simultaneous execution nf a
secondary rensurement task, such as simple reaction time or manual tracking. This method has much to
recommend it over the subjective measurement techniques, particularly with respect to objectivity. But the
behaiioral-interference method is difficult and time-consuaing to implement, and yields relatively little
data for the amount of time and energy invested in testing. As a consequence, this method has been of
more theoretical than applied interest.

The third major method is physiological, in which the response of the nervous systew to the load
imposed by an information-processing task is assessed. Momentary increases in processing load induce
short-latency, short-lived increpses in measures of central nervous system activation. These changes are
moat evident and most easily measured in the autonomic nervous system. Arnong the autonomic measures of
activation, changes in pupillary diameter appear to be the most sensitive and accurate (Kahneman, Tursky,
Shapiro & Crider, 1969).

This paper discusses the use of pupillometric measures tn the evaluation )f pilot workload. I begin
by describing the innervation of the pupil with respect co its connections with bratnstem activation
systems. Modern methods for pupillmetric measurement are than described. Next, a series of experiments
describing pupillary response in a variety of information-processing tasks is reviewed. Finally soot
possibilities for the use of pupillometric methods in the measurement of pilot workload are dinteisped.

INNERVATION OF THE PUPIL

Pupil1ary diameter is determined by the relative stats of contraction of the two .1prosing 2uscle
groups of the irls, the sphincter and the dilator pupillae. The dilator pupillee are radially oriented
bands of smooth muscle that are innervated by the sympathetic branch of the autonomic nervous syt4tem
through the cervical sympathetic ganglia. The sphincter pupillae are inncr-7ated by the parasympathetic
system through the ciliary ganglia, and act te close .he pupil when activated. Pupillary dilaticn,
therefore, can result from either sympathetic activation or paraeyvpathetic Inhibltion. Cortical Inhibi-
tion of the Edingcr-Westphal nucleus, thu brainstem nucleus that projects to the ciliary ganglia, has
been frequently hypothesized to accompany cortical activation. Both the sympathetic and parasampathtz.I
brainstem niclei involved tie regulation of the iris musculature are 'Intimately connected with the
reticular aztivating syaten. indeed, puplllnTr, meaaurel were used to &ssess reticular formation functions
in che ploxneeriug work of Moruzzi and 7illablancs (Moruzzi, 1972).

METHODS FOR PUPILLOHETRIC MEASUXYi47T

The pupillotetric measurement of information-processing load requires that accurate measures of

pupillary diameter to ohtiined during the course of an information-processing task. Early work in this
area employed photographic methods, in whic': infrared photographs of the pupil were obtained 3t the rate
of 1 or 2 per s9cond for the duration of the processing taek. Pupillary diameter vas determined by
subsequent direct measurements take, from the enlarged image ot each frame. Although accurate, the
photographic method suffers from two veaknesses: it is a laborious and frustrating procedure to implement
and, for this reason, it is tnot practical to obtain fine temporal resolution because of the resulting
proliferation of photographs.

The second principal method of pupillomttric measurement involves the use of a high-resolution infrared
video camera and a special-purpose image processor that extracts an estimate of pupillary diameter from
each frame of the video Image. Originally developed under a grant from NIH, this instrument is presently
manufactured by Gulf and Western Applied Science Laboratories, formerly the Whittaker Corporation. All
major laboratories now involved in pupillom-trtc research use this •nstrument.

In the basic video scan pupillometer the subject's head is restrained by a chin and forehead support.
&,.infrared video camera is placed outside the subject's foveal 'leld of vision, as is an infra-red
slit-lamp illuminator. Both the illuminator and the camera are focused on one of the subject's eyes and
the resulting image is sent to the image processor for extraction of pupillary diameter. This basic
configuration of head rest, illuminator and casera is adequate for most experimental work. Pupillary
measurements noy be made over a wide range of lighting conditions, including complete darkness. The
subject is free to move his gaze over a limited portion of the visual fAeld; strict fixation is not
104

required. Positioning of the subject in the head support is quickly accomplished. With appropriate
adjustments this testing arrangement results in little subject fatigue.

For purposes requiring greater freedom of head movement than allowed under this configuration, a
head-tracking pupillometer may be used. This permits recarding of pupillary diameter from a seated
subject with complete freeder of head movement. In this device, two video cameras are employed. The
second camera is used to locate the head of the subject in three-dimensional space and by the use of
servo-eechanisms direct the primary camera to the subject's pupil in that space. Although rather
expensive, this head-tracking arrangement seems to perform quite reliably.

In the basic pupillometer, pupillary diameter is estimated from the video image of the eye by the
following method: Each raster line of the image is first scanned for sharp light/dark contrast points
that might signal the boundary between iris and pupil. The use of an infrared vidicon minimizes the
effects of iris coloration on the contrast of the iris-pupil boundary. A single control is provided for
the adjustment of the sensitivity of the contrast detection circuitry. Sensitivity is individually
adjusted for each subject but, once adjusted, remains stable over long periods.

The second stage of imaging processing is the search for a semicircle of contrast points which
together define the leading edge of the pupil. The diameter of this samicircle provides a reliable
estimate of pupillary diameter. This measure is recomputed 30 times each second and is available for
computer input in either analog or digital form.

The performance of the image processor may be evaluated by means of a video display of the processed
image. Contrast points are indicated on the monitor as brightness-intensified sparkles. The extracted
image of the pupil is visually indicated by a darkening of all raster lines passing through the detected
pupil. Thus, if the pupillometer is functioning properly, the monitor displays a video image of an eye,
with intensified points along the left iris-pupil boundary with a dark band tangent to the upper and
lower boundaries of the pupil. Measurement qunlity can be assured by visual monitoring of the processor's
display.

Recordins Constraints: Accurate recording is essential to pupillometric measurement of workload since

the pupillary dilations reflecting zhanges in central activation, although highly reliable and observable
on single trials, are nonetheless exceedingly small. For this reason, other factors which affect pupillary
diameter must be carefully controlled.

Chief among the non-cognitive determinants of pupillary diameter is the well-known light reflex,
which reduces pupillary diameter as integrated retinal illumination is increased. The light reflex is
very sensitive and the maximum amplitude of the response is several millimeters. For this reason the
luminance of the visual field must be constant during measurement. In out experiments on visual infor-
matinn processing, we employed a computer-controlled CRT display in which task-relevant stimuli were
presented for short (100-200 msec) periods. At all other times equiluminance random dot fields were
displayed. Such control of the light reflax may not be possible if the subject is required to sc.n a
complex visual field of varying luminance.

The momentary state of the occulomotor reflexes mediating convergence and accommodation also must be
controlled as vergeace movements and accommodation reflexively affect pupillary diameter. In our work
with visual displays, the critical visual stimulus was piaced several meters from the subject to relax
accommodation and minimize convergence. At one time we were troubled with significant constrictions
occurring in some subjects while viewing prolonged visual displays. We attributed these artifacts to
uncontrolled vergence/accummodative movements and altered our task to utilize the more artifact-free
brief presentations. Nonetheless, visual stimuli can be employed in pupillometric research, but a great
deal of care must be taken in Aealing with such materials.

Th-se problema do not exist when auditory displays are employed. For this reason, presentation of
information in the auditory mode is recommended whenever feasible.

Recording Artifacts: The video-scan pupillouet~r is one of the most accurate reliable, and trouble-free
phychophysiological recording devices ever developed. Nonetheless artifacts in the pupillometric record
do occur and must be dealt with before the data are anaiyzed.

The major sources of artifact are blinks and partial lid closures. In these cases, movements ot
the eyelid obscuro a portion of the pupil, resulting in -.neous measurement. Such artifacts are easily
observed in the pupillary record and are sufficiently obh-..s to permit automatic computer artifact
detection If desired. In our own work, the raw pupillary data from a entire experimental session is
csored on disk memory for latee visual examination. 6wall artifacts are corrected by linear interpolation
and data segments with large artifacts are discarded. This editing procedure is rapid and assures accurate
pupillomecric data.

Another major source of artifact lies in the contrast detection threshold established for certain
subjects. If the illuninator is improperly foct'sed, or if the subject has long drooping eyelashes, the
recognition of the upper pupil boundary may be uncertain. This results in a characteristic jitter in the
pupillary record. When this occurs the source of the difficulty should be corrected. Data segments
containing such jitter should be discarded.

PUPILLARY CHANGES IN HUMAN INFORMATIG PROCESSING

There is a large body of experimental evidence that suggests that pupillary dli.ations under controlled
conditions reflect with high accuracy the momentary level of load placed upon the human nervous system by
infnrmation processing tasks of varying difficulty, content and complexity. These data have been reviewed
recently by Goldwarer (1972) and Janisse (197(). In this section I shall present a seiies of experiments,
primarily from my own laboratory, that illustrate the sensitivity of pupillometric measures to momentary
changes in processing load which are in some degree relevant to the applied problem of pilot workload
evaluation.

Meor Load: One component of pilot workload in the demand placed upon short-term memory in verbal
communication with other aircraft or groun4 sites. Detail d verbal instructions for example, must be
accurately retained. The limitations of short-term memory are well known to psychologists and htman
factors engineers alike. Pupillonutric measures provide a means of quantitatively assessing the
physiological lrad placed upon an individual by verbal informantion of varying amounts and complexity
which is to be retained for short periods of time.

Kahnsman and Beatty (1966) presented the first pupillometric analysis of the processing demands
encountered in a short-term memory task. Figure I presents pupillometric records obtained during a
short-term memory task in which string@ of 3 to 7 digits were auditnrily presented at the rate of I per
sec. Two seconds after the last digit was heart, subjects were required to repeat the digit string at
the sama rate. It is apparent from Figure I that the momentary degree of pupillary dilation accurately
reflects the cognitive workload imposed by the short-term memory task. Pupillary diameter increases in
a linear fashion with the preasentation of each digit, reaching the maximum in the 2-sec pause preceding
report. As digits are unloaded from memory during report, pupillary diameter decreases with each digit
reported, reaching baseline levels after report of the final digit. In unpublished work, it was deter-
mined that if the subject were requested to repeat the string a second time Immediately after reporting
the final digit, the pupil immediately dilates to the peak diameter for that string and then decreases
with each digit spoken until the entire string has been reported for the second time. The magnitude of
the pupillary dilation at the pause between input and output in Figure I is an increasing function of
string length. Beatty and Kahneman (1966) demonstrated that a similar pupillary function is obtained
when a string of items is recalled from long-term memory for report: On request to report, a large
pupillary dilation is observed as inforration is retrieved from long-term memory (see Figure 2). As each
digit in the string is reported, pupillary diameter decreases, reaching baseline levels at report of the
last digit. Thus it appears that the limited capacity portion of the human information-processing system
may be loaded from either long term memory or environmental stimuli and that the pupillometrically mea-
sured workload is similar in both of these cavaes.
Memory load is also determined by the difficulty of the to-be-remembered information. Remembering
unrelated nouns requires more capacity than remembering a string of single digits of equal length, as
measured by the difference in memory span for the two types of items. Figure 3 shows the pupillosetric
data obtained for strings of four items of different :ypes. The smallest dilations are observed for
strings of four digits that were to be simply repeated. Larger dilations were apparent for the string
of four words, indicating that both item difficulty and number of items determine workload in the manory
task. The largest dilations were obtained for the subjectively most difficult task of transforming each
of the four digits by adding one before report. These data provide strong support for the idea that task-
induced pupillary dilations provide a physiological index of the momentary level of workload imposed by a
memory task.

This idea was subsequently confirmed in an experiment by Kahneman, Beatty, and Pollock C1967) in
which both pupillometric and behavioral interference methods were utilized to assess workload in the
four-digit add-one memory transformation task. Using a secondary task of visual target detection, it was
found that the behavioral estimate of workload and the pupillometric measure of physiological load were
in exact agreement. A series of controls ruled out any peripheral interference of the pupillgry dilations
thev. elves on performance of the secondary task. In comparing the two data sets, the pupillometric data
was .or more detailed than the behavioral data, required fewer trials to obtain, and was of considerably
lower variance.

Decision Procosses: Even simple decision processes appear to impose some workload on the cognitive system
as indicated by pupillometric measures of activation. For example, Simpson and Bale (1969) measured
pupillary diameter in two groups of subjects who were required to move a level to one of four positions.
In the decision group, subjects were told at the beginning of each trial that either of two directions
was permissible (e.g., front or left). Seven seconds later a response cue was presented and the subject
initiated one of the two movements. In the no-decision control group, subjects were instructed exactly
as to the desired movenent on each trial (e.g., front). Pupillary dilation in the post-instruction pre-
response period was larger and more prolonged for thos subjects who had to choose between two movements
bafore responding.

Substantially larger pupillary dilations are observed to accompany more difficult decision processes.
In an experiment reported by Kahneman and Beatty (1967), listeners were required to determine whether a
comparison tone was of higher or lower pitch than the standard. Clear pupillary dilation occurred in the
4-second decision period between presentation of the comparison tore and the response cue. The amplitude
of this dilation varied as a direct function of decision difficulty, the difference in frequency between
the standard (350 Hz) and comparison tones. This relation is shown in rigure 4, which presents both the
6mplirude of dilation in the decision period and the percent decision errors as a function of the fre-
quency of the comparison tone. These dilations were highly reliable and did noc habituate over the
experimental session. Pupillary dilations during decision appear to vary as a function of cognitive
workload, as inferred from task parameters and performance data.

Complex Reasoning: More complex cognitive functions not unexpectedly impose a major load upon the human
nervous system during their execution. This may be most easily observed in the laboratory using mental
arithmetic tasks. Such tasks may be regarded as directly analogous to other types of complex reasoning
tasks that may occur in man/machine interactions.

Pupillary dilations accom,,anying complex problem solving appear to be related directly to the
difficulty of such processing, although behavioral assessments of workload have not yet appeared for these
types of cognitive tasks. For example, Hess and Polt (1964) examined pupillary movements as multiplication
"problems were solved mentally. Pupillary diameter increased during the period preceding solution, and
106

related to presumed problem difficulty. Payne, Parry, and Harasymiw (1968) also report a monotomic
relation between mean pupillary diameter and problem difficulty, but note that this relationship is
markedly nonlinear with respect to difficulty scales based upon percent corrert solution, time to
solution or subjective rating of difficulty. Pupillary diameter in %qntal multiplication appears to peak
rapidly as a function of difficulty, with more difficult problems requiring more time until solution is
reached. This suggests that cognitive capacity ia quite fully taxed in complex mental arithmetic problems
so that the workload per unit time remains relatively constant as problem difficulty is increased over
moderate levels, but that the total time to solution is increased.

These investigations using the older photographic methods of pupillometric measurement were not able
to discern the fine temporal structure of complex reasoning tasks which is clearly evident when more
de iled video-scan pupillometry is employed. Ahern and Beatty (in preparation), as part of a study of
individual differences and cognitive load, presented subjects with multiplication problems at three levels
of difficulty. The problems were computer-controlled using acoustically-presented digitized speech
stimUli. These data are suarized in Figure 5. Clear dilations may be observed in all cases at the
presentation of the multiplicant (a single digit, a low two digit numoer or a high three digit number).
This dilation quickly subsides and the pupil returns towards basal levels until the multiplier is presented,
at which point a major dilation is observed. The duration of this dilation is related to problem difficulty
being more prolonged for more divvicult problear. These data suggest that pupillometric methods nu, only
may serve to measure the workload associated with a single task or function, but also to measure the time
course of that load with some degree of precision.

Other types of complex problem solving tasks show similar relationships between pupillary dilation and
problem difficulty. For example, Bradshaw (1968) has reported that larger pupillary dilations accompany
the solving of more difficult anagrams, and that these dilations are maintained until solution is reached.

Summary: Pupillometric measurements have now been obtained in a variety of simple information-processing
tasks under laboratory conditions. They appear uniquely sensitive to subtle differences in procesring
load obtained in these tasks. Processing load appears to increase Lhe activation of brainstem arousal
systems in measured amounts. These activation responses are of short duration, of an extent that
accurately reflects load, and occur at short latency. The responses do not habituate, and therfore may
be assumed to reflect a fundamental physiological response to increase in cognitive workload. As such,
they suggest an alternative to traditional methods of quantifying workload, a possibility that is
explored in the following section of this repor-.

FUTURE APPLICATIONS OF PHYSIOLOGICAL MEASURES TO THE ASSESSMENT OF PILOT WORKLOAD

No investigation has yet been published in which pupillometric methods have been employed in the
measurement of pilot workload. Perhaps the most direct application of these methods to practical perform-
ance assessment is Peavler's (1974) use of pupillometric measures to assess fatigue in telephone operators
after working full shifts on different types of computer-based information retrieval systems. Peavler
found that the more automated methou, which was both more efficient and more taxing, resulted in greater
operator fatigue, as indexed by mean decrease in pupillary diameter from prernak to posttask mcasurements.
Thus, Feavler was not concerned with the question of task-induced pupillary dilations and instantaneous
workload levels, the topic of the present report.

The body of research summarized above certainly makes a theoretical contribution to the study of
workload, suggesting that workload can be measured by a physiological reaposse to task load, rather than
by behavioral interference or subjective report. In my opinion, these methods may be of practical cones-
sequence as well.

The most natural application to the problem of pilot workload wiuld seem to be in the area of design
of equipment and pilot procedures, in which the workload parameters of each of several design options
might be assessed separately using experimental methods similar to those outlined above. Here, one might
ask questions concerning optimal information formatting to determine a communication structure that
minimizes operator load. The me'hod is particularly well suited for the design of the more cognitive com-
ponents of the pilot's task, analogous to the mental arithmetic experiments described above. It is
precisely this aspect that would seem to be most difficult to measure by conventional workload assessment
procedures.

One could conceivably construct a simulator in which pupillometric measurements might be made to tnst
workload in a more realistic envirormt L. However, in my opinion, the problems of adequate control of
visual input in such a situation would seriously impede its usefulness. As mentioned above, strict control
of visual input is necessary for the pupillometric measurement of workload as the large magnitude changes
in pupillary diameter that are produced during a visual scan of a non-nomogeneous visual field introduce
serious artifacts in the pupillometric record. Until such problems are solved, the use of pupillometry
in more natural environments will be restricted at best.

Finally, some attention should be paid to the use of other physiological measures su, A as the EEG
in the assessment of workload effects. An inspection of the current literature is not prumising in this
regard, as no large maguitude and robust relations between EEG and workload have been reported despite
a reasonable amount of experimental work devoted to this problem. The development of an EEG measure of
workload would be of some practical interest, as the EEG is not dependent on small changes in visual input
as is the pupil. The question of an EEG meassure of workload is presently being pursued in my laboratory
under ONR support. We are using the mental arithmetic and short-term memory tasks which have such strong
and reliable effects on autonoalc indicators of load, including the pupil. Pupillometric data are also
being analyzed. EEG data is being systematically recorded from each of the 19 sites in the Ten-Twenty
recording system (Jasper, 1958) and stored for subsequent analysis. Bv proceeding in a systematic manner
in the analysis of the EEG and 1,ycontinuing use of the pupillometric -"easures to assess the effectiveness
of the manipulations of processing load, we hope to finally discei, the central signs of processing load
which are so clearly observable in the autonomic periphery.

~--..
-------------------------------------
--
 107

i iways unsatistac.tory, workloadprovide

methods ofmethods
In wt 3a-y, physiological measurement, alternative
a uniquesubjective to the traditional,
estimatton and behavioral in various
but interference

with a secondary task. Of the physiological measures, the task-evoked pupillary responses provide the
clearest indication of both the degree of load imposed by a pmrticular task or function and the fluctu-
ations of that load over time. Although soue restrictions are necessary to insure accurate pupillometric
recordings, the use of pupillometric methods for workload assessmcnt would seem to be feasible, particu-
larly in evaluating the load imposed by complex cognitive tasks.

RIF•UCES

Ahern, S. K. and Besatty, J. Activation and intelligence: Pupillometric correlates of individual

differences in cognitive abilites. In preparation.

Goldwater, B. C. Psychological significance of pupillary movements, Psychological Bulletin, 1972, 77,

pp. 340-355.

Hess, E. H. and Polt, J. H. Pupil size in relation to mental activity during simple problem solving,
Science, 1964, 143, pp. 1190-1192.

Janisse, H. P. (Ed.) Pupillary dynamics and behavior, New York: Plenum, 1974.
Jasper, H. H. The ten-twenty electrode system of the International Federation, Electroencephalography
and Clinical Neurophysiology, 1958, 10, pp. 371-375.

Kahneman, D. and Beatty, J. Pupil diameter and load on memory, Science, 1966, 154, pp. 1583-1585.

Kahneman, D. and Beatty, J. Pupillary response in a pitch-discrimination task, "erception &

Psychophysics, 1967, 2, pp. 101-105.

Kahneman, D., Beatty, J., and Pollack, I. Perceptual deficit during a mental task, Science, 1967, 157,
pp. 218-219.

Kahne-an, D., Tursky, B., Shapiro, D. and Crider, A. Pupillary, heart rate and skin resistance change
duting a mental task, Jovrnal of Experimental Psychology, 1969, 79. pp. 164-167.

Horuzzi, G. The sleep-waking cycle, Reviews of Physiology: Biochemistry and Experimental Pharmacology,
New York: Sprlnger-Verlab, 1972.

PaI e, D. t. Parry, M. E. and HarasYmiw, S. J. Percentage pupillary dilation as a measure of item

di ficulty, Perception & Psychophysics, 1968, 4, pp. 139-143.

Peavler, W. S. Individual differences in pupil size and performance, In M. Janisse (ed.), Pupillaty
dynamics and Behavior, New York: Plenum, 1974.

Simpson, H. M. and Hale, S. J. Pupillary changes during a decision-making task, Perceptual and Motor
Skills, 1969, 29, pp. 495-498.

L 4.3 A.

4.2 7 DIGITS
6 DIGITS
4.1L
5 DIGITS
!5 4.0 Y---- 4 DIGITS
"...... 3 DIGITS
a..
E3.8
"3.7

S3.7

3.5-
PALUSL

TIME IN SECONDS

Figure 1. Average pupillary diametc: during preseitation and recall of strings of 3 to 7 digits,
superimposed about the two second pause between presentation and recall. Slashes indicate
the beginning and the end af the memory task.
108

4.6

cc 4.4 IIo LONG TERM

4.3 -,--o sHo-, TERM

MEMORY TASK

MEMORY TASK
" 4.2
4.1

><4.0 CN

CONTROL A EN ýU4 REPORT CN L

3.9 Lit . I _ ., , p oi
TIME (in seconds)

Figure 2. Average pupil diameter for five subjectu during presentation and report of

seven-digit telephone numbers from short-term and long-term memory. The

long-term memory function is broken above the brace, with both points
representing the same pupillary measurements.

4.3

4.2g

1- 4.1
-4.0
I, - TRANSPORTATION

,3.9 --. 4 DIGITS

MJ53.8 -
WW3.7 -.--

3.6

3.5 CONTROL LISTEN PAUSE REPORT CONTROL

TIME IN SECONDS

Figure 3. Pupillary diameter duiring presentation and recall of four digits,

words and a digit transformation task.

; -
 109

.20 125%

.. I 0 •I
dE 26%

z I

DO.05 I.,. IoII.

826 uSll 830 562b 8174

FREQUENCY (in cpu)

Figure 4.
05 , -b
Aver'age pupillary dilation during the decision perte~d and per-'ent errors
as 8a
function of the frequency of the comparison tone, The frequenc.y of the standard
was 850 cps.

00
FI
SMULTIPLICAND MULTIPLIER DIFFICULT

0 p I 6I'52 o

MEIU
CTiME (SECS)

Figure 4. Average d pupl lary responses AS

, ing a me multdplicad ion task ofrthrees
levels of proble. difficulty.

I.
 AIRCREW PERFORMANCE RESEARCH OPPORTUNITIES USING
THE AIR COMBAT MANEUVERING RANGE (ACMR) J,

by

Clyde A. Brictson, Ph.D.

Anthony P. Ciavarelli, M.A.
Dunlap and Associates, Inc.
Western Division
7765 Girard Ave., Suite 04
La Jolla, California 92037

ABSTRACT

Three years of aircrew performance measurement related to air combat effectiveness using the Navy's
Air Combat Maneuvering Range (ACMR) are presented as evidence of ACMR's research potential. Performance
assessment methods used to evaluate pilot proficiency are described. The aircrew assessment methods have
been used to identify squadron performance differences, evaluate competitive exercises, and provide
diagnostic training feedback to operational users. The use of continuously recorded quantitative measures
from systems such as ACMR should stimulate more aircrew performance field research ideas. The avilability
of objective performance criteria promises to be of substantial benefit to both the operational user and
the research community in such areas as pilot selection and training, fleet combat readiness, and pilot
workload and stress.

INTRODUCTION

Background: The selection, training and assessment of military aviators, and probler-s associated with
the acquisition and retention of flying skill, have occupied aviaticn psy.hologlscs foi over 30 years
(Thorrdike, 1974). The major probletw in this line of research has been, and continues to be, the lack
of objective criteria (North and Griffin, 1977) for evaluating the effectiveness of aviation training in
general, and aircrev proficiency in particular. Traditionally, the use of subjective estimates has
provided the only means to assess training progress in acquiring and maintaining aviation skills.

The recent growth of the Navy's Air Combat Maneuvering Range (ACMR) has provided a unique opportunity
to obtain objective measures of aircrew performance that have not been avilable in the past. For the past
three years the authors have been involved in a research program to develop objective aircrew performance
criteria from ACMR quantitative output measures. Two technical reports (Brictnon and Ciavarelli; 1976 and
1978) have been written which detail the technical approach, performance assessment methods, and prelimi-
nary results of aircrew performance measurement on selected training objectives. The ACMR criterion
development research is sponsored by the Navy Aerospace Medical Research Laboratory, Pensacola, Florida
in order to provide them with performance criteria to validate, among other things, vision laboratory
results, aircrew selection practices, and training effectiveness. The availability of such criteria,
however, has perhaps more far reaching Implications for the expanoion of aircrew research efforts in an
operational wtivironineni.

Air Combat Maneuvering Range (AC(R): The ACHR is a sophisticated training fecility acquired b- the Navy
and now In use to train fighter aircrows in air-to-iea combat. The system is designed to train dircrews
in actual combat maneuvers af d ircrns in weapon delivery boundsses. ACHR provides data display
tof
features which greatly enhance air combat debriefs, and provide a rich source of continuously recorded
quantitative measures. Some of the capabtlitios of AC'NR include the following:

1. Real-time tracking of aircraft engaged in air combat training in a specified airspace,

2. Video tape playback of flight history data, complete with pictorial display of the air-to-air
engagemrant and voice transmissions,

3. Both digital and graphic hard.copy printouts of flight instrument data, Interaircraft pocitions,
cockpit view of rngaocd aircraft, mission data, and

4. Computer generated estimates of weapon launch outcomes.

ACOR as a system enables training and research personnel to monitoa in real-time various air combat
training exercises, and through exercise replay, provides the opportunity to review, debrief and evaluate
pilot tactics, decisions, and eapon delivery accuracy. In addition, ei,,-ted ACHR advances are designed
to obtain meaisures in attach mission roles as well. Planned system aub.en:ation will cover no-bomb-drop
scoring, mine laying operations, anti-radisLion and electronic warfar,: rsuonu The whole array of
operational missions and their slow-motion replay will soon be within tlj prcvincc of aviation research
teams to better undaretand and resolve the complexities of pilot/aircreft sAtzhu0s.

RESEARCH REVIEW

In-Fliaht AcsessmenL Methods: Our technical approach (Brictson, Ciavarelli, at. a1., 1977) describes an
appropriate systems fr,'neork, training content, and performance assessment methodology for the development
of reliable and valid ACHR criterion performance measurea.

Measures from over 600 ACNR dog fights have been obtained across a variety of aircraft and weapon
systems, and under vasying training missions and operating conditions. A performanc3 assessment methodology
was developed and used to avaluate aircrew and squadron air combat performance. The performance assessment
methods include anal.sis c* engagement outcomes (wins, losses, drawt., as well as task accuracy measures
associatad with successful veapon delivery. Recently, we have developed metrics from the analysis of
antecedent events (i.e. radar contact, initial visual acquisition, first engagement short) in order to
estimate the probability of any given outcome, given certain antecedent condftions.
 Collectively, these assessment methods provide a complete measurement system for estimating aircrew
and unit proficiency in all aspects of air combat maneuvering. We will soon be able to provide longitu-
dinal and objective dsaa on all critical phases of air combat maneuvering.

Performsane ResulA.: Lince ACHR instrumentation provides su many output measures there is a range of
...Uiscriminant selection of candidate measures of performance. We ran across many occasions where it was
tempting to measure 'everything that moves,' but we chose instead to look at the statistical and practical
aspects of the data--recognizing fully that if your results do not make sense to the operational community
they will not be used.

To arrive at a reduced set of candidate measures we first identified thirteen air combat training
objectives and, using various logical and documentary criteria, selected weapon envelope recognition as
the most critical to success. A comprehensive statistical analysis, using ANOVA, multiple correlation
and discriminant analysis, resulted in the selection of two statistically and practically significant
variables from the multitude of measures available on ACMR. In the final analysis a single error score,
which was defined as a deviation from ideal weapon delivery boundary zones, proved to be the most promising
measure of envelope recogrition task accuracy. Based on that conclusion we have now developed empirical
distributions of these error scores for high and low pilot performance and experience continous for use
as baseline data to evaluate any future training innovations or system improvements in envelope recognition.

In general, the progress of ACMR performance criteria development has produced some very promising
results. We have, for example:

o Identified key variables related to successful weapon delivery,

o Developed preliminary criteria for evaluating aircrew performance in envelope redognition, and

o Devised scoring metrics based on engagement outcomes and task accuracy meLsures which have
demonstrated their effectiveness in discriminating known performance differences.

Hore importantly, we now have in-hand a list of statistically and practically significant variable which
not only account for the major portions of variance related to air combat success but are also -- and
this is critical to measurement success -- understood and accepted by the operational user, i.e. pilots
and training officers.

Efforts are continuing to further refine and expand performance assessment techniques, and to
establish the statistical integrity of the data base for ultimate application in support of both operational
training and for validation of ongoing aviator research. While the training application of these data
are readily acknowledged, the research aspects and potential have yet to be realized in the research
community at large. We hope that this brief foray will entice other aviation research teams to utilize
the tremendous capability now available in ACMR systems emerging around the world.

ANEW ERRA

For the past 30 years aviation psychologists, given the lack of objective operational measures, have
been forced to do research designed primarily to enhance the reliability and validity of subjective and
second order 'criterion measures.' Ustally the criterion measures so developed rested on the use of
flight instructor subjective estimates or pee- training which met with various degrees of success. With
the arrival of training systems such as ACMR aviation psychology has crossed the treshold into a new era.

The avilability of continuously recorded and objective output measures, along with on-line computer
analysis and display, present the researcher with a completely new capability to evaluate performance 'on
the job.' Although much remains tc be done to demonstrate the generalizability of initial performance
assessment methods developed to date, the methods have already been successfully demonstrated across small
samples and show remarkable promise.

The utility of reliable and valid objective performance criteria can net, and should not, be under-
stimated. From an operational view point, the measures are essential for judging the progress of AOR
training, estimating aircrew proficiency levels, and for determining the combat readiness of operational
units.

On the other hand, the research cosminity now has at fts disposal operational measures as potential
validation criteria for ongoing aviator selection, ttairing and research programs. The air combat mission
is most certainly one of the most demanding tasks in terms of skills required and stresses experienced.
ACMR provides a vehicle for the field validation of research directed at understanding the acquisition
of these skills and the conditions under which they may be enhanced or degraded.

Going hand-in-hand with operational measures related to aviator ccmbat missions is the present
availability of aircraft carrier final approach lands ýcores (LPS) which have already been tested and
validated in the fleet (Brictson, et. al., 19.73).. -.i and carrier landing measures, used independently
or in combination, provide a unique opportunity to support ongoing research related to the selection,
training, and performance effectiveness of Navy aviators.

Given the aviiability of operational performance measures, researchers can more effectively address
some of the questions that have arisen over the history of aviation research. Some of these questions
are of very high priority to the nation's defense in general, and to Naval aviation in particular.
example: For

When is an aircrew, squadron or fleet considered combat ready in AD(?

W.

2. What are the effects of ACK on pilot physiological responses?

 113

3. How frequent is ACM practice required to maintain proficiency?

4. What is the range of pilot stress tolerance to ACM missions?

5. What is an acceptable standard of ACM performance for difterent training levels?

6. How can operational performance measures be used to select top aviators?

i. And finally, what are the effects of sustair.zd operations, prnlonged duty hours, and operational
workload on the performance effectiveness of Naval aviatorr?

The answers to these and other operationally relevant questions can now be obtained given access to
on-line performance measurement systems such as ACMR and represeets an unequalled opportunity for aircrew
performance research.

RESEARCH OPPORTUNITIES

Of prime importance to research workers dealing with aviator workload, stress and fatigue is the
intriguing notion of an on-line pilot monitor system during air combat missions. Long considered to be
one of the more stressful and demanding pilot tasks, an air-to-air engagement taxes the pilot physically,
mentally and perceptually. The possibility of complimenting on-line pilot performance measures with
on-line physiological measures such as heart rate, blood pressure, etc. would provide an ideal arrangement
for the research team interested in validating laboratory notiont of stress, fatigue or workload in an
operational 'real world' environment.

A word of caution is advised. Some research teams used to the controls and precision design of
experiments in the laboratory will be limited in their attempts to control the real world. But that is
exactly the point. Many laboratory studies stress the statistical significance of results withLut strong
support for practical or operational significance. In pilot workload, for example, the amount or sverity
of workload in either a 24-hour or flight segment is certainly useful to 'describe' the environment but
does not by itself have any practical significance unless it can be related to performance effectivenes3,
short or long term. Our physiological reactions to stress or workload can assuredly be measured but it
is only in the context of their relation to performance that they acquire operational significance.

With the advent of sophisticated instrumentation systems like ACIR and the concurrent development of
performance criterion measures the final building block in field calibrated research is in place. All
that now remains is the historical challenge of innovative and understandable test designs that can answer
operationally significant problems.

Out own approach in AOIR is to provide, first of all, valid and reliable performance criteria.
Sccondly, we want to obtain a longitudinal performance data bank based on pilot biographic, experience,
biochemical, sleep, moo6 and workload components. Third, and most important, is our interest in having
a field laboratory that cat provide an arena Lo explore, define and predict the influence of Pilot temporal
variables on aviation performance effectiveness.

The ACHR system, while now prevalent in the contental U.S.A., is also being made available to NATO
nations for training purposes at a location in Sardinia. NATO scientists, ideally, could have access to
the performance data through part-tima use of the facility for research purposes. Many of the papers
recently discussed at the 1977 Cologn AGARD Panel meeting on pilot workload could benefit from on-line
performance measurement data such as that provided by ACMR. In addition to land based ACMR systems there
is a strong likelihood that ACKR, with its vast potential for tapping continuously many aspects of pilot
performance and physiological responses, will also be available at sea, abcard various U.S. Navy aircraft
carriers. If that planned installation occurs then the use of ACHR for research purposes could greatly
expand due to greater availability of ACMR facilities at sea and ashore. Regardless, it is now possible
to obtain from ACMR reliable and valid operational meausres of air combat maneuvering. Such measures
should provide a wezlth of opportunity for research teams from NATO, USN and USAF communities.

REFERENCES

1. Brictson, C. A., Burger, W. J. and Wulfect, J. W. Validation and application of a carrier landing
performance score: The LPS. Inglewood, California: Dunlap and Associates, Inc., March 1973.

2. Brictson, C. A., Cievarelli, A. P. and Jones, T. N. Development of aircrew performance measures

for the air combat maneuvering range (ACMR) (U). Pensacola, Florida: Naval Aerospace Medical
Research Laboratory lechnical Report L53001, June 1977. (CONFIDENTIAL)

3. Brictson, C. A., Ciavarelli, A. P., Pettigrew, K. W. and Young, P. A. Performance assessment

methods and criteria for the air combat ma,.suvering range (ACMR): Missile envelope recognition (U).
Prepared for Naval Aerospace Medical Research Laboratory, Pensacola, Florida. Prepared under
contract N61339-76-C-0082, November 1977, (In press.) (CONFIDENTIAL)

4. North, R. A. and Griffin, G. R. Aviator selection 1919 - 977. Pensacola, Florida: Naval Aerospace
Medical Research Laboratory, Special Report 77-2, October 1977.

5. Thorndike, R. C. Research problems and techniques. Washington, D.C.: U.S Government Printing
Office, Army Air Forces Aviation Psychology Program Research Reports, Report No. 3, 1947.

Note: The research reported in this paper was completed under Navy contract N61339-77-C-0167. The
opinions expressed here are those of the authors and do not represent official Department of the Navy
policy.
 115

RoyaL Air Force Institute of Aviation Medicine

Farnborough, Hw htre, Unite6 Jingdom

SUMMARY

The use of speech patterns in the analysis of workload is examined. The rather sparse amount of
research effort expended in this field is reviewed in terms of a simple model of speech production and the
applications of current analysis techniques are considered.

INTRODUCTION

There is much intuitive evidence to suggest that high workload or stress may change the fundamental
characteristics of speech, and so although the voice may nor exhibit obvious variations during normal flight
profiles, ;,search for change in speech may prove to be a worthwhile approach in the investigation of work-
load in air operations. However, central to the possible use of speech patterns is the requirement to
reduce complex speech date to parameter sets of a manageable size, and to relate these sets to the psycho-
logical and physiological state of the pilot. Optimum choice of parameter sets constitutes a difficult
task, but there is an ever increasing literature concerned with speech processing which provides many
techniques of analysis.

Reliable voice parameters may be extracted from the relatively poor quality speech of existing flight
ctmnunication channels, and so speech patterns may prove to be useful, as they overcome the need for sub-
ject instrumentation and data collection (see for example Refs 1-3). Correct choice of speech parameters
may make it possible to assess changing workload patterns, and this may be important in the military
environment where rapid fluctuations in workload and stress are encountered, and where many other methods,
such as those which rely on biochemical analysis (see for example Ref 4), may be of little value.

THE SPEECH WAVEFORM

There are several questions raised by the use of voice analysis in the investigation of workload in
aircrew. In common with many other techniques, the processes underlying the variations in voice parameters
are uncertain. An effect may be produced in response to endocrine changes when it is likely that the res-
ponse time will be long in relation to the duration of the workload and the induced stress. If on the other
hand, the change arises essentially from increased neurophysiological activity, then the response time of
the effect will bp rapid. In each case, the basic mechanisms of speech production would be similar and are
documented in the literature (see for example Ref 5), but there is little information concerning the pro-
cesses which may invoke variation under high workload.

Further, vocalization is a conscious process. The majority of fundamental parameters which

characterise a given speaker may be split into those which transmit the information content of the spoken
word, and those which do not contain this semantic infoination. The former group, for example formant
frequencies, vary according to the particular utterance, uut over long periods of time, as changes due to
semantic information average out, the range of variation is beyond direct conscious control. The latter
group contain measures such as fundamental frequency or pitch, which are responsible for intonation.
Western languages do not require changes in pitch to transmit semantic information, but conscious control
may produce signifi-ant short term variations. Although the nature of the short term changes may be of
interest, mean fundamental frequency in the long term is again thought to be beyond direct conscious con-
tiol.

In view ol these considerations it is worth reviewing the way in which a set of parameter estimates
should be used. As an illustration, a voice parameter from a pilot is measured during the course of a
single flight, and it is ai.umed that, initially, there is no knowledge of the influence of high workload.
The time course distribution of the estimates of this parameter during the flight will depend upon the times
at which the pilot chooses to speak. The estimates are likely to be corrupted by noise due to poor record-
ings, problems of measurement and random or conscious variations in the pilot's voice. The absolute values
are likely to be of little value, but relative changes through the flight profile may be of greater
interest. Statistical methods are available to establish whether any trends exist, and to test if a
particular aspect of the flight profile shows a significant change. Similar methods would be applicable if
estimates of the same speech parameters from an unstressed situation were available, either in flight or on
the ground. Measures of the relative change could be quite useful, given a sufficient knowledge of the
flight profile which would identify times of high workload. However, utterances are often short and
randomly dispersed through the flight profile, and so in preliminary studies, it would be desirable to
correlate with other physiological data. Such data are eas~er to gather from transport flights because of
the ease of instrumentating the pilot.

Data from many flights are necessary to establish the existence or otherwise of specific trends
related to high workload, and if trends are found, it would be necessary to establish whether they are
reproducible in different pilots. Studies of changes in the voice under stress have demonstrated wide
inter-subject variability (Refs 6-7), and these observations raise a much broader question, concerning the
way in which various parameters from different pilots could be evaluated for indications of high workload.
Given speech data from a single pilot, it i possible to use techniques of ever increasing complexity until
changes are found which significantly reflect high workload situations. Such studies are time consuming,
and, even so, the final technique may or may not be relevant to other pilots. Alternatively, more simple
techniques may be applied to data from several pilots in an effort to establish trends across pilots.
Intuitively, the latter approach is felt to be more realistic, even if some aspects of speech requiring
 116

involved methods of extraction are ignored

It is also necessary to bear in mind the techniques of analysis which are available,
relate to the quality of the available data. For example, inverse filtering and how they
methods may be used to obtain
the shape of the glottal pulse, which is the bnsic element in the quasi-periodic
fundame~atal frequency. waveform responsible for
To do this, spotech recordings of adequate quality are essential (Ref 8). Finally,
when considering the existing literature, the relevant information is sparse,
but section 4 offers a brief
review, in which it is hoped the problem and various approaches may be put into
perspective.

SMODEL OF SPEECH PRODUCTION

Though a description of the physiology of speech production would be out of
place in this review,
nevertheless, it is worthwhile to describe the genesis of the speech waveform. This presents
the oppor-
"tunityto define aspects of the waveform which can be measured, and which may reflect a high workload
situation. Figure 1 illustrates the way in which speech may be broken
down into its phonetic components.

I
SPPEECH*AVEFORM

I
II

Fig 1
Phonetic composition of the speech waveform
Any utterance consists of periods of vocal activity and non-activity, known
respectively as speech
intervals and pause intervals. In isolation, the latter are of no interest,
but together they provide
information on the speech pause ratio, and on the overall rate at which the
pilot is talking. This
apparently trivial point is of some importance, especially when obtained as
part of an analysis of the
speech waveform envelope shape. The envelope shape reflects the duration of
phonetic segments as well as
overall articulation, or the precision with which different sounds are produced.
contain information on high workload situations (Ref 6), even though it is Such measurements may
likely that this information
only reflects changes in the pattern of respiration. Unfortunately, the discontinuous
nature of cockpit
communication rarely provides a speech epoch of sufficient length for this form
of analysis.
broad Figure 1 also shows
classification that speech
ia dependent uponintervals may be ordivided
the presence, into
absence, of voiced
vocal and unvoiced segments. This
chord activity. Speech intervals
can be described by the model illustrated in Fig 2, which is based on the acoustic
theory of speech pro-
duction (Ref 5). In digital form, the model has found extensive application in computer based
analyses
which extract voice parameters from the speech waveform (see for example Ref 8-12).
The first part of the

-- - -- --- --- ---r -I

Vocal InVoca I

foreue
trcct

ti GFig 2
• Model based on the acoustic theory of speech produr't on
 117

model comprises two possible excitation sources and a source filter. The type of speech depends largely
on the excitation source, with the random noise generator producing unvoiced sounds or fricatives. In
actual speech, a constriction is formed in the vocal tract and air is forced through it, generating tur-
bulance, and hence noise. A combination of the random noise generator and impulse tra, generator can
produce the so callud v'cicedfricatives, and plosive sounds are created in a transitional phase between
pause intervals arn! voiced or unvoiced speech intervals. However, none of these three types of sound has
any simple application to the current problem.

Mobre important are the vowel sounds, or voiced speech sounds, which are derived from the qunsi-
periodic impulse train generator. The instantaneous period of the pulses defines the fundamental frequency
of the voice, which usually ý,i•s in the range 80-300 Hz. Many speech analysia-synthesis systems are based
on voiced speech models, and as a consequence the vocal source spectrum and vocal t'act resonators need
only be considered in relation to this type of excitatlon. The concept of an impulse train is an
idealixation, because practically, puffs of air are relsased into the voncal tract by vibration of the voal
chords. The shape of each puff, known as the glottal pulse, is determined by the vocal source spectrum,
and is largely dependent upon the state of the larynx and vocal chords. Figure 3 illustrates a aiIple
electrical circuit which represents the sub-glottal system. The bronchi and trachea are represented as T

P
Lungs i Trachea Glottal am

Z, I Z •
.Ii
IL Z37 I R ýL

PI in Vl

Fig 3
An electrical circuit representation c~f the sub-glottal system

sections, driven by a voltage representing the alveolar pressure, Pav Elastic recoil in the lungs, ie
charge on the lung compliance capacitor C, is sufficient to produce normal expiratory airflow. The lung
tissue resistance is negligible and may be ignored. During phonation, however, inspiratory muscle activity
will produce a negative, (subatmospheric) intrapleural pressure and impede expiration. This produce-i a
highly regulated expiratory flow through the glottal area. Usually airflow is =mall and so the sub-glottal
pressure, P., and alveolar pressure are nearly the same. The resistance and inductance, R.and Lg respec-
tively, represent the variable area glottal orifice. For voiced sounds in the normal pitch range, the4
resistive term is dominant. In the context of stress analysis, the properties of this model are
conveniently summarized by the fundamental frequency of vocal chord activity and vocal source spectrum, as
viewed from the vocal tract.

The physical factors which control fundamental frequency and the vocal source spctrum are closely
related, and it~ has been suggested that they are important in evaluating high workload situaticnis (Raef 7
'II-

and 13). This implies that the larynix is subject to the normal neuromuscular manifestations of stressful
situations (Ref 14). once again, the respiratory pattern may be important since an increase in sub-glottal
pressure can change the shape of the vocal source spectýLum by effectively narrowing the rqlottal pulse.
Much of the literature concerned with stress in the human voice has used fundamental frequency as the
indicator, but the glottal waveform has found little application, presumably, due to -be computational com-
plexities involved in its measurement.

The final feature of the vocal source is the gain multiplier, which has the effect of controlling the
overall loudness of the speech signal. Except under controlled recording conditions, it is difficult to
make use of amplitude information, or equivalently, absolute values in power spectra. There is an added
complication in that an increase in the loudness of the voice is generally accompanied by an increase in
fundamental frequency. In the final stages of a let-down, approach and landing, a possible increase in the
fundamental frequency of the pilot's voice may not be due to high workload, but rather, an increase in
voice loudness related to increased engine noise.

The -vocal tract is characterised by a series combination of quasi-time-invariant linear band-pass

filters, which are often termed ferments. Man~h filter is characterised by a resonant frequency and a bend-
width. Realist'.C estimatea of the resonant frequencies or formant frequencies may be obtain-ad from
relatively straightforward processing schemes, but this is not true of the fermant bandwidths. However,
speech experiments have suggested that with ccin tant bandwidth parameters, time varying combinations of the
formant frequencies offer realistic syntheses of O~e vowel so3unds in sections of voiced speech.
 Ths essential structure of each formant filter is shown in Fig 4. The filter transfer ! ,nction is given as

H(s) -
s 2 LC + sRC + 1
And it follows that the spectral peak occurs at

p 1l- 52
max L' 4=V)

L, C and R summarize the properties of air motion in a cylindrical tube. L is an acoustic inertance and
remai.ns essentially constant. C is a compliance term which depends on the cross-sectional area ot the
vocal tract, while R is a viscous drag term, dependent upon both the crLss-sectional area and the circum-
ference of the voc&l tract. Essentially, R controls thu formant bandwidLn and C the formant frequency.
Both of these parameters vary relatively slowly and so the formant system may be regarded as invariant in
terms of short-time analysis. In this context the process is considered stationary during periods of 20 mS
or less.

eo ~H(s)el

Fig 4
Electrical analogue, of a single foimant resonator

In theory tUiere is an infinite num&er of formants, but in practice, three or four are sufficient to
characterise a voice, although the acoustic theory of speech requires further filter elements for the
correct representation of nasal consonants (Ref 5). In the male, empirical data suggest the fi'st formant
lies in the range 200-900 Hz, the second formant in the range 550-2700 Hz and the third formant in the
range 1100-2950 Hz (Ref 10). Physically, the formant resonators comprise the cavities of the pharynx and
oral and nasal cavities. The tongue, jaws and lips are also able to modify the 3ow order formants. Some
studies have considered possible interactions of stress with formants, essentially by examining changes in
spL.Jtral balance within the formant frequency range. Such studies have been qualitative as well as
quantitative (Ref 15-17', arn ý,ill be considered in more detail later. Intuitively, however, since it is
the gl.ttal waveform which actually nha-icterises the voice, the vocal tract will he of less interest, as
iL nereiy shapes the glottal waveform to produce semantic infor-mation (Ref 18).

Th? last component in the mcdpl of speech production is the radiation load. This filters the speech
signal according to the way in uhich the vocal tract is coupled, via the mouth, to free space. In speech
analysis applications it is often of greatest importance to ,btain fundamental frequency and formant
parameters, and so the characteristics of the vocal source spectrum and radiation load spectrum may be
lumped to,-:-her anc removed fron the speech signal, a- both may be considered time invariant.

ME.AS'RE3 OF STRESS IN THE SPEECH SIGNAL

In this section we are concerned with methods which have been used to establish whether stress modifies
the speech signal.

Voice Micro-tremor. Although the previous sec-.on presented a model of speech production and highlighted
the aspects of vocalisation which are likely to roflect stress, there is a further phenouenon known as voice
"micro-tremor, which does not fit into the scheme, but is nevertheless important. Tremor, or to be more
specific, an 1-12 Hz modulation in the human voice, is a fairly recent discovery'. Commercially, the
phenomenon has found application as a;nextension to polygraph lie detector methods, and appears to have mel.
with some succes, at least in a well structured interview situation. One of the first devices offered
simple strip chart recorder output, and required a skilled operator to interpret the results (Ref .9j. A
more recent device has a direct digital readout of stress level, but there is little technical information
on its operation (Ref 20).
The proponents of such devices have attempted to explain the principles behind voice tremor, and,
essentially, it is assumed that the muscles controlling the vocal chords exhibit the sort of tremor which
accompanies activity in any of the voluntary muscles. It is postulated that this will cause slight rhythmic
changes in vocal chord tension which wIll result in an inaudible 8-12 Hz modulation of fundamental
irequency. Similarly, the muscles controlling the throat, lips and tongue are thought to be sensitive
to the same kind of tremor, which will be reflected as a modulation within the first formant bandwidth. In
a stress situation it is assumed that increased nervous activity causes muscle tension to increase
throughout the body and, in the larynx at least, this will reduce the micro-tremor. In a high stress
situation voice micro-tremor may disappear altogether. In view of the supposed mechanism of voice tremor,
this is a rather curious observation, since other manifestations of muscle tremor appear to increase in
"the high workload situation (eg Ref 1). However, Inbar et al (Ref 21) have attempted to measure voice
tremor and correlate it with muscle tremor in the area of the larynx. This technique was used to
establish if voice tremor was due to mechanical "subresonances" in the vocal tract, or if it was
, .!
generated by increased nervous activity. Their
of the glottal waveform, and that it is generatedresults
•, suggest that micro-tremor is a frequency modulation
by nervous acti-vity. Frequency modulations were also
 I ri

deteted in the first ad third foruant bandwidth@.

Although the aomercial application of the voice trerxor phenomenon is not in line with the current
application, the underlying process would appear to make further investigation worthwhile. Several
studies of commercial devices have been undertaken (ey Ref 14 & 22). Older and Jenny (Ref 14), carried out
a comprehensive evaluation using the voices of astronauts from Skylab III and Skylab IV aissionr. Their
conclusAons suggested that the voice tremor principle, as exploited in comercial devices, would not detect
any possible stress, at least in the Skylab situation. This may suggest that such devices are of real
valne only in the st:.uctured into-view application. However, it should be pointed out that the commercial
devices appear to be of simple design, and since we are not aware of any adequate investigations intn the
use of the voicu tremor phenomenon tn a stress situation, the use of micro-tremor in the analysis of stress
may still prove to be a useful approach.

General ectroazaphic Measurements. These methods attempt to quantify sound spectrograms either by
visual inspection, or by direct measurement. Such methodi can be effective .n demonstrating changes in
the voice, but it is difficult to obtain .-ecise measures. The most important spectrographic analyses have
used either wide band filters (200-400 Hz bandwidth) to emphasise the formait resonances in the speech
spectrogram, or narrow band filters (less than 50 Ha bandwidth) to highlight the h&rmonic structure due to
fundamental frequency.

Kuroda at al (Ref 13) have defini al quantity from the narrow band spectrogram known as the vibration
space shift ratio (VSSR). This is simply derived from meaaurement of the frequency band spacing during
voiced speech, and relates to the relative changes in fundamental frequency between normal and high stress
situations. Thus if in the normal situation, frequency band spacing is given by SVS, and in the high stress
situation, by IVS, then

VSSR . x 100%
SVS

Real situations in which military pilots found themselves in difficultios were examined. Highly signif' 'ant
increases in fundamental frequency were reflected in the VeSR, but each case represented a catastrophic
situation and three are known to have resulted in a fatal accident. Generally, in such situations, machine
analysis is not necessary to demonstrate the gross increases in fundamental frequency attributed to both
intense fear and concomitant increases in voice loudness. In the more commonly encountered high workload
or stressful situation, changes in voice parameters would be expected to be much less dramatic, and only
then would it be necessary to use some form of machine analysis.

More general evaluations of the way stress may appear in the spectrogram have been carried out by
Williams, Stevens et al (Ref 6, 7, 18). Some early studies used data, and produced results, which were
very similar to those detailed above, although the fundamental frequency contour was also deemed to be of
importance. More comprehensive studies attempted to extract as much Jnformation as possible from the
spectrogram, largely by inspection. For instance, irregular structure in the second and third formant
regions of a wide band spectrogram is thought to reflect a non-stable glottal waveform. The results of
these studies have been summarised by the changes in voice attributable to four emotions, rnmely anger,
fear, neutral and sorrow. These four emotions, in that order, tended to produce a fundamental frequency
which decreased in magnitude and range. Irregular glottal pulses were often seen in the anger and sorrow
situations, while unusual pitch contours were characteristic of the fear situation. Changes were also
noted in the syllabic rate and duration of utterances. Howevor, it should be noted that the majority of
these results were obtained in the laboratory situation. Two methods have been used. Early methods
attempted to induce stress using an arithmetic task (Ref 6), but this is likely to produce failure stress
as well as task induced stress. Wide intersubject variability was observed. The second method used
actors, and the majority of the above results were obtained in this situation. Between the different
emotions, the actors were able to produce clear changes in their speech, but the application of these
results to the real situation is op-n to question. This is particularly true of the flying task, where the
range of emotions is not directly applicable, and where again, changes in the voice characteristics of
highly trained pilots may be expected to be subtle in all but tha most extreme situation.

Average Spectrum Measurements. A second example of stress analysis methods which makes use of spectral
information uses the average spectrum. This is the spectrum of a complete utterance, and may involve a
single word or a longer phrase. changes in fundamental and formant frequencies during the utterance
give pitch and formant peaks a wider and flatter appearance in the average spectrum, and reflect the
overall characteristics of the voice during the complete utterance.

Tishchenko (Ref 15) has suggested that formant frequencies tend to change in the stressed situation,
and that spectral intensities within these bands also change. This led to the definition of the formant
momentum, which is the product of a formant frequency and its intensity. The data in Tishchenko's study
consisted of speech from 23 students before, during and after their first parachute jump. All of the
spectral analysis methods were analogue. Generally, the first formant momentum increased in the stre3s
situation, and the second and third momenta usually decreased although greater variability was observed.
explained physiologically, but account had to be taken of the different vowel sounds
This behaviour ý--as
present in the single words which were analyred. Thus shifts in formants due to different vowels could
augment nr reduce apparent shifts due to strims.

Polpvw, Smonov, Frolov et al have attem )ted L aWalyse Ctress and emotions using spectral methods
(Ref 16, 1l;,24). Early work wae| di,:ected t •wards the measurement of changes in the average formant
structure of single words (Ref 16). Results •rimilar to those suggested above by Tishchenko were reported.
The data were obtained from actors, but further studies used speech from the cosmonauts in the Voskhod 2
spac6craft. Again a centroid spectrum method was used to give an indication of relative shifts in formant
peaks. Analogue spectral techniques defined the average spectrum cantroid as
u u
f E-l / -l P
if
 where the f are the filter tuning frequencies and the P are the average pover outputs of the
filters, measured ov• the effective time of output for each filter. &hoice of U appeared to be
empirical, but significant relationships wor established between relattive chances in the centroid and
heart rate, during various stages of the space flight. This, together with a. kawledge cf the cosmonauts'
tasks at each flight stage, suggested that changes in f, could reflect a stresefLl situation. Further
studies have analysed the envelope shapp of the output of each of the bandpass fiLters. Specifically, the
time integral of the output envelope of each bandpass filter is calculatnd as

T
a V (t) dt
0 t

where T is the analysis period and the Ui ar, the envelope shapes. it is suggested that empirical
combinations of the ai can actually distinguish between different types of emotion, labelled as fear,
anxiety, joy arn delight (Ref 16-17). However, reasonR for the choice of a, combinations are not explained.

Williams and Stevens (Ref 7) have described similar methods in which they have analysed the average
spectra from several seconds of speech. Their findings are merely consistent with increases in speech
loudness in the "anger" situation, and a decrease in speech loudness in the "sorrow" situation. The data
ware derived from actors' speech.

Direct Measurements on the Speech Waveform. Recent studies by Simonov at al (Ref 23) have suggested that
crude measures of fundamental frequency (designated by FaT) and first formant frequency' (designated by F,)
may be used to discern emotional states. Apparently, the&e parameters were extracted directly from the
speech
of the waveform,
parameters,butapproaching
the analysis methods
were were not described
1Wo%, reported. The use in
of any detail. However,
a discriminant functionvariations
tn the FaTin - each
f,
plane was therefore suggested to differentiate between so called states of rest and emotion. Again, the
main hulk of the work was performed with data derived from actors' speech, although the validity of the
methiod was supposedly confirmed using speech obtained from amateur parachute Juampers. To the amateur, a
parachute jumip clearly presents a highly stressful situation, but no mention 4as wade of possible physical
stress interactions.

We have performed similar experiments with the voice of a commercial airline pilot (Ref 25). Cepstrum
methods were used to obtain fundamental frequency estimates and to smooth log magnitude spectra, fiom which
formant information could be extracted (cepatrum methodologies are described in the next section). The
data consisted of 22 larmings into various international airports. For each landing, the baseline or
unstressed fundamental frequency and first formant parameters were obtained from about 30 seconds of speech
at the top of descent. Parameters in a stressed situation, as indicated by an increase in heart rate, were
obtained from about 30 seconds of speech taken around the touchdown instant. Thus for each landing,
specified by the index i, the data yielded fourparameters: foilu and foils are respectively the unstressed
and stressed mean fundamental frequencies, and Flilu and Flil are respectively the unstressed and stressed
mean formant frequencies. Normalisation of the data was carrfed out in terms of the overall mean parameter
values derived from the unstressed data in all 22 landings. Thus

f omean 22
T •E rJ.
i.1

Fimean 9- F lilu

and so f oilu foilu / fomean

oi f
Iu foils/f
o omean
flilu " lilu / olmean

Fiiiu F lul / F1lea

where signifies normalixed. These data are siarised in Fig 5 which plots parameter
variations from the stressed and unstressed centroids in each of the 22 landings. The ientroid of the
unstressed data lies at the origin, but, it can be seen that the stressed data centroia is shifted to a
position representing an increase in first formant frequency, but a decrease in fundamertal frequency.
The distance between centroids reflects the degree to which stress is manifested in these two speech
parameters. An application of the T2 test to the raw data demonstrates a difference in centroids
(P < 0.0002), but Fig 5 does suggest that the discriminating power of these two clusters may be somewhat
restricted (Ref 26). A section of speech obtained from either the top of descent or just before touch-
down, cannot be assigned to the stressed or unstressed group with any degree of certainty, at least not
solely on the basis of fundamental and first formant frequency measurements. These conclusions are at

I: variance with those of Simonov et al (Ref 23) who have claimed much greater stress induced changes in
their speech parameters.

It is apparent that some form of measure on fundamental frequency and its variations is essential in
an analysis of stressful situations. Various measures on the formants, in particular the formant
"frequencies, appear to be quite promising.
 121

SI I I I

..

€ "STRESSED"
H 2

X i-• •*UNSTRESSWD;

L U-
0L' -2_

-4

-6

-s-2u -is -i -S 0 S i
FUNDAMENTAL FREQUENCY VAR.'1TION (X)

Fig 5
Simmary of fundamental and forment frequency variations during 22 comercial landlings

SPEEM PROCESSING TECHNIQUES

Speech analysis literature provides a wealth of information covering many areas of application, all
of which are based on a requirement to reduce a speech signal to a concise met of parameters. The
applications fall into two categories.

1. Reduction of bandwidth requizements in speech counication channels and the automatic machine
generation of speech. Neither of these applications is relevant to the current discussion.

2. Speech recognition applications. This broad area requires recourse to statistical and pattern recog-
nition techniques to classify the sets of speech parameters. Speech recognition can mean the extraction of
phonetic or semantic information, and the methods which have been developed in this area are often bas-d
on prototype or template speech parameters (Ref 27). In the context of stress and high workload analysis,
the techniques of speech recogr.ition which are of greater interest are those aimed at identifying the
speaker rather than his speech. Considerable effort has been expended in developing means of assigning the
voice parameters of an arbitrary speaker to a specific "library" parameter group. It in possible to
identify particular speakers from relatively large populations if a section of their speech is available
for reference purposes. The relevance of these methods is obvious, and it may be possible to assign short
sections of speech from a given spaaker to known stressed or unstressed parameter groups. It may also be
possible to develop a system using several levels of stress, rather than a stressed-unstressed binary
quantisktion.
A description of all the available processing techniques cannot be attempted but two techniqueR which are
felt to be directly relevant in the analysis of stress and workload will be considered. The use of
cepstrum techniques is common and so the implications of these methods will be considered in some detail.

Cepstrum Techniques. Cepstrum analysis is a powerful methodology which may be used to analyse a voiced
speech signal by separating out the contriItion 'tue to the glottal pulse ind the contribution due to the
formant filters. It is possible to identify voiced and unvoiced intervals, and within the voiced intervals
vu obtain estimates of fundamental frequency. Further, the cepstrum technique can provide smoothed
ej~ectral estimates from which formant information can be obtained (Ref 9, 10). Using the model presented
in Fig 2 the voiced speech output signal may be assumed to be a convolution of the vocal source impulse
train with the impulse responses of the various filters in the system. Thus, denoting convolution with a *,

x(t) - p(t) * s(t) * F(t) * r(t)

wht,..e x(t) is the speech output

p(t) is the impulse train
s(t) is the impulse response of the source filter
F (t) is the combined impulse response of the formant filters
and r(t) is the impulse response of the radiation load filter.

Combining the offects of the source and radiation load so that tho vocal source output assumes the form
122

s (t) - p(t) * s(t) •r(t)

then x(t) - 8(t) * F(t)

or equivalently. since convolution in the time domain is identical with multiplication in the frequenz-f
domain,
X(M) - S(M). F(M)
where X(w) is the speech magnitude spectrum.
A logarithmic txnsform has the effect of separating the elements of X(w) into additive components, is

t,.i{X(M) ) - LnWS(W). 7(w)) - Ln{S(w) I + Ln{((w))

Ln[X(w) I has the appearance of an undulating function representing ferment structure with a superimposed
"high frequency" ripple representing the harmonic structute of the vocal source spectrum. The additive
components in Ln(X(w)I are maintained during inverse frequency transformation which results in the so
caiii cepstrum. Clearly, the harmonic structure in the lo> magnitude spectrum manifests itself as a
sharp p%ýak in the cepstrum fxom which pitch period may be determined. There are available, several
efficient algorithms which implement cepstral pitch peak picking, and we have developed &n algorithm
based on a design by Noll (Ref 28), which has proved to be very useful.

This tachnique, ther.:, is a relatively simple method of measuring fundamental frequen.y, and is based
on the harmanic structure of a log magnitude spectrum. As a consequence, the fundamental frequency com-
ponent neec. not be present in the signal being analysed. Further, our experience has shown that the method
works well, even in the presence of considerable noise, for instance, with a signal to noise ratio as low
as 5 dB, (as defined only during voPcei intervals). In this context, noise refers to the acoustic noise in
the cockpit environment ae well as to .my electrical noise intrtduced by the comuunication and recording
equipmaent.

It is of interest to comparE the cepstrum method with other simple pitch extraction routines.
McGonegal et al (Ref 30) have evaluated cepstrum methods together with low pass filtering and auto-
correlation techniques. The autocorielation function is quite similar to the cepstrum except that in the
latter, pitch peakp are more prcnounced due to the logarithmic tansform in the spectrum. With the excep-
tion of identifying ,iiced-unvoiced transitions, the three pitch extraction methods ment ;oed above were
shown to be quit. sm•ilar in operation.

Cepstrum analysis however, ha" other advantages when searching for parameters to characterise the
spectrum. Since the log magnitude spectrum and cepstrum are Fourier transform pairs, the low order
coefficients in the cepstrum contain spectral envelope shape information. This observation provides us
with two alternative methods for obtaining spectral information.

1. The low order cepstrum coefficients may be used directly as parameters which classify the speech
spectrum.

2. The cepstrum may be short time filtered and transformed back into the log magnitude spectrum. Ferment
picking algorithms may then be implemented to characterise the spectrum.

The first method is computationally faster, and haai found extensive use in talker identification
applications. However, the second method is less prone to corruption due to noise in the original speech
waveform and is physically more meaningful. It is suggested that in the current application, the second
method offers the more viable proposition. Within the framework of cepstrum analysis, there are several
methods of extracting formants. Schafer and Rabiner (Ref 10) have provided a robust method (peak picking)
which makes full use of empirical data. Alternatively, Olive (Ref 29) uses the model of speech production
in an iterative spectrum matching technique (analysis by synthesis). Both of these methods make use of
amplitude information, but this is not practical in the current application. We have had some success using
an algr, ithm based on ,*chafer and Rabiner's design. The algoriium disregards amplitude information except
fc¢. • t'ative changes w.thin specific ferment ranges, and currant formant peak picking decisions are, in part,
based on previous decisions.

At this stage we must consider the choice of analysis interval, that is the length of the speech epoch
used to obtain fundamental and formant frequency estimates. Since these parameters will vary during an
utterance, the analysis interval should be arbitrarily short. In practice of course, a compromise is
necessary. At least four pitch periods are desirable to obtain a strong peak in the cepstrum, but within
this time at is quite likely that one or more of the formant peeks will have moved, prc-ducing a smearing
effect in the spectral envelope. Generally speaking, the analysis interva, is chosen to contain up to four
pitch periods (20-40 mS), but successive intervals are overlapped, to have centres which may be only 10 mS
apart. Individual fundamental and formant frequency estimates can be used to form contcurs or profiles
covering a complete utterance.

It is usual to implement cepstrum analysis using fast Fouxier transform (FFTr methods, either in hard-
ware or software. An important problem which is closely related to the choice of analysis interval concerns
the choice of sampling rate and FVT transform size. Assuming formant information is required, a minimum
sampling rate of 8 kHz is desirable: 10 kHz is more usual. For fundamental frequency extraction alone,
lower sampling rates may be used, but this will result in significant quantisation error. It can be shown
that time resolution in the cepstrum is given by

TCR - I/FS

where FS is the sampling frequency.

Consider a pitch peak at the nth cepstrum coefficient. Fundamental frequency is then given by
 123

fo(n) = l/(nTCR)

Resolution in to(n) is inversely dependont upon n and so for a given sampling frequency, the maximm
quantisation error incxeases as fo(n) increases. This is illustrated in Fig 6 which plots cepstrally
derived fundamental frequency against the maximum quantisation error, AQ, at that frequency. At the nth
cepstrum coefficient the quantisation error is defined as

efo(n+l) - fo(n-l) 1
n 2 (n9-1) TCR

Intuitively, for the expected changes in fo induced by stress and high workload situations, AQ should not
exceed 2 Hz. Thus if fo does not exceed 150 Hz, a minimum sampling rate of 8192 Hz is sufficignt.

,i6-
O FS = 2948HZ
e FS =•124IHZ
w
7- 12

4
H

25 so $5 is@ 125 SI 175 291

FUNDAMENTAL FREGUENCY CHZ)

Fig 6
Effect of sampling rate on fundamental frtquency resolution

Given a suitable sampling rate, choice of transform size is restricted by the analysis interval
requirement. But, if formant picking is to be implemented, a large transform sime is desirable to give
good resolution in the spectrum and to avoid aliasinq problems in the cepstrum.

Now B Fs/S

where 8 is the spectral resolution

and S is the transform size.

Given an 8192 Hz sampling rate, a minimum transform size of 1024 points should be used. This implies
an analysis interval of 0.125 seconds. It is therefore usual to pad the actual analysis interval with
zeros for tranrfornmation purposes. A schematic illustration of the complete cepstrum analysis procedure ih
given in Fig 7.

Linear Pr~dictive Coding. Linear predictive coding i! a form of inverse filtering which models the speech
waveform itself rather than various aspects of the speecb spectrum (Ref 11, 31). The contxibution of the
vocal sourcm and vocal tract to the speech signal are not separatd out and it is possible to track rapidly
changing speech processes which may be lost in the relatively long analysis intervals associated with
Fourier methrdAs.

In ersence, during the segment of speech to be analysed, the nth speech sample, Sn, is given as a
weighted sum of the previous p values. n

S kE ak Sn-k

t
The weighting coefficients, ak, can be obtained by first calculating the prediction error, n, as

p
9n = sn Sn S n- =n as -k

where S is the value of a speech sample and Sn is its predicted value. E 2 is then averaged
over all n in the current speech segment to _!orm a mean square prediction error whicfl is minimised by
choice of the ak. The number of coefficients needed to reoresent a speech segment is given by p, and
 124

RE
RE

AANI

I REAL TIME SERIES CIINDOIED)

FURR
.. TR1 . .ANSFR

LOS MAGNITUDE SPECTRUM CEPSTRUM

RE RE

IE F

ImI

SHORT TIME FILTERED CEPSTRUM SMOOTHES

LOG MAGNITUDE SFECTRUM

Fig 7
Stylised representation of the stages in a cepstrum analysis

depends on the model chosen to represent the vocal source and tract. It can be shown that twelve
coefficients are generally adequate. Thus for a speech signal digitized at 1O k~Iz, the 100 samples taken
over a 10 mSec analysis interval, may be represented by just fourteen parameters, that is twelve weighting
coefficients and a pitch period, together with a binary voiced-unvoiced decision.

Fundamental frequency estimates are obtained easily with this method since it can be shown that the
prediction error, En, is a maximum at the start of each pitch period (Ref 11). A simple peak picking
procedure may be used on the En series to identify the points in the speech time series at which a pitch
impulse occurs. Peak picking is independent of the analysis interval and so very short inteivals (as low
F? as 5 mS) may be used, even when fundamental frequency information is required. This offers distinct advan-
tages over cepstrum methods which can only evaluata an average fundamental frequency over the duration of

a much longer analysis interval. Furthermore, since predictive coding is a time domain method, it is sig-
nificantly faster than corresponding frequency domain methods. However, when using predictive coding for
pitch extraction it is desirable that the pitch fundamental be present in the digitized speech signal.
This, together with the poor quality speech in existing communication channels, makes linear ixrediction
less attractive in the current application.

The advantage of using very short analysis intervals is of more interest if we consider formaynt extrac-
tion. It can be shown that the weighting coefficients, ak, define the poles in the vocal tract transfer
function, and it is possible to obtain both formant frequencies and formant bandwidths. However the
problems associated with poor speech quality will again be an overriding factor.

STATISTICAL TECHNIQUES

In the previous section we have shown the way in which speech analysis allows parameters such as
fundamental and formant frequency estimates to be derived during a short time period, and, it has been
suggested that time seý-ies profiles of such parameters may indicate strebs or high workload. It is
necessary to establish the validity or otherwise of this hypothesis, and so we will briefly consider some
of the statistical tuchniques which may be used to classify or to group series of speech parameters.

The first question concern3 the length and nature of the speech epoch to be analysed. To some extent
this will be dependent upon the type of the statistical analysis. For example, the examination of single
words or single phonems is only feasible under a limited set of conditions. The same phonem or word must
j
be chosen for analysis from the different stages of a flight profile, and this is particularly important in
7
•12S

the case of an analysis attempting to use formant information. With regard to fundamental frequency, even
though it is expocted to remain essentially constant, our experience has suggested that in the course of a
single word lasting less than one second, variations due to intonation mask any possible change due to the
workload situation. This problem may be partially overcome by restricting the choice of phonem or word to
those which may be obtained from similar circumstances during the flight profile. In this respect, call-
signs reprusent useful information, although the automatic and often emotionless manner in which they are
uttered may or may not be an advantage. The analysis of single word call-signs is being actively pursued.

Problems due to intonation, varying formant structure or simply random variations, may be overcome by
analysing longer sections of speech and averaging the resulting parameters. Although a simple averaging
procedure is a valid approach, any other measure which classifies the shape of a parameter profile should be
considered. For example, moments about a fundamental frequency mean are useful since it is inwise to dis-
regard intonation information completely. When considering such methods, the question of a suitable length
for the speech epoch arises. Long term feature %veraging experiments (Ref 32) have suggested that a mean
fundamental frequency obtained over a 20 second epoch reduces the sample variance due to intonation and
random variations to acceptable levels. In this context, the epoch describes 20 seconds of voiced speech
which represents a considerably longer section of normal speech. In the current application, it is
unlikely that such lengthy sections of speech wil l be available. It should also be noted that this
methodology derives long term average parameterF ,ing a short time analysis technique and results similar
to those obtained using the spectrographic methoý described in section 4 are to be expected. It is
desirable therefore to use parameters other than the simple average.

The above discussion suggests that the voice may be searched for signs of stress of high workload in
terms of single phonems and words such as call-signs, or in terms of the properties of longer sections of
speech. In either case, a complete data set may be viewed as a series of vectors, Xn: n - 1,2 .... m.
Each vector represents the p speech parameters which are chosen to characterise a particular situation.
Thus
_Xn- fXnl, Xn2 -- X-n

In the case of single phonem analysis the components in X may represent the elements of a pitch or formant
profile, while for longer sections of speech, the components in X will represent different properties of
the whole speech epoch. The m vectors are obtained at various stages of the flight profile, and it would
be hoped that differences in the structure of the vectors reflect changes in the stress and workload
situation.

Such data require recourse to multivariate statistical methods. Principal componert xnd factor
analyses (Ref 33) are obvious candidates. Such methods reduce the dimensionality of the vectors &nd as a
consequence, can be used to demonstrate possible groupings in the original parameters. If consistent
changes can be produced in speech parameters, then well defined stressed and unstressed situations will
resolve into two distinct groups in the vector space. Of greater importance however, is the fact that
these methods form the basis of techniques such as linear discriminant analysis which effectively optimize
the ability to distinguish between different groups of parameter vectors. These methods have proved useful
in speaker identification experiments, using parameters extracted from long sections of speech (Ref 34),
and are felt to be useful in the current application. For example, the data presented in Fig 5 forms the
basis of a discriminant analysis using just two parameters: the inclusion of further parameters which may
provide better separation of the centroids is desirable.

Mathematically, discriminant analysis is a powerful tool, but well defined "training" data sets are
required. In the current application this implies that for each subject considered, it is necessary to
obtain sections of speech in both stressed and unstressed situations. If such data can define diseinct
groups then it is possible to assign an arbitrary sample of speech to one of the two groups. The
reliable collection of training data constitutes a major problem, particularly in the stressed sitaletion
which is not easily definable. In this respect, the availability of other physiological data is of some
importance, at least during training procedures. Thus for the data presented in Fig 5, the stressed-
unstressed decision was based partly on a knowledge of the workload patterns in the flight profiles, but
mainly ci. the measured heart rate patterns during the flight. A simple correlation analysis between
physiological data and speech data is also proving valuable in establishing which speech parameters contain
useful information.

CONCLUSIONS AND RECOSSENDATIONS

The salient features of this review lead to recommendations, which may constitute a methodology for
the investigation of high workload using speech patterns.

.. The aim should be to reduce the dimensionality of speech and provide a succinct description of the
data. This must be done in a way which preserves the possible stress or high workload Information. Also a
•tatistically robust method must be found to classify the reduced ipeach data into at least two groups of
stressed-unstressed parameters.

2. The nature of the data requires some attention. It Is suggested that an attempt to extract relativily
simple speech parameters from many flights acrosm several subjects is the most viable approach. In the
long term, a complex and sophisticated analysis on a limited set of data may not be profitable.
3. For a given flight profile it is desirable to have a knowledge of the likely stress or high workload

patterns, together with some indication of their rates of change. This information will influence the type
and length of speech samples which may be used. Thus for %apidly changing high workload profiles, only
single words or phonems may be used, but longer sections of wpeech may be employed for slowly varying
stress patterns.

4. Our experience has suggested that no matter what type of speech sample is chosen, a cepstrum analysis
technique offers a realistic compromise between the degree of processing power required and the amount of

...........
126

information preserved in the reduced data. At this stage, more involved processing methods, ,while
possibly being more powerful, are not considered to be worthwhile, and indeed, would not offor such
reliable results given the poor quality of the original speech samples.

5. Capstrum analysis can offer short time smoothed spectrum and formant information together with
fundamental frequency information. Pitch an.d formant profiles over longer periods of time may be easily
constructed. Previous research in thia area has suggested that such measures are of consideraole value.

6. Whatever statistical methods are employed, they must be capable of assigning arbitrary speech data to
some point on a stressed-unstressed scale. Our experience has suggested that initially, only a binary
quantisation of the scale may be pous!.ble. In any event, multivariate methods are necessary and linear
discriminant analysis looks very prouising.

7. It is unlikely that an absolute estimate of stress or workload can be obtained from a single speech
sample in isolation.

8. Simple correlation analysis of' speech parameters with physiological data appears to be a realistic
means of establishing which parameters will be of use in the long term.

9. Finally, voice micro-tremor has found limited uses in commercial devices, but a rigorous study of its
possible usage in the current application should not be neglected.

REFERENCES

1. Nicholson, A.N., L.E. Hill, R.G. Borland & H.M. Ferres. Activity of the nervous system during the let-
down, approach and landing: A study of short duration high workload. Aerospace Ned., 41. 4, 436-446,
1970.
2. Nicholson, A.N., L.E. Hill, R.G. Borland & W.J. Krzanowski. Influence of workload on the neurological
state of a pilot during the approach and landing. Aerospace Ned., 44. 2, 146-152, 1973.

3. McFeely, T.E. Pupil diameter and the cross-adaptive critical tracking task; A method of workload
measurement. Thesis, Naval Postgraduate School, Monterey, California, 1972.

4. Various. A preliminary study of flight deck workloads in Civil Air Transport Aircraft. Flying
Personnel Research Committee Report No FPRC 1240, 1965.

5. Flanagan, J.L. Speech analysis, synthesis and perception, 2nd Edition. New York, Springer Verlag, 1972.

6. Hecker, M.H.L., K.N. Stqvens, G. von Bismarck & C.E. Williams. Manifestations of task induced stress
in the acoustic speech signal. J. acoust. Soc. Am., 44. 4, 993-1001, 1968.

7. Will'.&ms, C.E. & K.N. Stevens. Emotions and speech: Some acoustical correlates. a. acoust. Soc. Am.,
52. 4, 1238-1250, 1972.

8. Hess, W.J. A pitch-synchronous digital feature extraction system for phonetic recognition of speech.
IZR Trans. Acoust. Speech and Sig. Proc., ASSP - 24. 1, 14-25, 1976.

9. Cppenhaem, A.V. Speech analysis - Synthesis system tased on homomorphic filtering. J. acoust. Soc.
Am., 45. 2, 453-465, 1969.

10. Schafer, R.t. & L.R. Rpbiner. System for automatic formant analysis of voiced speech. J. accust. Soc.
Am., 47. 2, 634-648, 190.

11. AtAl, B.S. & G.L. Hanauer. Speech analysis and synthesis by linear prediction of the speech wave.
J. acoust. Soc. Am., 50. 2, 637-655, 1971..

12. Schafer, R.W. A survey of digital speech processing techniques. TREE Trans. Audio and Electroac.,
AU 20. 1, 20-35, 1972.

13. Kuroda, I., 0. Pý%jiwara, N. Okamura 6 N. Utauki. Method for determining pilot stress thruugh analysis
of voice comunication. Aviat. Space Environ. Ned., 47. 5, 528-533, 1976.

14. Osr, N.J. & L.L. Jenney. Psychological stress measurement through voice output analysis. NASA
140port No CR-141723, 1975.

15. Tishchenko, A.G. Dynamics of formants Li-the spectrum of speech as objective irdicator of differences
betwen rAwitive and negative emotions. Envircn. Space Sciences, 2, 371-375, 1968.

16. Wopov, V.A., P.V. lieonov, M.V. Frolov & L.B. Khachatur'yants. The articulatory frequency spectrum as
on indicator of the dweree e•n nature of emotional stress in man. NASA Tech. Translation No TT F-
137%•, 1971.

state of attention. Aerospace Med., 44. 3, 256-258, 1973.

18. Williams, C.S. & K.N. Stevens. On dsterminirnjg the emotional state of pilots dnring flightt An
exploratory study. Aerospace Ned., 40. 12, 1369-1372, 1969.
 127

"•i !' 19. Counterintelligence

Springfield, and Security Inc.
Virginia. 'Dektor' Psychological Stress Evaluator. Port Royal Road,

20. Communication Control Systems Inc. Voice Stress Analyser Mark IX-P. Third Avenue, New York.

21. Inbar, G.F. & G. Eden. Psychological Stress Evaluators: EMG correlation with voice tremor. Biol.
Cybernetics, 24, 165-167, 1976.

22. Gallagher, G. Psychological Stress Analyser. Protection of public figures symposium, Fort Belvoir,
Va., 1972.

23. Simonov, P.V., N.V. Frolov & V.L. Taubkin. Use of the invariant method of speech analysis to discern
the emotional state of announcers. Aviat. Space Environ. Ned., 46. 8, 1014-1016, 1975.

24. Luk'yanov, A.N. . N.V. Frolov. Signals of human operator state (Chapter 6: The Speech Signal). NASA
Tech. Translaticir 14o TT F-609, 1970.

25. Cannings, R., R.G. Borland, L.E. Hill & A.N. Nich,. n. Voice analysis and workload during the let-
down, approach and landing. Aerospace Med. Association meeting, Washington, 1979.

26. Anderson, T.W. An introduction to multivariate statistical analysis (Chapter 5). New York, Wiley,
1958.

27. White, G.M. & R.B. Neely. Speech recognition experiments with linear prediction, bandpass filtering
and dynamic programming. IEEE Trans. Acoust. Speech and Sig. Proc., ASSP-24. 2, 183-188, 1976.

28. Noll, A.M. Cepstrum pitch determination. J. acoust. Soc. Am., 41. 2, 293-309, l1C-.

29. olive, J.P. Automatic formant tracking by a Newton-Raphson technique. J. ac5ust. Soc. Am., 50. 2,
661-670, 1971.

30. McGonegal, C.A., L.R. Rabiner & A.E. Rosenberg. A Semi-Automatic Pitch Detector (SAPD). IEEE Trans.
Acoust. Speech and Sig. Proc., ASSP 23. 6, 570-574, 1975.

31. Makhoul, J. Linear prediction: A tutorial review. Proc. of the IEEE, 63, 561-580, 1975.

32. Markel, J.D., B.T. Oshika & A.H. Gray. Long term feature averaging for speaker recognition. IEEE
Trans. Acoust. Speech and Sig. Proc., ASSP 25. 4, 330-337, 1977.

33. Morrison, D.F. Nultivariate statistical methods (Chaptem7 and 8). New York, McGraw Hill, 1967.

34. Hunt, N.J., J.W. Yates & J.S. Bridle. Automatic speaker recognition for use over counication
channels. Proc. IEEE Int. Conference, ASSP 77. Hartford, Ct., 1977.
 129
AN EXPLORATORY STUDY OF PSYCHOPHYSIOLOGICAL MEASUREET

AS INDICATORS OF AIR TRAFFIC CONTROL SECTOR WOR•LOAD

by

Richard E. McKenzie, Ph.D.

Edward P. Buckley, Ph.D.
Kiriako Sarlanis, M.A.
Federal Aviation Agency
Natiunal Aviation Facilitee Experimental Center,
Atlantic City, New Jersey

INTRODUCTION

There has been an ever-increasing concern within the Federal Aviation Agency for the possible adverse
effects of stress inherent in the character of the work of Air Traffic Control Specialists (ATCS). Some-
one has characterized the lob of the pilot as involving "hours of routine monotony interspersed by moments
of sheer terror." Perhaps this is no less true of the job of the controller whose basic task is to main-
tain an orderly flow of air traffic, maintain the safe separation of enroute and coverging terminal
traffic, and to assist the pilot, often under adverse flying conditions.

As pointed out by Dougherty, Trites, and Dille, who compared health information between ATSC and
non-ATSC personnel, those who are engaged in this particular occupation, as well as external observers of
the job situation, feel that there ir inherent stress involved in the work which may have adverse effects
(1). These effects undoubtedly involve internal stress factors such as fatigue, aging, and job experience,
as well as external factors of potential aircraft conflict, workload, critical incidents, and other
aerospace events. The major concern of human engineering has been to develop command and control systems
wherein better displays and more functional controls would enable the controller to better perform his
demanding task and ultimately render it less stressful. Basic to this concern has been an attempt to
define the controller's task and to identify certain aerospace events, such as number of aircraft, aircraft
speed, control sector size, etc., which may be c.ucial factors in the controller's job performance (2, 3).
However, such studies have served only to point out that the real need i" evaluating the efficiency of
control systems, or of the operator himself, is the establishment of relevant criterion measures. Studies
in this area, to date, have demonstrated that simple measures of various aerospace events which comprise
the controller's workload do not fully relate to the complex stresses that are experienced in the job
performance.

Since external job-related measures do not offer satisfactory criteria, we have turned to internal
operator-related measures in an effort to determine their usefulness in evaluating the stressors inherent
in the work of the ATCS. Therefore, this study was designed to explore the possibility that certain
physiological measures could be related to some aspects of the contr,>ller's task, namely, workle'd defined
in terms of number of aircraft (traffic density), and the occurrence of aircraft conflicts.

PROCEDURE

Oti=nu.us Materials, Since the research goal was to determine if selected physiological variables were
related to controJ- workload, the stimulus materials were selected to provide two extremes of work level.
This was done by simulating a PPI (Plan Position Indicator) display cf an enroute sector by means of
specials films using the "CODE" (Controller Deicision Evualtion) technqiue developed by one of the authors
(4). One film presented a traffic pattern of low density, i.e., few nircraft and few conflicts. The other
film presented a high density traffic pattern, i.e., many aircraft, muy more conflicts, and higher
aircraft speeds. An aircraft conflict is defined in terms cf aircraft in flight that approach each other
in such a manner so as to violate established separation criteria. This is a potential collision situation.

The problems were approximately 40 minutes in duration. The low density sample had an average of 6.6
aircraft under control at a givin time. The average aircraft speed was 470 knots with a range of 380
knots to 550 knots. The number of conflictions occurring during the problem was four.

The high density sample had an average oi 19.4 aircraft under control at a given time. The average
aircraft speed was 476 knots with a range of 346 knots to 566 knots. The number of conflictions occuring
during the problem was 16.

Subjects: Ten subjects, all Air Traffic Control Specialists, were selected. Their chronological age
range was from 29 to 46 years with a mean of 35.7. Their experience as controllers ranged from 7 to 18
years with a mean of 11.5.

Instructions: In crder to standardize the subject's approach to the experimental task, the following
instructions were read to each man individually:

You are being asked today to take part in an experiment cancerning controller
workload and certain related physical changes. We have developed a film showing
traffic on an enroute sector. The picture will change every six seconda to give
an approximation of a six second sweep on a PPI scope. Some 3f the aircraft
develop conflictions in the film. Your task will be to discern when conflictions
are developing and to indicate this fact by pressing the button on your left and
noting the identities of the aircraft in confliction on the sheet before you.
The number of aircraft you will be responsible for may be unusually high and be
going unusually fast--but do what you can. There are two sectors displayed. You
are responsible for the conflict detection task for both sectors. Your separation
standard is 5 miles and 1,000 feet for the total area. YOUR PERFORMANCE IS SCORED
ON HOW ACCURATELY YOU PERFORM. Avoid reporting too soon as this may be only a
potential conflict. Avoid reporting too late--prevention action could not be
30

taken. Both of theme factors will be considered errors in determining your score.
The records of your performance are for research purposes only and will not be
divulged to anyone for any other purpose.

Note that you are asked for two things: (a) to press the button on your left at
the time which you would normally instruct one of the pilots to take preventive
action, (b) note the identities of the aircraft involved.

All aircraft in the system are identified by alphanumerics. Information on

altitude and changes in altitude is also given. Frame 6 (sample display) shows
how this information will be given throughout this fili. The scale of the map
is in the upper left.

Are there any questions?

Test Schedule: All subjects were scheduled for one session during which they monitored the display and
performed the required task of conflict identification for both the high and low dengity traffic patterns.
The order of presentation of either the high or low density stimulus was counterbalanced across subjects
to rule out order effects. Each traffic pattern was monitored for 40 minutes. Thus, the subjects were
instructed, instrumented, calibrated, and then monitored either the high or low film for 40 minutes. Then
they had a 10-minute break during which time the film was changed. Then followed another 40-minute seassln
using the alternate film.
Physiological Measurements: Two physiologic measures were selected as dependent variables: Heart Rate
and Psychogalvanic Skin Response, also called the Galvanic Skin Response (GSR). Heart rate was recorded
via two electrodes attached at conventional locations on the chest approximately 6.5 cm above and below
the left nipple on the mid-clavicular line.

GSR was recorded from electrodes attached on each hand in the fleshy area commonly called the "heel."
Anatomatically, this area can be described as laying over the fourth metacarpal bone and about 2 cm. to
the ulnar side of the palmer aponeurosis. Prior experience has demonstrated that excellent GSR responses
can be obtained at this site without the movement artifacts usually associated with a central palmar
location. In fact, our subjects were above to write without producing noticable artifacts.

The electrodes were of local manufacture using silver-silver chloride material 1A cm. in diameter
and mounted in a plastic cup measuring approximately 3 cm. in diameter. The electrolyte used was EKG-SOL
(Beck-Lee Corporation). The sites were prepared by sponging with acetone and the electrodes, filled with
the electrolyte, were attached by Eastman 910 adhesive.

Leads from this sensor were fed into appropriate couplers of an E. and H. Physiograph. Both heart
rate and GSR were condenser-coupled. Read-out was continuous at 120 me. per minute.

Subject calibration for GSR determinations was done by the "sniff" method. Prior to starting the
experimental run, the subjects were requested to sniff (a rapid inhalation through the nose) at k minute
intervals, the GSR amplifier was adjusted to yield a 20 mm. per excursion. When the subject's "sniff"
response stabilized at this level the session began. In order to maintain the GSR at a "standard" level
throughout the experimental session, the subjects were requested to give a sniff response every minute.
By this method, each subject's GSR could be calibrated and maintained throughout the experiment.

Data Reduction: Heart rate was scored on a minute-to-minute basis and an average determined for each
experimental session. GSR was evaluated in two ways. First, total GSR frequency was determined in terms
of a "standard unit." This standard unit was arbitrarily selected as a 5 x 5 me. square area. All GSR
responses which were less than this size area were ignored. The records were then hand scored in terms of
this standard unit. In order to check on the accuracy of the hand scoring, GSR's from each record were
randomly selected and the area scored by means of a planimeter (reuffel and Easer 4211). By this means
the hand scoring was determined to be in error less than one percent. Hand scorinR for area is a laborious
procedure but, in the absence of an qlectronic integrator, reasonable accuracy can be obtained, although
we are not advocating the procedure.

Statistical Analyses: The means and medians for the high and low density situations were computed for
all four measures. Both Sirametric and non-parametric tests for the statistical significance of differences
were done for all four measures. The matched pairs 't' test was the virametric test used. The arcsin
transformation was used before applying the 't' test to the percentages of confliction detection. The sign
test was the non-parametric test used. One tailed tests were used in all cases.

RESULTS

Establishment of Difference Between Traffic Samples: The confliction detection performance is shown in
Table I. That there was a very significant difference in the two traffic samples shown is indicated in
the significant difference in the confliction detection performance. This establishes the fact that we
were in fact dealing with traffic sampler which were markedly different in difficulty.
The fact that some confliction detections were missed in the heavier traffic sample shows that it
was probably unrealistically difficult, in accordance with the rationale stated earlier for this pilot-
study, of using two markedly different conditions to examine the physiological measures. In addition,
the fact that there were considerable overwrites among the alphanumerics on the film undoubtedly markedly
increased the difficulty of detections. In general, then, the reader is cautioned not to regard these
percentages of conflict detections as operationally valid but rather to keep in mind the rationale for
this test as an exploratory study.
 131

S~TABLE I

• CONFLICTION DFTECTION IN LOW TRAFFIC DENSITY AND

-,' HIGH TRAFFIC DENSITY SITUATION FILMS

Subject Low Density )HighDes:ty (2)

1 100 56

2 100 75

3 100 62

4 100 44

5 100 75

6 75 50

7 100 69

8 100 69

9 75 69

10 100 69

Mean 95 64

Median 95 69

Sign test p - .001 t (transformed data) p - .0005

t (untransformed data) p - .0005

Evaluation of Physiological Measures: The major purpose of the study was to see whether certain physio-
logical discriminated between two traffic samples of different difficulty, i.e., whether they would
reflect workload.

Tables II, III, and IV, present the data from three measures: Heart rate, GE. frequency, and GSR
area, respectively. It is clear from the tables that the best discriminator is GSR area; that the GSR
frequency measure is of moderate definitiveness as a discriminator; and that heart rate is least effective,
although discriminating.

TABLE II

HEART RATE DURING LOW TRAFFIC DENSITY AND HIGH

TRAFFIC DENSITY SITUATION FILMS

Subject Low Density High D l.lty

1 96.68 99.35

2 72.48 7J.39

3 90.88 1)9-37

4 90.05 85.64

5 99.40 105.23

6 105.88 96.58

7 63.49 75.51

8 79.47 79.86

9 65.03 67.66

10 88.09 102.99

Mean 85.14 89.56

Median 89.07 91.11

Sign test p - .055 t test p - .10 ---- .05

 132

TABLE III

GSR FREQUENCY DURING LOW TRAFFIC DENSITY AND HIGH

TRAFFIC DENSITY SITUATION FILMS

SubjecL Low Density High Density

1 107 110

2 64 72

3 36 75

4 46 112
5 51 99

6 77 76

7 59 82

8 109 92

9 51 97

10 11 128
Mean 67.1 94.3

Median 61.5 94.5

Sign test p - .055 t test p - .01 ---- .005

TABLE IV

GSR AREA DURING LOW TRAFFIC DENSITY AND HIGH

TRAFFIC DENSITY SITUATION FILMS

Subject Lcw Density High Density

1 198 433
2 134 279
3 119 207

4 88 296

5 146 199

6 221 478
7 150 206

8 197 234

9 164 272

10 196 324

Mean 161.3 292.8

Median 157.3 275.5

Sign test p - .001 t teat p - .0005

This pilot study, then, within its limitations, has accomplished its purpose of determining whether
extensive examination of this type of measure was warranted since it has shown that psychophysiological
i i!widely meaaurements could at least discriminate the human reaction to traffic situation which were known to differ
in difficulty.

DISCUSSION AND CONCLUSIONS

The resder should be aware, on the one hand, that the traffic situations portrayed here were different
in ditficulty in the extreme and that these physiological stress measures may not be as successful in
discriminating the human effort differences associatLd with more normal sector-to-sector or hour-to-hour
variations in traffic. On the other hand, if further studies confirm the results obtained here, a tool
for systems rasearch and development of considerable importance has been found. As only one example of
this utility, this methodology could be useful to verify, refine, and improve the recent formulation by
 133

Arad of a mathematical index of the complexity of airspace events (2, 3). Another obvious use is as a
criterion for new systems which may have, as one of their values, a reduction in controller effort,
fatigue, and stress.

The study has had another outcome, important in the technology of psychophysiological measurement.
As previously noted, GSR changes in the subjects were more detectable using variations in measured
amplitude area, as compared to frequency of GSR changes. While our methods of evaluating GSR area were
laborious due to equipment limitations, the availability of integrating methods for automatically yielding
measures of amplitude change should yield important data in evaluating the GSR relative to workload and
other stress studies. In this, our work appears to closely parallel a Russian study (5) where
Kozarovitskii reports: "In inexperienced airline dispatchers the galvanic skin responses deviated t.:a
normal due to fatigue at the end of a working day or under tension. The character of the tracings showed
diminishing skin resistance, either an increase or a decrease in amplitude in different individuals, and
a decrease in frequency. Distraction of attention tended to lower the skin resistance %hich depends
(on) stimulating and suppressing processes." He also points out that, "Experienced subjects shoved fewer
sharp variations, less fluctuation in amplitude, a faster fading of the response, and leas reaction to
the extraneous stimuli."

While this study was of limited scope, it seems that the study of physiological parameters of ATCS
workload may yield important criterion measures of external factors of aircraft conflict, task overload,
critical incidents, and other aerospace events. Future studies will explore other physiological
variables for use as criterion measures, as well as tools for the evaluation of internal stress factors
in relation to fatigue, aging, and traffic control experience.

REFERENCES

1. Dougherty, J. D. et. al. Self-Reported Stress-Related Symptoms Among Ai.- Traffic Control
Specialists (ATCS) and Non-ATCS Personnel, Aerosp. Med., October 1965.

2. Ared. B., Mayfield, C. E. et. al. Control Load, Control Capacity, and Optimal Sector Design,
Franklin Institue Laboratories, Philadelphia, PA and Systems Research and Development Service,
Federal Aviation Agency, Atlantic City, NJ, Report No. RD-64-16, December 1963.

3. Jolitz, G. D. Evaluation of a Mathematical Model for Computing Control Load at Air Traffic Control
Fa.ilities, Systems Research and Development Service, Federal Aviation Agency, Atlantic City, NJ,
Report No, RD-65-69, June 1965.

4. Buckley, E. P. et. al. Pilot Experiments Concerning Air Traffic Control Decision Making, The
Franklin Institute, Philadelphia, PA, April 1960.

5. Kozarovitokii, L. B. Dynamics of Skin-Galvanic Reactions in Control Panel Operat..rs During Their

Work of Regulating Airplane Traffic, Moscow University, Department of Physiology, Moscow, USSF,
Zhurnal Vysshei Nervnoi Deratel' nosti, Vol. 14, May-June 1964, pp. 387-396 (In Ruesian).

S -
 f4

I
I

I
 135

INDIVIDUAL AND SYSTIL( PERFORMANCE INDICES

FOR THE AIR TRAFFIC CONTROL SYSTEM

by

Edward P. Buckley, Ph.D.*

William F. O'Connor, Ph.D.
Tom 9**b*
Federal Aviction Agency
National Aviation Facilities Experimental Center,
Atlantic City, New Jorsey

This in a simulation study to examine the relationships between field air traffic controller perfor-
mance indices and system performAnce measures. The study encompassed performance criteria developed
within two distinct environments, the controller's home facility where he controlled live traffic, and
a specially designed microsystem or "one-iman ATC system" with simulated traffic. This microsystem
simulation was done at the National Aviation Facilities Experimsental Center. Thus, the experiment
represented a comparative examination of several quantitative measures of system functioning derived from
air traffic control simulation and an investigation of these measures as indices for the objective
evaluation of the individual air traffic controller.

The initial impetus for this study arose from a concern over the relationship of age and experience
to controller proficiency. To a large number of controllers in the field, there appeared a definite
trend for older men to be unable to adequately handle the new and complex demands related to the increasing
pace in the air traffic control system. While this question of age versus proficiency formed the initial
experiment, the basic interest lies in the necessity of the FAA to maintain a highly competent workforce
of air traffic control specialists. The central assumption is, that in order to evaluate the effects of
age or any other variable upon air-traffic controller performance, we must first develops and validate
an objective and reliable criterion of performance that has known relationships with controller task
functions. Until the establiahment of such criteria any question such as the matter of age versus
proficiency could only be appraised in terms of indirect or anecdotal measures. Another aspect of this
study was that simulation which had been designed anid utilized to study procedural and syst~m differences
had never been employed for the assessment of performance differences associated with the individual
controller. Through the use of simulation in this manner ii technique could be developed which would permit
the evaluation of each controller operating his test sector as a "micro ATC system" and utilize the system
to measure varying system load levels and related controller behavior/efficiency.

Thirty six (36) journeymen enroute air traffic controllers served as subjects having been chosen as
a randomized stratified sample of the personnel from four enroute air traffic control centers. The
controllers were brought to the NAFEC center for one week in groups 4t four to receive an orientationI
to the simulation task including the fictitious geographical area which was to be simulated and the
traffic control local procedural rules which were to be in.effect in this sector. Each subject performed
traffic control during six one-hour runs. with two runs at each of three traffic densities, an experimental
protocol designed to produce scores which would measure individual rather thian team performance. In
addition, each subject was tested on an abbreviated simulation method called CODE, which stands for
each subject was tested on an abbreviated simulation method called CODE, which stands for Controller
Decision Evaluation. During the main simulation, various performance measures derived from counts or
timing of events, for example, the number of aircraft delayed, the delay time, and so forth, were taken.
In addition to these system performance measures, stress sensitive measurements of physiological functions
were obtained under this dynamic simulation. In addition to the physiological variables of heart rate
and galvanic skin response (GSR) measurements, a number of psychological measures were obtained through
the use of the 16 PF Test.

The conclusions based upon the results of this project are as follows:
1. The current chronological age of the 36 subjects ranging in age from 31 to 45 years possessed
weak, negative relationships with indices of controller proficiency In both field ratings and
simulation performanc'o measurements. Age alone, within the range studied, is not a very guod
performance predictor.

2. Controller age, modified in various ways by experienc.j, does have some effect on performance.
This age effect operates in the direction of greater caution and safety, with tendency toward
delay of traffic. However, there are wide individual differences within age groups and con-
siderable overlap in proficiency indices between age groups.

3. Current age and age at entrance on duty were highly correlated in this journeymen level group.
At the journeyman level, differences apparently due to current age many in fact be due to
age at entrance on duty.

4. The personality scale scores based on the 16 PP scales have an unusually large number of
statistically significant relationships wi :h controller performance. This suggests that the
controllers task which requires sustained performance under complex circumstances, makes such
stressful demands as to involve his total personality as well as his skills. The use of these
scores so predictors of controller efficiency is not validated. The 16 PF tests reflected that
superior controller performance might be linked with the following characteristics: freedom
from depression, lack of timidity, socially realistic and relaxed with a relative absence of

4 tenseness or anxiety.

*Abstracted with permission of the senior author from final report No. NA 69-40, Federal Aviation
Administration, September 1969, by Richard E. McKenzie, Ph.D.
136

S. In terms of physiological measuremnto, the controller's heart rate rises in a manner

corresponding to increases in the level of traffic and the heart rate levels observed are probably
indicative of the stress inducing nature of his task. Unfortunately, the GSR measures did not
correspond to the previously reported exploratory study, however, this may be due to the method
of evaluation of to the method of GSR collection which was obtained by foot electrodes as opposed
to palmer surface electrodes.

6. Simulation system performance measures are reliable and sufficiently precise to measure
individ~dl differences in controller proficiuncy.

7. Simulation measures and field indices of controller performance possess sufficient overlap to
establish a meaningful correspondence between the simulation test environment and the tive
traffic environment. Stimulation technology, then, is capable of providing reliable and
objective measurements of controller proficiency.

8. Part measures of the controllers task using the Z•DE technique appears to be a good objective
measure of certain fundamental controller abilities which warrants further development.
9. Results from factor analysis indicate that nine speci.fic system performance criteria are
sufficient to describe system functioning over the rttagc of traffic studied.

10. An index which represents the quantification of a trade-off function between volume handling
capacity end the occurrnce of delayG to aircraft cnlled Rdv offers promise as a measure of
system load. This index appears suitable for utilizaaton in both the live and simulation
system environments for assessing workload differences associAted with various sector configu-
rations, staffing patterns, and different feographical artal. The index Rdv is mathematically
defined as the correlation between delays and volume in terms of nuwber of aircraft handled at
a given traffic level.

Tb. •chut the history of attempts to evaluate individual and mystom performance, factors simulation
techniqý nd psychologizal tests have not always proven effective. In this study we see thAt a
simulatio, ýystem can yield measures of controlleT proficiency and tOat at least sow psychological test
scores can depict superior controller characteristics. Since scores ov, the 16 PF scales were correlated
with both the simulation system performance meatures and the firir rating measures, further exploration
of this test as a predictor of successful controller qualities or as a posse.ble method of evaluating
decrements in the performance of career air traffic controllers sefum itid~cated.

' it

p|

S"_
 137

WORKLOAD AND STRESS IN AIR. TRAGIC CONTROLLERS

DY

Carl 3. Melton
Aviation Physiology Laboratory
Civil Aeromedical Institute
FAA Aeronautical Center
Oklahoma City OK 73125

Abstract

DatA nollected at 14 air traffic control facilities regarding air traffic controller (ATCS) workload
and urizary stress indicator hormone (SIHE ..xcretion is reviewed. The data sho% a significant relationship
betweevi objective workload measures (radio transmission time and traffic counts) and indexes of cate-
cholamin., excretion. Mean epinephrine excretion by ATCSs at six air traffic control towers ranging from
very low to very high traffic dsnsity was significantly (R - 0.96) related to annual tLaffic counts at
those towers. The sympatho-adrenomedullary axis that prepareL the organism for "figh., or flight" described
by W. U. Cannon in 1929 apparently is applicable to ATCSe. The q,,estion of underload, optimum lood, and
overload is discussed.

I. Introduction. The workload experienced by air traffic cont-,,ir-a (ATCS) is difficult to define.
One may cousider imposed load objectively in terms of numbers of 1rcaft hrndled, but the subjective load
perceived by the controller may be a greatly different quantity.

Many factors may operate as workload modifiers either making the v-.k easier or more difficult:
(1) Type of traffic handled. One aircratt in distress may cause more "work" than all the other traffic
being handled. (2) Weather. Controllers' perceived workload always iiwveases when pilots cannot maintain
visual separation in instruments meteorological conditions. (3) Equipment outages and malfunctions causing
reversion to manual methods .f control. (4) Disruption of circadicn rhythms -aused by rotating shifts,
and (5) General physical and eb-otional condition resulting from a variety of off-duty activities and on--
duty problems with management or peers.

It is perceived workload that gives rise tc the poorly-defined entity known as stress. Excessive
stress has generally been assumed to be a component of air traffic control work and has been legally
recognized as such in Public Law 92-297 which providos full retirement for controllers over 50 years of
age after 20 years of work controlling air traffic.

Estimates of stress in ATCSs are rendered difficult because of the interaction of off. d%. and on-
duty experiences. The ATCS undoubtedly brings off-duty problems to work with him and, just as certainly,
takes home with him concerns connected with the work place. Thus, a complete representation of strens hI
ATCSs must integrate all aspects of the ATCS's life.

There has been a strohng tendency in the popular press to describe stress in ATCSs in terms of con-
ditions at "hot spot" facilities such as O'Hare and Atlanta Air Traffic Control Towers (ATCT). General-
izations from these descc ptions give a skewed idea about stress in the entire population of ATCSs, most
of whom work in facilities with far fewor operations.

For the last .0 years this laboratory bar carried out studies aimed at providing a general description
of stress in ATCSs. These studies have encompassed several variables including numbers of air traffic
operations, shift rotation effects, automation, different kinds of air traffic control (ATC) work, and
geographical distribution. This report represents an attempt to provide a general concept of stress and
workload in ATCSs.

II. Methods. Estimates of stress were derived primarily from urine biochemical analysis for 17-ketogenic
steroids (17-KGS), epinephrine (E), and norepinephrine (NE). In most cases values for these stress
indicator hormones (SIH) are expressed as creatinine (CR)-based ratios (Pt SIH/100 mg CR). Urine analysis
for SIH was carried out as previously described (1). Urine collected at field facilities was frozen at
the work site; when a sufficient number of specimens had accumulated, they were shipped to the Civil Aero-
medical Institute. (CAMI) by air freight. Upon receipt the specimens were placed in a freezer where they
ware kept until analyzed. Specimens were in transit for 3-5 h and detectable thawing did not occur.

Subjects were all volunteer male air traffic control specialists. They were 'nstructed to void and
discard urine just prior to retiring the night before a workday. They were then to collect all urine
voided ontil they arose; normally there was only one voiding end that one ipon arising. They were then
instructed not to collect urine until they arrived at work. At work they were told again to void and
discard and to collect in one container (or two, if that one became full) all urine subsequently voided
during the workday. ATCSs theik repeated this collection regimen for various periods of time, depending
on the facility being studied. In some studies urine was collected for a whole 5-day workweek; in others,
urine wrs collected for 2 days only because of changeable shift patterns. Each 24-h rest-work period was
represeated by two specimens.

Urine was collected in cuboidal, plastic 1-quart receptacles containing an excess of dry boric acid
as a preservative. When ATCSs delivered the containers to the technical crew the containers were labeled,
logged, and frozen.

Workloads were estimated in two ways. One involved the recording of all radio trans*_a1ono received
by and coming from the subject ATCS. From these recordings total radio transmission t1w'z (RTT) was derived
by use of voice-actuated relays and digital counters. Workload was also derived from traffic counts.
 In order to amalgamate a large volume of biochemical data, a stress index (Cs) was formulated. The
details of the index have been published (2). Briefly, the _adex is based on the idea that the product
of the resting and working values for each SIH gives a more realistic view of stress than does the
excretion increment (or decrement) from rest to work. However, because the SIHs appear in such unequal
quantities in the urine, each individual mean thL.t value (rest and work) is adjusted by dividing it by
r a grand mean derived for that SIH from all Lhe measnrements made on ATCSs in all past studies in this
laboratory. This adjustment causes all SIHs to assume equal importance in the calculation of Co. Cg is
the average of indexes calculated for each of the SIHs, cat (17-KCS), c. (E), and Cne (NE). This index
allows different controllers and facilities to be readily compared.

The indexes for each of the SIHs can be presented diagraatically to show composite stress and the
relative contributions of each SIH thereto. The diagram is based on the theorem that the sum of the
lengths of internal lines emanating from a comon point and perpendicular to the sides of an equilateral
triangle is equal to the altitude of the triangle (3,4). The values for cat, ce, and cne can be repres-
ented as lines originating at a common point and diverging at angles of 120, the lengths of which are
proportion.l to the values of cat, ca, and cne. Lines drewn perpendicular to the free ends of the
diverging lines form an equilateral triangle, the area of which is proportional to Cs, the average of
cto, ce, and cne.

III. Results. Field Experiments. Table 1 shows the correlation between Cs, cst, ce, cne, and RTr at
Opa Locks (OFF) Air Traffic Control Tower (ATCT) located on a very busy general aviation airport in
Greater Hiami, Florida. It is apparent that RTT is significantly related to ce and, with less signifi-
cance, to cne. RTT is not significantly related to cat.

Figures 14 show graphically the relationship between stress indexes and RTT at OPF.

Between 1972 and 1974, Los jingeles (LAX) and Bay Area (Oakland (OAK)) Terminal Radar Approach
Control (TRACON) facilities were given automated radar terminal system equipment (ARTS-Ill). This equip-
Sament displays aircraft identification ant altitude oxithe radar cathode ray tube, and thus contributes
greatly to safaey. This equipment was also expected to reduce significantly the workload of radar
controllers. There were other changes associated with the new equipment, also. The TRACONs were moved
from thair towers to separate buildings with adjacent parking lots; the dress code was relaxed; lounge
facilities and the general work environment in the control rooms were greatly improved.

Studies were carried out at these two TRACONs prior to (1972) and after (1974) installation of
! ARTS-III. Table 2 shown changes in stress indexes for individual controllers before and after ARTS-III
installation. It is ciear that there was a uniform drop in 17KGS and an increase in catecholamine
'
Sexcretion.
Table 3.
The workload in terms of number of aircraft worked and number of radio contacts is shown in

The traffic count at LAX increased by 3 percent and at OA" by 4 percent from 1972 to 1974. The
number of radio ,intact# increased by 3 percent at LAX and by 1 percent st OAK, while composite stress
increased by 21 i,4rcent at LAX and 20 percent at OAK. This increase can be seen diagrammatically in
Figure 5. Thus, the disproportionate increase in stress, entirely due to elevated catecholamine excre-
tion, is not explained by che objective workload. The explanation most likely lies in work elements not
reflected in traffic counts, KTT or number of radio contacts. The new TRACONs had been in use only
L about 5 months at the times of the second studies. There were still equipment difficulties to be
worked out; outages were fairly frequent. Controllers liked the reduction in coordination with other
facilities nod within the TRACONs; however, the concensue among the controllers was that ARTS-Ill had
not reduced and, in fact, had increased thz total workload, primarily because of unfamiliarity with the
new equipment.

Catecholamine Excretion and Traffic Count. Because of the demonstrated relationship between E
excretion and RTT, annual traffic counts for ATCTs where studies have been conducted were graphed
against mean E excretion for controllers at those towers. Such a graph ýs shown in Figure 6 where
annual traffic count is plotted against mean working and resting E excretion for ATCSs at ATCTs ranging
from low to high density. The relationship between traffic count and mean E excretion is significant
(R a 0.96). The working value for O'Hare ATCSs has been displaced tc the left to reflect the fact that
ORD ATCT was effectively operated as two facilities with separate control positions for the north and
south aides of the airport; one side wos customarily used for departures and the other side for arrivals.
The workload impinging on each ATCS was tbus about half of the total airport traffic. When the data
point Is roved to reflect this division of work, it falls near the line of best fit.

Laboratory Experiments. Bocatse reallitt stress arises from a mixture of streseors, the physio-
logical responses to those streseota are difficult to interpret. One cannot separate off-duty exFeri-
ences from workrelated factors. Therefore, an attempt was made to expose paid experimental subjects to
"pure" stressors in the laboratory in order to delineate the FpecifiLity of the hormonal response,
should there be such.

The subjects (10 young men) were each exposed to a purely physical task with no competitive element
(treadmill, 3 talles per hour with no grade) and, on another date, to a purely competitive- but nonphysical
task ("Pong," a video game based on pingpong). One of the researchers acted as opponent for all subjects;
site was an expert and was rarely beaten. Order of presentation of the tasks was balanced; each task
was presented in 50-inin episodes. In the 10- min following each episode, urine collections were made,
rest was allowed and water was imbibed to replace tha urinary loss. Urine was analyzed for 17KGS, E,
and NE. Values are expressed as the total quantity of each SIH excreted during each 50-min episode.
The schedule in ?.ach instance vas maintained for 2 h

Prior to either experimental exposure, each subject rested for 50-mtn in the supine position on a
cot, ..CG electrodes having been previously attached for registration of ambulatory heartrate on small
battery-operated ECG tape recorders.
 139

The results of urine and heart rate analyses are shown in Tables 4 and 5. Corresponding episodes
of the two tests did not cause significantly different excretion levels in urinary metabolites. Heart-
-.. werestatistically
ateawas significantlysignificant
higherS~tion
for for
treadmill
the Ponethan for whereas
Pong tasks. last-to-work
task, the difference was difference in E excre-
not significant for the
Streadmill
task. Rest-to-work differences in excretion of NE erd 17-KGS were not statistically significant
for either task.

IV. Discussioa. Data collected over 10 years from several ATC facilities point to catecholamines,
principally E, as being a good indicator of the response to an applied workload. The adrenal cortical
response (17-KS) primarily indicates chronic stress arising from unresolved conflicts such as labor
management disputes, marital difficulties, financial problems, etc. NE and E usually go in the aeme
direction-seldom does one go up and the other down. It is always difficult to separate the so-called
physical and mental stressors in a field study setting. Our laboratory studies indicate strongly that
mental activity without significant physical effort engenders a significant output of E above :.:he resting
state that does not occur during episodes of purely physical effort.

It thus appears that there is a degree of stressor-response specificity. The large unanswered
question relates to the significance of the magnitude of the response. When is a person underloaded,
optimally loaded and overloaded? It is also clear that ATCSs at ATCTs with low traffic density have a
low E output while ATCSs at high density ATCTs have a relatively greater E output. Obviously, each group
of ATCSs is doing at least an adequate job and one cannot say on that basis whether overload, optimal
load, or underload is present. However, it is distinctly possible, in view of the known effects of
catecholamines on the cardiovascular system, that the cost of adequate- performance is greater for the
ATCSs at high density facilities than it is for ATCSs at low density ATCTs. Rose (5) has recently shown,
as have others in the past (6), that hypertension is more prevalent among ATCSs than among the general
population. We have shown that E excretion level is significantly and directly related to heartrate (7).
Further, we have shown in a limited group of ATCSs that elevated HE excretion is predictive of later
hypertension (8). These data suggest that the cost of adequate performance at a high traffic density
ATC facility may result in breakdown of physiological systems.

At the other end of the workload spectrum it is obvious that adequate performance can be maintained
without great arousal brought on by high blood levels of cat-cholamines. In short, it appears probable
that arousal necessary to meet workload demand is mediated by sympathoadrenal output of catecholamines.
This idea was first put forward by Cannon in 1929 in his descriptiorn of the "fight or flight" reaction
(9) and is applicable to the ATC task.

These data also are consistent with the data reported by Schaad, Gilgen, and Grandjean who showed
a statistically significant relationship between urinary catecholamine excretion and level of difficulty
of work in European air traffic controllers (10).

REFERENCES

1. Melton, C. E., J. M. McKenzie, B. D. Polis, H. Hoffmann, and J. T. Saldiver, Jr. Physiologicel

Response in Air Traffic Control Personnel: Houston Intercontinental Tower. FMA Office of Aviation
Medicine Report No. FAA-AM-73-21, 1973.
2. Melton, C. E., J. M. McKenzie, J. T. Saldivar, and S. M. Hoffmann. Comparison of Ops Locke Tower
With Other ATC Facilities by Means of a Biochemical Stress Index. FAA Office of Aviation Medicine
Report No. FAA-AM-74-11, 1974.

3. Streng, 0. Eine Volkerkarte. Eine graphische Darstellung der bisherigen Isoagglutinationsresultate.

Acts Soc. V¶ed. Fenn. Duodecim., Vol 8, pp. 1-17, 1927.

4. Streang, 0. Das Isoagglutinationsphanomen vom anthropologischen Gesichtspunkt aus. Acts Pathol. et

Microbiol. Scan., Suppl. 5, pp. 59-62, 1930.

5. Rose, R. H., C. D. Jenkins, and M. W. Hurst: Air Traffic Controller Health Change Study. A report
to the Federal Aviation Administration on research performed under Contract No. DOT-FA73WA-3211
awarded to Boston University.

6. Dougherty, J.D.: Cardiovascular Findings in Air Traffic Controllers. Aerosp. Med., 38:26-30, 1967.

7. Melton, C. E., 1. H. McKenzie, J. T. Saldivar, and H. Hoffmann: Studies on Stress in Aviation

Personnel: Analysis and Presentation of Data Derived from a Battery of Measurements. Advisory
Group for Aerospace Research and Development, AGARD-CP-180, 1975.

8. Higgins, E. A., H. T. Lategola, and C. E. Melton: Three Reports Relevant to Stress in Aviation
Personnel: I. Development of the Aviation Stress Protocol--Simulation and Performance, Physiological
and Biochemical Monitoring System: Phase I. II. Assessment of Cardiovascular Function After
Exposure to the Aviation Stress Protocol--Staulation. III. The Relationship Between Stress-Related
Netabolites 3nd Disqualifying Pathulogy in Air Traffic Control Personnel. FAA Office of Aviation
Medicine Report No. FAA-AM-78-5, 1978.

9. Cannon, W. B.: Bodily Changes in Pain, Hunger, Fear and Rage. C. T. Branford, Boston, MA., 1929.

10. Schaad, R., A. Gilgen, and E. Grandjean: Excretion of Catecholamines in Air Traffic Control
Personnel. Schweizerische Medizinische Wochenschrift (Swiss Ifedical Weekly), 99:889-892, 1969.

L. t.-- ---- ~____

140

TABLE 1. Correlation Coefficients of Stress Indexes and RTT at OFF

INDEX RTT

C 0.64*

:t 0.19

ce 0.77**

c ne O.55*

* P < 0.05

** P < 0.01

TABLE 2. Stress Indices From Individual Controllers

Before and After ARTS-III Installation

CS Cat Ce Cne

LAX 1 2 1 2 1 2 1 2

Subject

1 0.90 1.25 0.96 0.34 0.48 1.80 1.25 1.61

2 0.42 0.37 0.64 0.20 0.20 0.50 0.42 0.41
3 0.30 0.36 0.48 0.32 0.09 0.18 0.32 0.57
4 0.48 1.28 0.33 0.21 0.38 1.11 0.72 2.51
5 0.62 0.47 0.55 0.14 0.43 0.67 0.39 0.60
6 0.52 0.47 0.68 0.34 0 20 0.42 0.69 0.65
7 0.42 1.42 0.59 0.33 0,,9 0.59 0.38 3.42

OAK 1 2 1 2 1 2 1 2

Subject

1 0.45 0.68 0.58 0.51 0.53 0.99 0.24 0.54

2 0.73 0.72 0.50 0.27 1.21 1.17 0.50 0.72
3 0.57 0.47 0.56 0.21 0.49 0.59 0.67 0.60
4 0.35 0.71 0.31 0.29 0.29 1.56 0.44 0.28
5 0.46 0.69 0.90 0.37 0.23 1.08 0.25 0.63
6 0.42 0.86 0.59 0.43 0.24 1.27 0.43 0.88
7 0.27 1.36 0.20 0.21 0.28 2.83 0.34 1.05
8 0.18 0.57 1.17 0.05 0.15 1.01 0.22 0.15
9 0.40 0.84 0.73 0.15 0.28 1.81 0.20 0.56
10 C.33 0.47 0.61 0.16 0.22 1.03 0.17 0.22
11 0.22 0.74 0.31 0.10 0.16 1.33 0.20 0,79

TABLE 3. Number of Aircraft Worked and Number of Radio Contacts In

a 5-Day Workweek.

FACILITY NO. AIRCRAFT NO. CONTACTS CONTACTS/AIRCRAFT

LAX (1972) 1,803 13,806 7.66

LAX (1974) 1,860 14,210 7.64

OAK (1972) 1,190 8,712 7.32

OAK (1974) 1,233 8,827 7.13

.•I
 141

TABLE 4. Comparison of Excretion Values and Heartrates for Pong and Treadmill Tasks*

Total Amounts of Hormones Excreted

17-KCS E NE Heartrate
Task us ng na (Beats Per Minute)

Rest (Fong) 0.70 1,237 3,603 64

Rest (T-Mill) 0.67 1,214 4,274 64

P NS** NS NS NS

Pnng 1 0.70 1,619 3,809 73

T-Mill 1 0.59 1,741 4,384 101

P NS NS NS 0.05

Pong 2 0.62 1,720 3,379 73

T-Mill 2 0.67 1,463 3,813 100

P NS NS NS 0.05

Pong 3 0.59 1,750 3,833 70

T-Mill 3 0.58 1,491 3,581 98

P NS NS NS 0.01

A Group Averages
** T-test

TABLE 5. Statistical Significance of Rest-To-Work Differences for the Various Measurements*

Level of Significance of Difference Between Rest and Task (P**)

TASK 17-KCS E NE HEARTIATE

Pong 1 NS 0.01 NS NS

Fong 2 NS 0.01 NS NS

Pong 3 NS 0.05 NS NS

T-Mill 1 KS NS KS 0.01

T-Mill 2 No NS NS 0.01

T-Mill 3 NS NS NS 0.01

gee
S Table 4 for actual values.
•,Ps.rod, t-ten.e
 142

.2

I x
ai.i

.3

U DPI' L.O•KS "DSI4Ift

*4 X

WrLRKLnFMLDF <M I N>

FIGURE 1. Relationship between RTT and Cs at OPF ATCT. The relationship

is statistically significant (p < 0.05).

¼x
I V

K
. x
.3

.4
x
. x

WORILD <M I N>

FIGURE 2. Relationship between RTT and cat at OPF ATCT. The relationship is not
statistically significant.
 143

I.

I .3
.x x

...

.M ~~OlPmln
LDCkR •DIC"W

W R L DI < 'I N

FIGURE 3. Relationship between RTT and ce at OPF ATCT. The relationship is

statistically significant (p < 0.01).

2s.-

2.'
"Cm~W I m.r

Xx

x
.

0 WE on 7 so on lo

WEIRI'KL-F:S < M I N >

FIGURE 4. Relationship between RTT and cn. at OPF ATCT. The relationship is
statistically significant (p < 0.05).
Fv !44

STRING TRIANGLE REPRESENTATION OF CHANGES

OCCURRING IN *q6 ae. AND one AFTER INSTALLAIION
OF ARTS M AT OAKLAND AND LOS ANGELES TRACONS

OAK ('72) OAK ('74)

CS 0.60 0.72

LAX ('72) LAX ('74)

0.60 0.75

FIGURE 5. Diagramatic representation of the relationship between cgt, ce,

and cne on Streng's triangle. Coupar-lson of LAX and OAK TRACONs
before and after installation of ARTS-]II.

4
S3.0

•2.0.

S 1 .5 o S W 3

oo
S1.0,
0060

., .2 .3 .4 .5 .6 .7
ANNUAL TRAFFIC COUNT X 10e

FIGURE 6. Graph of annual traffic count (in millions of operations) vs. mean
urinary excretion levels of E of controllers at the various
facilities. Crosses represent on-duty excretion levels of E;
circles represent corresponding resting levels (ORD graphed at
actual traffic count and adjusted value (+) as explained in the text).
F'
•ASSESSMENT
S~by
CORRELATES OF WORKLOAD AND PER.FORMANC
'45

Richa-ýv E. McKenzie, Ph.D.

Crew ,echnology Division
USAF School of Aerospace Medicine (AFSC)
Brooks Air Force Base, Texas 78235
USA

INTRODUCTION

A few years ago Dean Chiles (1), while discussing objective methods for developing indices of pilot
workload talked about a hypothetical research vehicle having a system with the following capabilities:
(i) it would have an exact assignment of the nature and number of pilot duties that could be developed
for any given mission; (2) it would be possible to vary those duties in any combination over time; (3)
the control and display characteristics of the vehicle could be manipulated at will; (4) precise and
reliable quantitative indices of the task demands placed on the pilot by the system would be available
for all task elements; (5) precise and reliable quantitative measures of the skill with which the pilot
meets these demands would be available; and (6) an adequate criterion measure of total system performance,
would be available. If we had such a hypothetical vehicl, it is obvious that we would be able to deter-
mine the priorities that a pilot assigns certain tasks. We would also be able to assess the attention
demandd on the pilot by the system, and we would be able to determine which tasks or performance functions
are most sensitive to variations and total demand, we might be able to solve some of the problems we have
experienced in terms of the human being acting at certain optimal work rates, ignoring certain signals
from his display, working at how own pace, failing to pay attenzion to certain instruments or signals, but
in general, always managing to perform at a satisfactory enovgh level to complete the mission. Perhaps
a hypothetical vehicle being a realistic flying system would also help to weed out the variable caused
by non-flying laboratory tystems where the subject allowed to crash the system because of decremented
performance. Whereas in "real world" systems we find that the pilot werks harder and harder; performing
more and more control responses and movements, but the end result is usually to make a landing that he
can walk away from.

Chiles also pointed out that the first and foremost factor to keep in mind in choosing a methodology
for assessment is the purpose or goal of the research. Unfortunately, the entire histiry of assessing
workload, performance or stress in the human operator is one of compromise. We have to compromise because
of safety, because of operational requirements; we have to devise laboratory-type tasks because the real
thing is not available or is unasailable to our measures and often times we mupt rely upon human beings
other than pilots to perform these tasks because of the demands upon pilotapt. time in the real systems
world.

The assessment correlates of workload, performance, and stress can be divided into several areas:
those of physiological correlates, psychological correlates, stress correlates, psychophysiologic
correlates and finally central nervous system (CNS) correlates. 1.erealize that this is an artificial
taxonomy and that many areas of overlap exist; however, we thoutht we would arbitrarly subsume under the
heading psychologic correlates those tasks of vigilance, monitoring, tracking, reaction time, and so
forth. It should be noted that we're talking about the operat.ioual aspects of psychologic correlates,
that many of the psychological tests have been used as selection devices or as measures of skill in order
to predict successful training as a pilot or aircrew member. TheEe tests, for the most part, have shown
little relationship to the prediction of workload and performance abilities. Many of these tasks have had
the disadvantage of single operators looking at single displays and measuring single scores which are often
poorly related to the real world of the man-machine interface. Some of these measures are difficult, if
not impossible to measure in the operational environment. However, we should keep in mind that it may
not always be necessary to measure everything in the operational environment, that is, while the aircrew
member is piloting or doing his thing. It may be possible to ascertain whether he is capable of initiating
this particular flight-mission and it may be possible to ascertain upon his return from a mission the
amount of decrement that re3ulted by using relatively simple type, fie]' type tasks. Very often in pursuit
of psychologic variables we have resorted to the use of subjective eval-ation of such factors as fatigue,
stress, irritation, etc. Unfortunately, we are finding increasing evidence that certain aspects of these
factors may not be amendable to accurate subjective evaluation.

Some 20 years ago this author in his thesis study which had to do with the effect of binaural beats
upon performan:.e (2) found out there was a subjectively experienced quality of beats produced externally
in an audio mixer as compared with those auditory beats generated centrally in the central nervous system.
These two kinds of auditory beat phenomenons were preceived as differentially disruptive by the subjects.
In one experiment the subjects said they felt that the stimulus in neither session bothered them in any
way especially in terms of their performance on the tests that they were required to do. They ignored
the fact that they were not doing any better in consecutive performances and, in fact that they were
actually doing somewhat poorer. In another experiment, the subjects invariably reported that the sound
of an externally produced beat really bothered them and that they were sure that their performance was
effected. This was in spite of the fact that they could observe that they were completing more items
and doing better than they had on previous sessions. We were forced to conclude that the neuralmechanism
by which binaural beats influenced performance is not open to correct subjective evaluation.

In terms of physiologic correlates of workload and performance we are considering the electromyogram,
the electrocardiogram, the measurement of various metabolites in the parotid fluid and urinary tract, etc.
Unfortunately, these physiologic correlates, while telling us that a human being has been stressed, do
not tell when in the course of time the stress occurred, or what was the nature of the stress. They simply
tell us end result of workloads and performance which alter the body's physiology in such a manner that
the effects can be measured. Nevertheless, some of these measures do yield interesting correlates of
performance.
146

Stress correlates of workload and performance are somewhat more ambigious since they bridge both
the psychologic and physiologic factors. Perhaps, it would be better to consider terms such as environ-
mental and operational stresses. We have to consider the problem of the acuteness of the stress versus
chronic stress versus the cumulative effects of stress. Perhaps, we will be able to explore the usefulness
of the concept of task-induced stress, where the stress lies clearly within the task in such a way as to
present the humon operator with a unique situation.

Some of the psychophysiologic correlates that we will consider are the critical-flicker-fusion rate,
the psychogalvanic skin response and electro-oculography. These and other measures can be useful in terms
of revealing parameters of central nervous system function. We will elaborate considerably in thia area
because it leads us to the imporLant concept of central nervous system correlates of workload and perfor-
mance. Nevertheless, psychophysiologic correlater are difficult to relate to actual performance or to
workload effects because they may reflect subjective evaluation of task difficulties and they may be
related to subjectively calculated probability measures. Nevertheless, if one cen tease out thece
factors one is left with some correlates that may reflect the activity of the cenLral nervous system.

In spite of tie various problems of correlates of assessment, we will try to explore some of the
various correlates, both old and new, which uay offer some help in the quest for measures and assessment
of human workload and performance. No attempt is going to be made to make this a global overview. Rather,
we have been highly selective in eliminating many measures which appear to offer no fruitful results for
the amount of effort: expended. We have carefully eliminated any measures which can be retarded as
selection tools or measurements of ability, skills and so forth which have little to do with the ultimate
question of trying to evaluate or predict human performance in the operational environment.

Psychologic Cozrelates: The tasks of vigilance, monitoring and tracking seem to have a common root. Most
of them involve relatively long times at the task; they involve the detection of a signal or the detection
of a nonsignal or nonoccurrence and some form of motor response. Usually, the tasks are of a simple
nature, although they may be made more and more complex in terms of additional targets, etc. Such tasks
also may be used to evaluate the effects of other kinds of stimuli on vigilance, monitoring or tracking
type performance.

In general, we have found that this type of task when considered individually, that is, one task
performed by one operator, in one session differ considerably when they are embedded in a multiple task-
type simulator or a multiple task paradigm. The vigilance monftoring or tracking task is felt to measure
alertness and provides for minimal requirements for intellectual and neuromuscular function. A typical
such task is described in the Neptune system (3) wherein the display consists of three meters with a zero
centered needle which deflects either left or right and six push buttons, two for each meter. The subject
monitors the meters until the needle deflects then he pushes the correct button and the needle returns
to the center position. The measure of this performance is response time. Programing of the signals
is aperiodic. This is a simple, very undemanding task element, but the behavior it measures is considered
important at low levels of arousal. Trumbo (4) points out an experiment in which they had subjects track
step-function sequencec with six possible target positions. From each position there were two alternative
steps, unequally pr~iulle and either in the same or opposite directions. The Lubjects had to anticipate
to minimize error, therefore, each step presented them with either an amplitude or a direction prediction
problem. The outcome scores which they obtained showed a relationship between input uncertainty and
tracking error. Evidence for response strategies or organization came only from continuous records. These
records reveale that subjects clearly used different strategies in the two prediction situations, matching
event probabilitie-6 and prcdicting direction, but averaging probabilities in predicting amplitude. The
importance of this finding is that not only were these strategies unavailable in the outcome scores which
are usually neassred in terms of response time, but the averaging strategy could not have been identified
if subjects had seen limited to discrete response alternatives rather than a continuously graded response.

In the meah;vrement of vigilance, monitoring, and tracking activities, usually we have found a number
of hybrid tasks put together to evaluate a particular problem. However, in general we find that a
nonmechanical ele(.cronically driven system such as an oscilloscope provides a display with the most
advantages. It.is accessible to automated scoring and is readily adaptable to pursuit or compensatory
displays and to one or two dimensional courses. Ideally, the response or coatrol apparatus should permit
the operator to produce a continuum of graded responses, especially if one is interested in evaluating
the aspect of motor skills along with the vigilance or monitoring activity. There are many off-the-shelf
function generators to provide a programming system, but the use of a analog computer system would
certainly be gdvisable if one is seriously interested in the determination of probabilities and the ability
to vary stimulus organization along many dimensions. Most investigators are not concerned wit.4 motor
skills for the: sake of investigating motor skills themselves, but have used a motor type response as a
convenient vebicle for testing other hypothesis, usually involving procedural variables which have been
derived frogeneral behavioral learning theories.

Trumbo po:Lnts out that the rotary pursuit apparatus is a good case in point. He states that it is
certainly the b:est known and most widely used motor skills apparatus available. It is, of course, a motor
tracking task and it has been used in a host of studies on distribution of practice and other procedural
variables. Yet, in turns of task variables it is limited to little more than rate of turn, target size,
and stylus weigtt. The rotary pursuit generally yields a single time-on-target score. The time-on-target
score has definite limitatious in that it does not use all of the data in the error distribution. In
order to make use of the data in the error distribution, it is necessary to measure root mean square error
or average error or integrated error. These values are readily obtained with an electronic scoring system.

Bahrick, Fitto, and Briggs (5) point out in a 1957 paper that time on targeL scores, which art the
amount of time during a particular trial that a person is able to remain within an arbitrarially specified
region around a target using some kind of tracking device, presents some real problems in turns of the
derivation of learniag curves and their meaningfulness relative to the basic process of vigilance, tracking
and monitoring. The) advocate the use of the root mean square measurement (RMS) as the best method of
avoiding difficulties because it simply substitutes a single function for an unlimited number of functions
 147

determined by all possiblr target or activity dimensions. In general, response characteristics may
follow a continuous and normal distribution, but learning or practice results in a diminished variance
of this distribution; however, performance is scored according to an all or none criterion of frequency
of occurrence. This scoring practice accounts for the lack of predictability of such tests as the
steadiness teat, the dotting test, tweezer dexterity tests, pegboard tests, etc., whenever success is
scored agaiust an all or none criterion.

Kennedy (6) reports on a vigilance task which increased in difficulty from one to three channels.
When he stressed a group performing both one and three channel vigilance monitoring, the stress group
performed better than the nonstress indicating a certain level of arousal or an alerting response to
the threat of shock.

In another study, Demaio, et. al., (7) reported comparisons between student and instructor ViVots
using a visual scanning task, showing that instructor pilots learned to attend to critical features more
efficiently than do individuais with little or no flight experience. This suggests the interesting
possibilities of using a variety of scanning tasks in the undergraduate pilot training program to facil-
itate the more rapid development of adaptive scanning strategies. Thus, instead of using a scanning or
monitoring type task to evaluate stress or other such factors, we have the use of the task as a training
device.

Reaction time has also suffered in the total context of suspicion about the ambiguity of simple
performance measures. This again relates to the concept of performance versus effort. This can be seen
in the form of simple performance In two working situations which are obviously different because the
operator sees it as his task to achieve a particular output rate and he adjusts his effort accordingly.
Now this effort can be detected by various refinements of performance data, but it remains true that
straight forward speed-error data will not easily reveal difference of effort. Using speed as a measure
of complexity makes unwarranted assumptions about the essentially sequential rather than parallel nature
of humEn performance processing. Nevertheless, reaction time measurement has been an attractive area
for sc~entific research for some time. Apparently scientists have been intrigued with the attempt to
quantify the absolute speed to which a human can react to a signal. Perhaps the landmark paper in this
field was one by F. C. Donders in 1865. O'Donnell (8) reports that Danders developed a "subtraction
method" of reaction time. Basically, this theory stated that the elements of the reaction time response
were addictive, in that each component of the response begins immediately when the only, when the
preceeding component has ended. Thus, decision "is a component of a reaction involving choice. When the
decision is being made, nothing else is going on and when that process ends, the subjective immediately
moves into another, perhaps, movement phase, if this and other assumptions are true than dezision time
could be calculated by knowing the total reaction time and substracting from it the time required for all
other components of the response." Using this approach, Donders distinguished between three types of
reactions. The "A" reaction involved a single response to a single invariant stimulus. This is what
is now called simple reaction time. The "B" reaction involves two stimuli and two responses with the
stimulus response relationship always constant. This is the most simple form of choice reaction Lime
presented. The "C" reaction also used two stimuli but in this case only one response was used and that
response was to be given to one of the stimuli, but not to other. Consequently, one could calculate,
response selection time by calculating "B" minus "C" and stimulus categorization time by "C" minus "A",
etc. In spite of over 100 years of use of reaction time testing with various controversies we still have
many measures that are only indirectly related to reaction time per se, but they use reaction time as a
measure of performance. These secondary techniques, if we may call them that, in general differ from
the classic few of reaction time where the stimulus is a relative discrete signal introduced by the
experimenter for the subject's discrete response. These secondary techniques involve self-pacing by the
subjects, simultaneous performance on a number of tasks and/or use of a complete series of reactions to
obtain a total score for task completion rather than a reaction time score per se. For example, Hartman
and I have used reaction time component for mental arittmetic and tracking, etc. Recently, Wood (9) has
undertaken a study of the neurophysiolog-cal basis of reaction time change as a viable means of exploring
physiological mechanisms of local muscular fatigue and fatigue effects on aensori-motor performance. Here,
he used measures of reaction time, evoked potential, and W.. as indicators of central and peripheral
activity, With this particular reaction time model he is able to fractionnte total reaction time into
component latencies, he is able to study central versus peripheral issues which are featured predominantly
in both fatigue and reaction time. Wilkinson (10) reports on a small battery-powered fully portable
device for administering a four choice serial reaction time test and recording the results on a standard
magnetic tape cassette. In preliminary performance trials, this test appears to reflect fatigue due to
continuous vepetitive responding in a way similar to classical nonportable multiple choice serial reaction
tests. Gaillad, et. aX., (11) renorts some effecta of ACTH 4-10 using a serial reaction task, concluding
that this particular drug uo"'nteracts the usual decay and performance as a function of time on task due
to increasing boredom and mental fatigue. Bartz (12) describes an experiment using peripheral detection
and central task complexity where reaction time is measured relative to the peripheral stimulus. He
supports Hebb's arousal theory which wo'ld predict that increasing the complexity of the central task
would heighten the subjects' vigilance performance. Salzman and Jaques (13) explored the relacionship
between heart rate changes and reaction time. They found no relationships between reaction time and the
heart beat immediately precqding the stimulus or with the beat during which the stimulus was presented.
Therefore, response latencies in terms of reaction time did not differ significantl; as a function of
phase in the cardiac cycle as predicted by J. Lacey who suggested that feedback from cardiac events can
effect central functioning by a negative feedback regulatory loop mechanism. Thackray, Bailey, and
Touchstone (14) reporting on boredom and monotony while performing a simulated radar control task showed
that a high bG..edom/motonony group revealed greater increases in response times, heart rate variability,
and "strain", and a greater decrease in attentiveness. They conclude that the pattern associated with
boredom and monotcny seems m~re closley related to attentional processes than to arousal. Holt and
Brainard (15) reported an exIeriment using reaction time and a condition of selective hyperthermia where
they raised cortical temperatures. In tLis task, (a simple choice reaction time task) response times and
response variabilities were decreased compared to performance in either control or placebo condition.

t......v•h.•r .,tea| Although it has always been recognized that both psychologists and physiologists
148

heart rate, etc., as physilogical and others such as we have just discussed, vigilance, tracking,
monitoring, as psychological. However, as singleton (16) has pointed out, it now begins to looks as
though the complexities and interactions within the human body are such that neither discipline iw
adequate alone for the study of any problem of man at work. The physiologist can be accused of too
narrow an approach with Insufficient regaMd to cortical dominai•ce and tending to deal with endocrine and
autonomic parameters. Similarly, the psychologist can be accused of treating the human operator as too
"91pure" an information processing device without sufficient regard for sibcorLical. and scematic tactors
which clearly influence performance. Directiug attention now to the physiological correlates we find
that physiological measures of heart rate, muscle physiolcgy, body metabolites such as 17 keto-steroids,
etc., have been used to provide a method of measurement and to provide a set of standards. In terms of
a set of standards derived from these kinds of measures it must be pointed out that it has been difficult
to determine that a partciular task requires an energy expenditure of so mNny calories per minute or
hour, but it is even more difficult to determine whether this energy expenditure, or beart rate level, or
outpeLring of metabolite constitutes a light workload, a light energy stress, or a b),avy, or even an
intolerable amount of effort on the part of the human. Another aspect of the general difficulty about
physiologicai measures is that the stress on the operator tends to have similar effects whether it is due
to work, development, fear or environmental factors such as noise, vibration, etc. Nevertheless, there
is some value in using physiological concep..s to attempt to predict the behavior of the h.man operator
keeping in wind that the relationship between simple physical measures of the environment and the
corresponding effects upon the operator invariably turns out to be a multidimensional problem with
dominant influences from many variables difficult to measure or control especially those in the psycho-
logic realm of attitude and motivation. Sharkey, McDonald and Corbridge (17) point out in a paper in
which they evaluate pulse rate and pulmonary ventilation as predictors of human energy cost that this
human energy cost and efficiency are of considerable importarce in the evaluation of the equipment for
industrial tasks. Now this is particularly true if we look at Industrial tasks and related equipment in
the light of aircrew protection garments for altitude effects, thermal effects, and chemical defense.
Pulmonary ventilation rate as a predictor of human enery costs has long been known and used; however, the
accurate assessment of ventilation -ate depends on cumbersome gas analysis techniques and still requires
that %as he collected in the fieWd and transported to the laboratory where the time spent in analysis
still restricts sample sizes to those relatively small. Therefore, this particualr investigation attempts
to compare the precision of prediction of human energy costs afforded by both pulse rate and ventilation
rates. In short, in spite of the attiactiveness of using relatively simple determination and rezording
of pulse rate, the use of pulse rate alone in lieu of ventilation rate would indicate the possibility of
larger errors in predicting energy cost. In spite of the drawbacks of pulse rate alone, it should be
pointed out that Sharkey, et. al., 2ndicate that predicted energy costs were over-estimated rather than
under-estivated (17).

In another study relating to tamk and load diffi:ulties using the EKO by Schwarz and Ekkers (18),
they discus. the task of developing the optimal functioning reliability of a complex system from three
aspects (1) the development of analyais of the reliability of the system (2) the organizational rules
of procedure by which unanticipated emergencies can be forestalled and (3) equipping the individual
operator physically and mentally to ragulate tasks and load difficulties. In this study, they found t.at
EKG was significantly related to the pe-rceived gravity of an unannounced or emergency situation. In
another study relating task demand reflected in physiological variables, Frakenhaeuser and Johausson (19)
measured catecholmine excretion and heart rate variance pointing out that the physiological arousal
indices were more susceptible than performance measures to the level of task demands. In other words,
the higher demand imposed by a double conflict task was reflected in relativeYy larger increases of
adcenalin excretion and heart rate where as performpnce measures which were psychological remained
unaffected. A study by George Montomgery (20) on the effects of performance evaluation and anxiety on
cardiac response in anticipation of a di:ficult problem solving task showed that analysis of second-by-
second changes in cardiac rate revealed that waveform components were sensitive to both anxiety and
failure within the evaluation stress condithoa only. Initial cardiac acccleration responses covarted
with performance measures across anxiety ,%rc ýps apparently reflecting differences in confidence or moti-
vation. Concept ef the anticipated problem solving task was reflected in a cardiac foreperiod deceleration
response which is very likely related to attentional readiness for the beginning of the problem.

In a relatively loing-tem study of the activity of the nervous system during pilotage activities of
letdown, approach, and landing Nicholson aad his colleagues (21) have related pilot subjective assessment
of his workload to changes in heart rate using the RR interval and the finger tremor measured by an
accelerameter. They have concluded that the mean heart rate interval around touchdown reflects the work-
load of tl~e crew's letdown, approach, and landing phases whereas changes in finger tremor are associated
with untoward events during the approach which relate to difficulties in the dynamic flight situation
involving weather, wind shear and other factors. A follow-on study of four years of workload assessment
vas dore to determine how effective their measures were in terms of reliability. They report that the
subjective assessments of the pilot are meaningful. However, they note that thý cLgree of neurological
changes associated with the possibility of impaired subjective analysis of workload may be related to
the fact that under difficult circumstances a pilot may have a degree of central nervous system arousal
above that which may be associated with optimum performance. Their finger tremor technique is interesting
Lecause it may be related to the release of catecholmines which are in turn associated with finger tremor.
On the other hand, ballisto-cardiographic effects may play a role in this mechanism, since finger tremor
has been observed with muscular contraction during pronounced tachycardia. As they conclude, "the
peripheral changes in nervous activity observed during the letdown, approach, and landing may indicati
two physiological states both of which arise from central nervous arousal. In the case of high workl ad
letdowns without untoward events, profound cardiac acceleration and limited finger tremor are the physio-
logical changes of neurogenic origin. In letdo'wns in which the approach is complicated, profound finger
tremor dominates the pict6re and may be associated with circulating caterholmines."

In another study relating physiological correlates to changes during a mental task, Kahneman, Tursky,
Shairo and Crider (22) had subjects perform a paced mental task at three levels of difficulty while they
recorded pupil diameter, heart-rate, and skin resistance changes. They reported a similar pattern of
sympathetic-like increase found in the three autonomic functions during performance intake and processing
 149

followed by decrease during toe report phase. The peak response of each of these three measures was
ordered as a function of task difficulty. There is considerable evidence that problem solving performance
as well as other tasks are associated with activation of the sympathetic nervous system indicated by
increased electro-dermal activity, Increased heart rate, increased blood pressure and peripheral vasoco:i-
, striction.

It has been long known that pupil dilation occurs during mental activity. More recent research by
Kahneman and Beatty (23) suggests that this indicator may Le p:rticularily sensitive to mental activity
in a special way. While it is true that pupillary chan~e* ar associated with activitation of the
sympathetic nervous system an even more important index of arousal is the fact that the oculomotor nerves
which act to change pupillary sizi originate in the ascending reticular formation and provide ue with an
important window into that system. It is unfortunate indeed that pupillary measures are so difficult to
obtain in terms of equipment and the physical constraints imposed. Nevertheless, important work is
underway that will hopefully relate pupillary changes to other more easily obtainable physiological and
psychophysiological correlates of workload, peeformance, and stress. Before turning to these phycho-
physiological correlates, we will direct immediate attention aud cowent to some of the remaining
physiological .orrelates, namely body metabolites, and the electromyograph.

SBy measuring s:Lmultaneously the urinary excretion of most of the known hormones, it has been
established that the organism's response to stress involves a total neuroendocrine apparatus. As Dukes-
Dobos (24) has stated, these hormones can be divided into groups according to their excretory pattern.
One group of hormones is excreted in increased amounts during the stress exposure and the other group
shows a biphasic change inasmuch as these hormones are excreted In decreaaed amounts during the stress
and In increased amounts during the recovery phaeo Studies performed on the urinary mucuproteins suggest
that the excretion rate of this substance is an indicator of the speed 0f catabolic processes in the body
reflecting the balance of the total neuroendrocrine response to streias. These measures while important
present us with certain problems in the interruption of such changes, •'tr instance, we only know that
the individual has been stressed; we do not know the exact time in wli-c, the stress occurred. Als', a
reduction in excretiun after repeated exposures to a stress may be due to either adaptation or to fatigue.

Since the active state of the human operator is connected with the sympathetic tonus, one could assume
that the hormones of the sympatho-adrenal-medullary system (adrenalin and noradrenalin) must always be
excreted in increased amounts during physical exercise or mental work. However, as Dukes-Dobos points out,
while some investigators have found increases in one or the other cathecolmines in the urine after physical
exercise or physchological stress others did not find such changes at all. One reason for the confusing
results may be that the blood-brain barier permits only a small amount of noradrenalin to cross through
from the brain to the blood and then show up in the urine. Therefure, the urinary noradrenalin level
depends upon the activity of the peripheral sympathetic nerve endings which may or may not be related to
noradrenalin release in the brain. Thus, urinary noradrenalin does not give a reliable estimate of the
total noradrenalin excreted in the sympathetic nervous system, On the other hand, urinary excretion of
adrenalin may reflect completely the activity of the adrenal medulla. According to the classic experiments
of Von Euler (25) excretion of catecholmines will increase after physical exercise only if the subject
discovers that the performance requires a special effort.

In one of the many studies on airplane pilots performed by Hale (26) urine was sampled over a 28-hour
period every four hours from the crew members during the first transatlantic helicopter flight. The flight
was a risky undertaking and bad weather conditions often threatened its success. The average adrenalin
and noradrenalin excretions of the crew members were elevated. What can be considered a unique finding
was that an increased adrenalin excretion during the flight was observed in all ten of the subjects.
Other urinary metabolites measured were excreted in Increased amounts by some subjects and in decreased
amounts by others compared to the controls. Thus, adrenalin excretion seems to be the beat parameter
for accessing the magnitude of stress brought about bv a task which is not demanding as far as physical
exertion is concerned, but is connected with stressful work conditions and is in fact hazardous.

Reflecting Selye's (27) stress concept, physical as well as mental work can be considered as a stress
factor which may evoke the general adaptation syndrome, thus, activating the pituitary-adrenal-cortical
axis. Many studies have demonstrated that the excreted metabolites of this endocrine system show quanti-
tative changes after physical exercise as well as psychological stress. This mechanism has been explored
by studies utilizing measurements of 17-hydroxycorticosterolds (17-OHCS). In general, we have found that
during periods of stress, the 17-OHCS excretion incre&ses; however, we have also found that an exposure
of day-to-day stress may bring about a state of "chronic-adaptation" fatigue which will cause a drop in
17-OCHS excretion instead of an elevation. A relatively unique tool for the evaluation of 17-OCHS levels
was pioneered and developed by Shannon (28) using parotid fluid collection and analyzing these samples for
free 17-OCHS levels. The development of this technique is very tnteresting to follow and represents a
determined effort to avoid some of the dangers, discomforts, nnd logistics problems involved in collectig
in-flight specimens of blood and urine. The refined technique for collecting parorid fluid involves a
plastic collecting device using an acrylic bite-block molded to the individual bite of each subject. This
technique allows easy and rapid self-positioning of the device over the parotid duct opening. I' should
be noted that in a study by Warren, Ware, Shannon and Leverett (29) they state that the in-flight parotid
fluid collection technique has been developed to the point whnre it represents a valuable adjunct for
in-flight physiological studies. This is especially true because rises in steroid levels in parotid fluid
does not demonstrate the lag that is characteristic of urinary steroid responses.

The electrical measurement of muscle activity, the electromyograph, has a long history. Most of the
derived muscle physiology relates to laboratory studies in which the particular muscle of muscle group has
been stressed to bear maximum muscle contractions during a relative uhort period of time. Not many muscle
studies have been related to operational purposes where muscles are stretched, fatigued or otherwise tasked
over long periods of time and unique workload situations. If we set aside the work of those various
researchers and clinicians interested in muscular skeletal relaxation techniques perhaps the most important
investigations in the neurophysiology of muscle has been done by Basmajian (30). We will not review all
of hbi work becau•u• mot of It is of a basic nature and is not of Immediate operational interest. However,
control that can be obtained by humans to indicate that this important physiological correlate should
not remain ignored in the assessment of workload, performance, and stress.

saamajian has developed a technique and the necessary instrumentation in the form of bipolar
intramusciar electrodes to study the changing patterns of activity of individual motoneurons through the
application of modified electromyographic methods. In a series of studies he has demonstrated that human
beings can learn to activate or repress any number of spinal motoneurons in a given pool. The human
subject can also learn to voluntarily select individual motoneurons and to control'the firing of these
neurons through the assistance of auditory and visual feedback. Some of his subjects learned such exquisite
control over these individual motoneurons that they were able to produce various rhythms and patterns by
deliberately speeding and slowing the firing of the individual neuron. Basmajian's technique is to be a
proi:sing method for the study of many fundamental phenofona in the nervous system relating to cortical
and sa"cortical effects upon the motoneuron and the of conditioning and learning. One other important
area of investigation would be that of pharmacological agents on various parts of the motor pathways and
muscle activity itself. Of a more applied nature in the area of electromyographic investigation, Lafevers
(31) reports on a work task performed in a full pressure suit. Lafevers performed a power spectral density
analysis of EMG recordings from several muscle groups involved in a push-pull task at various reach
positions in both a space suit and in shirt sleeves. He feels that the power spectral shifts indicated
significant findings relative vo the performance and stress requirements for these muscle groups. The
reason that these would not be expected to appear in ordinary electromyographic determinscions is the
fact that the task requirements of the study were not of a fatiguing nature nor did they stress the muscle
groups to their upmost. This type of work suggests many areas where the relationship between the man-
machine interface in terms of motor activity and response requirements might be explored. Here, it should

I
be possible to identify task requirements that promote muscular fatigue and the resultant effects of this
fatigue on both man and the particular task involved.

Stress Correlates: The stress correlates of workload and performance might be considered the environmental,
operational, and internal results of acute, chronic and cumulative effects of psychological and physio-
logical activities. In response to environmental stressors such as heat, noise, vibration and so fosth,
and the operational requirements of a particular task, duty, or mission the internal ei~vironment of the
human begins to respond in a more or less predictable fashion. The end result of the stress correlates of
behavior is a deterioration in activity and a series of determinable changes which we usually call fatigue.
Fatigue, as Grandjean and Kogi (32) report in their introductory remarks to the Kyoto symposium on methology
of fatigue assessment, is a subjective sensation in many ways where we feel not only tired in our bodily
parts and clumsy in psychomotor activity, but we feel hampered o.d inhibited in doing either physical or
men.al work. This inhibition of activity continues until we are constrained against doing any form of
active endeavor. These sensations of fatigue can be assumed to have a protective function in that they
force us to avoid further stress and allow recovery to take place. The concept of fatigue is not a
popular scientific term because it is difficult to evaluate and to quantify. A series of studies reported
by Wolf (33) and later by Snito (34) report that the sensation of fatigue has three major components: (1)
a sensation of bodily tiredneS.s and drowiness; (2) a sensation of weakened motivation or concentration
-owards a task and, (3) a group of physical complaints that relate very closely to what are commonly called
the psychosomaLic disorders. Thcse psychosomatic complaints are usually those of headache, palpitations,
tzchycardia, shortage of breath, loss of appetite and indigestion or sleeplessness. A predominance of
these kinds of -omplaints is usually referred to as clinical fatigue. In the presence of clinical fatigue,
absences from work predominate due to "illenss," and there arises a general negativa attitude towards one's
work, one's superiors or the place of work which obviously can just as well be a cause of clinical fatigue
as well as be a result of it. Compounding the problem of evaluating chronic fatigue is the fact that
clinically it is well known that people with psychological conflicts and difficulties are especially prone
to this state. This makes it difficult to separate tI-e psychogenic factors from exogenous causes of
fatigue. In spite of the fact that the actual components of fatigue are somewhat difficult to scientically
quantify, it is not difficult to assume that the commonly experienced sensations of fatigue are very likely
a biological sign of the necessity for man to enter into a recovery phase by informing us that the relative
inflow of fatigue is exceeding our capacity. As Grandjean and Kogi report the following signs are observed
in conditons of chronic fatigue: (1) a general weakness and drive and loss of initiative; (2) a tendency
to depression associated with unmotivated worries; (3) increased irritability and intolerance (occasionally
exhibited with unsociable behaviors).

In considering the stress correlates of workload and fatigue we n~ed tobe aware of the role of the
activating system and inhibiting systems of the CNS. We know that the brain contains neural structures
responsible for maintaining wakefulness anO5alerting the cortex. It has been shown that lesions of the
medial mid-brain make animals inattentive with low motivation and drowiness. This structure is located
in the reticular formation of the mid-brain and is called the activating system. Stimulation of this
system arouses the individual or animal, while destruction of it causes the animal to go into a permanent
coma. There are also neural pathways leading impulses from the ceberal cortex ba-k to the activating
system. These corticof-igal pathways converging on the reticular formation have a function similar to a
feedback system, that is, impulses originating in the cortex are capable, through this feedback, of
stimulating the ascending reticular activating system which in turn maintains the cortex and the behavior
of the organism ir a state of arousal ai.d alertness. All of the classical afferent pathways coming from
the sensory organs send collateral impulses to the reticular activating system. This means that impulses
from the environment, through the sense organs, or from muscle activities, can stimulate the ascending
activating system and thereby increase cortical activity. Though there is recent evidence that lateral
brain stem regiuqs are as important as medial regions for attention and arousal, it is generally admitted
that unspecific neurons decisively regulate arousal and attention. Related to the neurological aspects of
workload/fatigue are other investigations that have shown that stimulation of the activating system can
spread to the autonomic nervous system giving rise to hormonal changes in the internal organs such that
the organism may poise itself for energy expenditure.

The work of Hess (35) showed that electrical stimulation through chronically implanted electrodes
produced a tendency to fall asleep and to produce pronounced muscular relaxation in cats. This discovery
later confirmed by many others has shown a active inhibition mechanism which spreads from the subcortical
structures to the ceberal cortex and acts to depress cortical functions. These systems have a direct
depreesing influence on the ascending reticular activating systems. Therefore, cortical inhibition can
result from two different causes. On the one hand, cortical activity may decrease as a result of lowered
sensory inpuits or a lowered corticofugual feedback. This might be called a passive inhibition. On the
other hand, cortical acti,-ty can be reduced by an active inhibitory function which would be elicited by
4
increased activity of the inh .bitory system. It is interesting to note that we see changes in the brain
wave which involve a flattenink' of electrical activity which are associated with suppressed behavior in
both fatigue, in states of chronic anxiety, and in certain drug effects whica act to suppress the central
nervous system. It is important also to remember that the organism regulates its feelings of fatigue or
relative arousal not only through the neuralmechanisms, but through endocrine factors which are ultimately
responsible for maintaining a certain functional state for hours or longer periods of time.

In spite of the neurologital and endocrinologic relationships discovered and understood, it is still
a problem for its to remember that fatigue is still subjectively evaluated. In other words, just because
the conditions of fatigue exist does not necessarily mean that performance decrevent occurs. It may be
true that subjective feelings will preceed a loss in performance ability, but not necessarily. It is well
known that in spite of great fatigue the human organism will response with adequate, if not, lifesaving
performance levels. Further, we know thmt there is a situation of unacceptable fatigue which people
classify as a kind of fatigue called overwork, overload, or exhaustion or other kinds of terms. Since
these kinds of fatigue concepts are related to subjective judgment we, therefore, get back to the psycho-
physiological implications of workload, performance, and stress. We will see that some investigators
in assessing fatigue feel that there are indexes of fatigue through physiological measures such as increase
in heart rate, reduction ok sinus arrythmia, and so forth. Regardless of whether one is concerned with
the physical aspects of fatigue or the mental aspects of fatigue, certain symptoms can be considered as a
consequence of cortical inhibition activity. The following symptoms of what might be called "cortical
fatigue" are those which need to be evaluated, (1) decrease of attention, (2) slow and impaired perception,
(3) impairment of thinking, (4) decreased motivation, (5) decreased performance speed, (6) decreased
accuracy, and (7) decreased performance reserve for physical and mental activity. While most of these
factors have been investigated to some extent it is probably the electrical activity of the cortex which
may give a better picture of activity which can be considered as having a direct regulatory effect.

In a factor analytic study of mental fatigue, Kogi and Saito (36) were able to demonstrate that
certain changes in cvztical functions were related to various phases of a 24-hour period. The measure of
cortical activity they selected was the critical flicker fusion test, but changes in CFF were also
reinforced by changes in a choice reaction time test.

A study by Ettema and Zielhuis (37) investigating the physiological parameters of mental load demon-
strated that a simple, binary choice test providing different me"tal loads or levels of difficulty,
showed systematic changes in heart rate, sinsus arrhythmia, systolic and diastolic blood pressure, and
rate of respiration.

Kahliwagi (38) was able to constuct a fatigue rating scale which allows a judgment of human fatigue
Lhrough a person's appearance. The use of such a scale might be very helpful in th• field as far as
management or field commanders are concerned and might be of use to some of the mission crew fatigue
studies done at the School of Aerospace Medicine by William Storm and his colleagues. Presetntly, subjecti,
fatigue and sleep data are collected from various mission groups. These measures are used to assess the
overall effects of mission requirements upon sleep loss and workload requirements (39).

A system using a concept of task-induced stress was developed by this author and used in the stress
testing of special mission personnel in the U.S. Air Force (40). This concept was structured around the
tasks an operator must perform in an advanced space system. He muat perform a relatively large array of
discrete, discontinuous operations against a background of moittoring and information processing tasks.
In terms of information theory, the discrete, discontinuous functions would constitute a source of noise
in the form of unwanted or distracting signals when the operator was trying to monitor and process a
continuous input. Increasing the signal rate of the discont..nuous tasks makes the detection and identi-
fication of the continuous task mote difficult in the same manner that tncreased noise acts to degrade
audible signal detection and recognition. By structuring the task situation so that the cperator is
uncertain as to what is signal and what is noise, it is possible to cause him to continually shift his
attention from signal to noise and noise to signal. This is the natura of competing tasks and the end
result of such a situation can be regarded as task-induced stress. BYv further structuring the situation
so that the operator has been allowed to find out that he can in fact perform both the discrete, discon-
tinuous task and the continuous monitoring task independently, he is quite apt to assume that he should
be able to do them together with perhaps only a little more effort. When he finds out that he has much
difficulty doing both tasks he is led to conclude that there is something wrong with him and he would
much better if he could only find the optimal technique or "a gimi-ick". This sort of structuring tends
to invite the formation of internalizvd, psychologic stress which is not relievv- much by hostility
towards the tasks themselves. Since there is no obvious source of the proficitncy problem presented by
the competing task, the psychologic feelings generated by failure to perform well tend to be self-directed
rather than task directed. In this situation, the induced stress is more than the sum of the stress of
performing each task independently. The results indicated that a criterion group of those finally
selected for the special mission using various other criteria was better able to adapt to the two competing
tasks and was less suscept.ible to the signal noise, ambiguity and the induced task stress than the special
mission personnel group as a whole.

It must be noted that the evaluation of highly specia.izi'd groups selected by virtue of years of
experience and special talents presents a unique problem to those of us charged with the reEponsibilites
of investigating the effects workload, performance, and stress within and upon the human operator.

Another important area of performance decrement in relationship to environmental and operational

working conditions is that of exposure to hot and humid environments. In an important study reported by
Welch, Longley and Lomaev (41) they found that sweat loss and pulse rate were found to be unreliable
methods of measuring fatigue and that skin temperature was completely unreliable as an index of fatigue
152

except vhen the temserature and humidity were high. Their experiments confirmed that a rectal temperature
of 38.8 C to 38.9 C will in most cases coincide closely with the onset of actual exhaustion. Another
study in this area by Grivel (42) indicates that the permanent, specific heat effects on psychomotor and
mental performance are related to preferentially in that different aspects of the same activity were
considered to determine the effects of climatic stress. In other words, heat acts differently on the
reactivity aspects of performance than those aspects of performance associated with continuous attention.
In the field of time-varying heat effects, studies have examined the possible transitory effects of heat
as well as long term evolution of effects found at the time of first heat exposure. Here, a sequence
of events can be distinguished, each characterized by a particular kind of ambient heat effect upon
performance. This suggesti some type of learning or conditioning takes place relevant to heat stress.

The Psychophysiological Correlates: We have seen so far that we have both common and scientific knowledge
that individual and combined streses, both physiological and psychological, can adversely effect mental
performance and judgement as well as physical performance. We have just discussed how both hyperthermia
and fatigue can produce deteriorated, objective judgements regarding environmental situations as well as
degraded performance. The individual's subjective judgement or insight concerning the quality of his own
performance is similarly degraded. With sustained exposure to stress, he tends to overestimate his
capability and to discount his errork. Subjective identification of degraded central nervous system (CNS)
function is generally based not on recognition of degraded performance, but on secondary indicators such
as diming of vision in case of hypoxia and reduced span of attention in the case of fatigue just to use
one example. We are primarily interested in looking at psychophysiological parameters which relate to
central nervous system function. Our interest stems from the fact that to the extent that these CNS
changes are detectable through analysis of peripheral physiologic measures they can provide a sort of
warning system of primary higher CNS functional decrement in the same sense that an oxygen partial
pressure meter provides primary hypoxia warning.

Many years ago the Cambridge cockpit series of performance/fatigue studies examined behavioral
changes observable through several hours of continuous performance when subject's "flew" a specially
instrumented simulated aircraft. Bsrtlett (43) summarized these studies in terms of skill-fatigue effects.
The experiments showed beyond question that, under the conditions imposed, "operator fatigue" does occur
though in most cases the operator himself did not realize it. Within a maximum of 8 hours of simulator
operating, the experimenter concluded that the subjects were still able to perform the operations, but
only if they were especially careful to avoid known deficiencies characteristic of fatigue. Some
inexperienced subjects developed significant deterioration of performance after only 14 hours. One highly
motivated, experienced subject went 8 hours without appreciable deterioration. Most subjects, as fatigue
progressed, showed a lack of coordination between the recognition of the required operation and the
necessary response. This is related most logically to impairment of the integrative function of the
association areas of the cerebral cortex. Marked increases in lability and irritability were also observed
along with changes in judgement. As time progressed during the performance, the number of small errors
increased, but were later replaced by large errors. T1,is was interpreted as reflecting degraded neuro-
muscular control with increasing levels of frustretion. This was compensated for by a judgemental change
in the standards of accuracy. The subjects were utiaware of this change in their judgement of acceptable
performance unless it was called to their attention. This idea of subjective lack of awareness is crucial
in the operation of high performance man-machine systems.

Other studies of fatigue have demonstrated that subjects become tired of a specific task and show
rejuvenated performance upon changing tasks. From a neurophysiological standpoint, this relates to a
reduced level of general CNS activity upon habituation to a monotenous task. However, with the introduc-
tion of a novel stimulus there is a marked increase in CNS activity. This Is the so called arousal or
activation rusponse. Subjective appraisal of performance and its relation to objective criteria under
conditions of fatigue produced by prolonged wakefulness using skin resistance measurements as well as EEG
tracing has been reported by Burch and Greiner (44). Generally, they found the subjective evaluations
showed a high correlation with their bioelectrical measurements during pre-experimental control periods
and the earlier portions of fatigue. However, as the fatigue progressed the subject's ability to evaluate
his own state of consciousness begins to break down.

Several studies of mild hypoxla at the School of Aerospace Medicine have shown frequent lapses in
simple performance tasks lasting only a few seconds and suggestive of a momentary loss of awareness. In
fact, one 'f the tasks incorporated into a multielement psychomotor test device previously ceferred to as
Neptunt- (an acronym for "neuropsychiatric test unit") was designed to provide a relative measure of
operator consciousness during periods of experimentally induced hypoxia. This task was called Auditcry
Monitor and involved monitoring three Morse Code signals "A", "N," and "M" which are played in random
order at a preseaected speed. This task indicated momentarily loss of awareness and these periods of
loss correlated with EEG changes indicating reduced cortical arousal.

The stress of sleep deprivation also shows brief, often dramatic intermittent pauses or lapses in
ongoing behavior. Many studies show that these lapses increase in frequency, duration, and depth as
sleep loss increases. Between lapses subjects aie able to think and act under challenge almost as well
as under preexperimental conditions. As lapses deepen, it is increasingly difficult for the subject to
hold a stable frame of reference while performing a series of mental operations. If a deep spell of
drowsiness occurs in the middle of a serial operation, the subject will stop the sequence for a brief
time and often 'oose track of where he had been in the series. Luby (45) has attributed decrement 14
psychological test performance under conditions of 118 to 120 hours of sleeplessness to fluctuations of
attention. T.,e frequency of periods of inattention increases as a function of the hours of sleep loss.

In comparing measures of physfologic activity with observed behavior it is necessary to note that
an organized psychomotor response pattern involves three factors which must be integrated at a relative
high cortical level. These factors are: (1) detection of the signal, (2) selection of the response, qnd
(3) execution. Under conditio-s of disorganization the response pattern is fractionated ratner than
coordinated so that certain indications of responae pathology occur. For example, under stress conditions
detection may be accompanied by a startle response of varying degree to signals having high attention
value. Signal detection may be degraded by a breakdown of scanning behavior as the attention span is
 "153

attenuated. The wrong response my be selected and execution may be characterized by gross spatial errors
and psychomotor movement, that is, moving first to the general area of the control and then to the cortrol
itself. Other execution errors involve operating the wrong control or using the proper control incorrectly.

In sumary then, we see that various physiological and psychologic stressors individually produce
variable degrees of decrement and behavioral performance, some of which are predictable. However, in
combination, the effects of these stressora on performance become difficult, if not impossible to predict.
Nevertheless, it is possible to monitor neurophysiological states and events to the extent that it is
possible to identify CNS functional changes related to the primary cause of performance decrement. At the
present state of the art, the likelihood of detecting a specific erroneous judgement using CNS functional
criteria is not possible. However, it is possible at present to detect specific mental activity related
to specific external and internal processing events. We will discuss these at a later point. At present
we will deal with the identification of a CNS functional state correlated with unreliable or pathologic
performance and judgement. It is interesting to speculate upon the fact that many of the early symptoms
of some organic brain diseases and some focal brain disorders or not unlike some of the behavioral changes
just mentioned and yet to date no organized study has been made relatiag psychophysiological variables
with symptomatic or usychometric factors in such disorders.

rsychophysiologtcal Monitoring of CNS Function

The physiologic evaluation of central nervous system function can be approached from two points of
view: (1) the general state of arousal or level of consciouspess, and (2) the quantitative and qualitative
aspects of indivudal or specific 1.NS responses. We will discuss the first approach from a neurophysiological
standpoint and relate it to some existing data under the title, "General Levels of CNS Activation". The
other approach will be discussed under the heading, "Individual CNS Response."

Be.ore delving into CS monitoring itself, we should be aware of some of the considerations in terms
of measures and analysis. Individual physiologic measu:es can be analyzed from at least two aspects: (1)
averaged or integrated values, and (2) quantitative analysis which reveal changes due to individual CNS
responses. These methods have been devised to permit a reduction of data to speed analysis and to allow
for computer handling of the data. Four such measures, the electroencephalogram, the electrocardiogram,
respiration and electrodermal responses will be discussed in some detail. In each case we will try to
relate Che two kinds of analysis to the two corresponding types or modes of CNS function.

The Electroencephalogram: The potentials observed from scalp electrodes measure a part of the electrical
activity that underlies superficial cerebral cortex. The specific areas of cerebral cortex are iden-
tified with primary, sensorimotor activitj and with the integrative function of the'association areas
adjacent to these sensory areas. The sensory association areas play a major role in deriving meaning
from the impulses received in the primary sensory areas. The frontal area contributes to the integration
of the sensory aasociation areas permitting abstract and conceptualization. One can thus expect observable
changes in EEG patterns relating to changes in activity involving these higher mental functions. Much of
the problem in intrepreting ERG patterns is due to the highly complex wave forms developed. Most forms of
analysis have been borrowed from engineering approaches to vibration str.ss which also presents multiple
frequency wave forms. In engineering terms this is called frequency spectrum analysis in which the various
frequencies are partialled or split-out for individual analysis for a specific time period. A modification
of this approach developed by Burch (46) shows promise. The Burch method analyzes EEG wave forms in the
time domain expressed as major and minor periods. The major period represents the dominant EEG frequency
for a specified time interval and is defined by baseline crosses of the raw EEC. The minor period repre-
sents the superimposed waves between the baseline crosses. Major and minor periods are each sumntud and
represented as a total count during a time interval or epoch such as 10 seconds. The Burch method involves
an additional display referred to as spectral analysis. This divides the raw EEG spectrum for each given
epoch into 10 frequencies bands with reference to both major and minor periods. This analysis a- -s in
a form similar to the Grey Walter frequency analysis system. The readout provides a value for , of the
10 major and minor period frequet:y bands luring every epoch. The amplitude of this writeout i. ates the
total time in the preceding epoch during which the analyzer detected periods with values fallin. within
the frequency limit assigned to that particular band. The model frequency band for either the major or
minor period is that band in which the greatest accumulated time is scored during the epoch. A total of
major and minor period counts represents a characterization or signature of the EEG frequency spectrum
during the epoch selected. Total counts over epoch's of a few seconds reflect individual reflexes involving
a major portion of the cortex. Changes in modality of the frequency spectrum during the corresponding
epochs are related to the quantitative aspects of such refJexes.

In contrast to the previously described frequency-period analysis which is concerned with time domain
only, Riehl (47) developed a method which relates both time and amplitude domains. He defined an activation
response which he called Us. This can be written in the form of a equation where Us equals F (the dominant
frequency) multiplied by the reciprocal of the average amplitude. In this equation the dominant frequency
is defined for the major period count and the average amplitude is that which is obtained by full wave
rectification and integratier. In order to derive this function, an analog computer is employed to integrate
the wave form, and to obtain a continuous real-time representation. The integration of Us over epochs of
10 seconds recorded at a relatively slow chart speed provides a convenient readout of an activation response
over specific time periods. The Us itself will exhibit major fluctuations of only a few secondo duration.
These can be evaluated as identifiable CNS resporses to known stimuli.

A more recent approach is to use power spectrum analysis using a "set Fourier transform. This yields
a representation of extremely small power shifts over very small epochs. This result can be obtained
on-line in terms of percentage of power in each selected frequency band width or in terms of puee pow, r.

The Electrocardiogram: Nervous control of heart rbte is classically described as mediated Lhrough vagal
parasympathetic cardic-inhibitor fibers and through symprthecic cardioaccelerator fibers. The vague
nerve cardio-inhibitor fibers originate in the bilaterally paired dorsal motor nuclei of the Vagus. These
nucle! lie in the floor of the fourth ventricle thro'iEhout most of the length of the medulla oblongata.
The sympathetic ecelpa"erator '~Matrally i 41red W1ei 6n0out V$T~ the rtaA
154

substance of the Medulla. Beat by beat values of heart rate are obtained by measuring the period (R-R
interval) of each cardiac cycle. An analysis of heart rate or trend, or accelerator, vs decelerator
information may be obtained by averaging the frequency of a number of cardiac cycles. This is most
conveniently done by using a cardiotachometer. The beat by beat analysis of cardiac rate represents
a very promising method of observing individual CNS reflex responses. This analysis shows two contrasting
patterns: (1) during sleep, the record consists almost exclusively of a rhythmic increase and decrease
of heart rate coincident with respiration. This is referred to as respiratory coupling. (2) During periods
of wakeful sensory motor activity, such as speaking, walking, etc., the beat by beat pattern of heart rate
shows a preponderance of nonrespiratory coupling or decoupling showing frequent csrdioaccelerator reflexes
as opposed to those seen only occasionally during sleep. The number of premature ventricular contractions
per unit time is observed to increase under conditions of stress. Other specific electocardiovascular
changes have been reported in the literature, but at this time is not yet clear whether these are a
function of direct nervous control or indirect humorial influences.

Respiration: Control of respiration is mediated through autonomic and voluntary pathways. The primary
respiratory centers lie in the medulla oblongata and in the pons. The medullary centers are described
as paired bilateral half-centers which include both an inspiratory and expiratory half-center on each
side. The half-centers are contained within the medullary reticular substance. The pontine reticular
formation contains an inhibitory pneunotaxic center and apneuistic center which exerts a strong tonic
effect on the bulbar inspiratory center. Voluntary control of respiration originates in the cerebral
cortex and is mediated through the hypothalamus. Both inhibitory and acceleratory cortical influences
appear most specifically localized in the frontal cortex.

It is interesting to note that the medullary centers for respiratory control and cardioaccelerator
control lie close to each other within the medullary reticular formation. Thus, it is not surprising
that there should be a strong interaction between respiratory activity nnd heart rate. As we have noted
during quiet periods of CNS activity heart rate is predominantly coupled to the respiratory cycle while
during periods of CNS arousal the respiratory coupling is frequently replaced or decoupled by cadio-
accelerator reflexes associated with brief respiratory arrest. Similar transient increases in heart rate
are occasionally associated with a marked increase in respiratory rate. This observation suggests that
the cardioaccelerator reflexes may derive from two clearly distinguishable neurophysiological mechanisms.

In a comprehensive review of sinus arrhytkmia reflex mechanisms, Heymans cites clear evidence that
in lower animals the cardiac vagal center is subject to two inhibitory, cardioaccelerator influences, one
arising from the lungs exhibits increasing activity with mild pulmonary inflation and the other is mediated
directly from the respiratory center (48). The latter influence in fully capable of producing typical sinus
arrhythmia in the complete absence of pulmonary ventilation. This fact makes is reasonable to postulate
changes in heart rate arising from cortical activity mediated directly through respiratory centers to the
cardiac vagal center.

Electrodermal Responses: Electrodermal responses (EDR) which include galvanic skin response and the basal
skin response, are predominantly, if not exclusively, mediated by the sympathetic nervous system which
produces changes in skin resistance highly correlated with sweat gland activity, the so-called galvanic
skin reflex.

A comprehensive review of galvanic skin reflex neurophysiology by Wang discusses stimulation, trans-
section, and oblation techniques employed in the CAT to identify CNS excitatory and inhibitory centcrs.
The suprasegmental excitatory areas include the sensorimotor area of the cerebral cortex, the hypothalamus
of the diencephalon, and the facilitatory reticular system in the diencephalon and mesencephalon. Two
pathways of the GSR which are separate at the cortical and diencephalic level converge on the preganglionic
sympathetic sudomotor neurons in the spinal cord. The fakilitatory influence of the diencephalic and
mesencephalic reticular system is characterized as followg: When the facilitatory reticular system is
stimulated in both the interbrain and the midbrain, the response or eaffect varies with the strength of
the stimulating current. Weak current elicits no response itself, but augments the reflex. Moderately
strong currents evoke a small response itself and also enhances the reflex. A very strong current which
calls forth a large response by itself, suppresses the reflex during and immediately after stimulation,
but has a late, long lasting facilltato'y effect on the reflex. This effect begins one minute after
stimulation, reaches a peak in two or three minutes and then gradually declines to zero in 30 to 40 minutes.

The inhibitory centers identified include the frontal ceberal cortex, the caudate nucleus, the anterior
cerebellar lobe and the bulbar medial reticular formation. The cerebral cortex has the least inhibitery
effect and the bulbar medial reticular forwation the strongest.

A number of stimuli characteristically elicit the GSR on jelected sites of human skin. These include
startle, painful or other strong sensory stimuli, v'.olent respiratory activity, generalized muscular
activity, and strong emotional stimuli. GSR activity during arousal conditions occur spontaneously in
response to no apparent stimulns. Tnis is the so-called nonspecific GSR. GSR's which occur in response
to known atimuli are they called specific GSR's. The sensitivity to stimuli evidenced by the frequency
and magnitude of GSR's are observed to fluctuate through relatively wide ranges in the course of normal
daily activity. This suggests a threshold-type mechanism, that characteristically effects a wide span
of control.

Individual GSRs are characterized by a transient drop in skin resistance in a period of a few seconds.
This expression of reflex activity has been analysed in terms of the number of responses per ut.it time, of
the amplitude of the individual responees, and response latency or the period of time betw-en an administered
stimulus and the onset of the GSR. Some time ago this author •emonstrated that the area subtended by the
recorded GSR, a measure which integrates both time and amplitude is a more sensirite indicator of etress
than frequency or amplitude only (49).

Basal skin resistance or BSR varies slowly o%'er a wide range as the iindividual fiuctuates through
states of consciousness on the sleep-arousal continum. High resistance values are ast..:iated with low
levels of consciousness such as sleep, and low reaistanc valies with high levels aD with intense
excitement. Basal skin resistance has been observed to vary over a range of 10 to I in a perlod of 10
minutes during the period of transition from sleep to aroused wakefullness in the morning. The BSR tends
to be lower during periods of frequent GSR and high during periods of infrequent GSR. The reduction of
BSR during frequent CSR is apparently due to the cumulative drop in resistance resulting from the failure
of complete recovery of the response mechanism to the prereflex level of resistance before the onset of
the next stimulation. This same phenomenon is seem in the repeated stimulation of other neural response
mechanisms. The relationship between BSR and GSR activity is apparently a function of the rate and
magnitude of individual reflexen and the recovery rate of the skin resistance towards higher values.
Thus, BSR can be seem as a form of integrated function of GSR activity. Analysis of both forms of
electrodural activity provides further quantitative and qualitative information concerning some aspects
of CNS reflex activity.

From this discussion, it is apparent that anatomically the central pathways mediating EDR provide
numerous sources of influence upon the observed reflex. As a practical indicator of CNS function the
enormous volume of GSR literature published emphasizes the fact that the GSR pattern produced is the
result of multiple influences at the CNS level. To date these influences are rather poorly identified
and their separate effects on GSR patterns are not clearly distinguished.

The discussion of these four measures referenced to CNS activity does not imply that other meaquren
may not be of equal or greater value. Additional measures deserving consideration include blood pressure,
pulse wave velocity, EMG (electromyography), eye motion (REM) as in dream studies, and pupillary measures.
Further research efforts will be required to adequately determine the usefulness of the information given
by each measure concerning the functional state of the central nervous system.

Central Nervous System Activation: With some insight into measurement procedures, we will look at general
levels of CNS activation. The level of CNS activation or arousal relates to the state of consciousnese
normally ranging from deep sleep through wakefulness to intense arousal. Obviously, level of arousal is
influenced by many factors including circadian periodicity, workload, emotional stimuli, and internal
ideation. From a neurophysiologic standpoint, the state of consciousness is intimately related to the
activity of the reticular formation. It, in turn, may be influenced strongly by afferent motor activity,
the amount and kind of sensory stimulation, and the emotional state of the individual.

The Reticular Formation: Anatomically, the reticular formation occu- es a central location in the brain
stem joining the cerebral cortex with the apinal cord. It is composed of a network of interlaced fibers
and contains nuclei surrounded as a group by the primary sensory and motor pathways connecting the cerebral
cortex with the spinal cord. The central cephalic brain stem which includes the diencephalic and mesen-
cephalic reticular formation is essential for awareness of i:he environment and voluntary purposeful
move ment.

In terms of the relationship of various CNS structuresj to conscious activity the ascending reticular
activating system has been identified as having great functional signifance. Stimulation of the anterior
portion of the reticular formation elicits electrocortical arousal in animals and has been used to produce
wakefulness in human analeptics. This same anterior portion originates impulses distributed widely over
most the cortical surface, particularly the association areas. It is interesting to note that there is
a clear distinction between the mere meaning of impulses received in the primary sensory receptor areas of
the cortex and the meaningful, purposeful activity evoked by concurrent activation of association areas,
via the ascending reticular pathways. Thus, we find that impulses corresponding to a visual image arriving
in the primary visual receptor area remain devoid of meaning unless the adjacent, visual assocation areas
are concurrently activated. The arrival of sensory impulses devoid of meaningful association is charac-
teristically demonstrated in sleep (or dreams).
The complexity of the relationship between the cortex and the reticular formation is emphasized by
the important role played by the descending pathways which strongly influence the core of the brain stem.
It is through these corticifugal pathways that emotional arousal and goal directed behavior of conscious
processes are mediated,

The Limbic System: The functionally related neural structure called the limbic system surrounds the
attachment of the cerebral hemispheres to the brain stem. This system is posi.ively associated with the
subjective and autonomic motor expression of emotion. Recordings of the electrical activity within the
limoic structures have revealed two patterns of electrical discharge associated with excited behavior.
This kind of behavior apparently involves the reinforcerent mechanisms of the limbic system which serves
both to increase the amplitude and to generalize the distribution of activity in other parts of the brain,
including the reticular formation. Any informati'3n from physiologic measurements indicating the level of
activation of the reticular formation should be helpful in determining the behavioral level of conscious-
ness. We have previously noted the investigative window pro-Tided by pupilography into the relative state
of the ascending reticular formation. Additonal physiologic patterns or identification of activity which
would serve to indicate the contribution or involvement of the limbic sytem to arousal would help to
determine its emotional component. At the present stage of development of biomedical monitoring most
reports of physiologic measures obtained under stress present them in the form of average or integrated
values over relatively long periods of time. The results generally correlate w~th trends in the level
of CNS arousal. However, thexe is increasing evidence that detailed analysis of differences in the
central processing of CNS responses in going to be evidenced by relative small transient changes in the
EEG. These will be related to small changes in heart rate and electrodermal responses which together will
provide more specific and reliable indicatoro of CNS arousal and the functional state of the central
nervous system.

EEG Indicators: One of the most thoroughly studied features of the normal EEG is the 8 to 13 cycles per
seconds alpha rhythm often observed most clearly over the occipital cortex. The complexity and multi-
variant nature of this rhythm has been demonstrated, to the extent that it is possible to distinguish
individual•, who demonstrate either persist alpha, responsive alpha or absence of alpha. However, it is
necessary to reclize that this is not a fixed classification and that individual alpha patterns actually
encompass a Lontinum in which "oerslstent" and "absent" types represent ends of the scale. It has been
 found that the sensory modality of the imagery characteristically employed by the individual largely
determines his alpha rhythm. Visual imagery is associated with the absence of alpha, while non-visual,
or auditory and tactile imagery is associated with persistent alpha and responsive or fluctuating alpha
patterns are related to variance in the individual's imagery modality. This highly variable expression
of alpha rhythm is further complicated by the fact that what appears as simple rhythm on a primary trace
is really often a complex of frequencies from multiple sources. Finally, it is known that alpha may be
absent due to nonspecific stress effects as seen in chronic neurotic anxiety states.

When present alpha rhythm appcdrs most predominately in the relaxed, eyes closed, awake condition.
However, it is possible to train a human being to produce predominant alpha with his eyes open, while
fully awake, and fully conscious, and fully mobile. The disturbance or replacement of the predominant
alpha frequency is called "alpha block" and is seen as a low voltage, higher frequency pattern. This is
4,• characteristic of an attentive state or alerting response. Although alpha rhythm responds sensiti.•ely
to a number of features of CNS function, so many factors are involved that no simple unambiguous conclusion
can usually be drawn from the presence or absence of alpha rhythm alone.

Generally, increased cortical arousal Is associated with an EEG of lower voltage and hieher frequency,
Sleep or certain drugs act to produce a slowing of EEG frequencies, as does training and the controlled
relaxation response of Jacobsen. In deep sleep, three waves per second are seen, the so-called delta
waves. In moderate sleep, sleep spindles or bursts of 14 per second waves occur. breazing is associated
with rapid eye movements (REM) and takes place in the range of drowsy to light sleep, the so-called
emergent/Stage II type sleep. The term emergent is used to indicate a stage of sleep occurring following
of period of one of the deeper stages of sleep. This phenomena will take place periodically through the
night with many individuals exhibiting a particular sleep pattern unique to them alone. No dream activity
is known to take place in delta sleep.

Additional studies are needed to establish a clear relationship between physiological measures,
performance, and the level of CNS arousal in the drowsiness-extreme alertness continuum of wakefulness.
In one study, it was observed that there were high levels of major period counts over ten second epochs
during increased levels of arousal and lower major period counts during decreased periods of consciousness.
Ine converse was generally true of the minor period count. Spectral analysis in this particular study
showed quantitatively the shift of the major period modal band to slower frequencies and a shift of the
minor period modal band ta faster frequencies with decreasing arousal as sleep became deeper.

The combination of both frec'uency and amplitude domains in the activation analysis previously
described, shows its sensitivity to some situations while indicating some ambiguity as a eimple arousal
indicator. Johnson and Ulett (50) examined 50 college students on three occasions using a modified EEG
spectrum analyzer. Each subject was examined three successive occasions under quiet, eyes-closed con-
ditions. Average values of the frequency spectrum for all students grouped by visits produced three
curves of comparable frequency distribution; however, the curve corresponding to the first visit was
approximately half the amplitude observed on subsequent visits. The authors concluded that the increased
anxiety level of the subjects generated by their apprehension of the intial EEG examination produced this
depression in amplitude at all frequencies. This shows that in this group of subject3, a decrease in
anxiety was observable as a decrease in EEG amplitude at all observed frequencies between three a-rd 33
cycles per second. The activation analysis which is sensitive to such amplitude changes may be a useful
indicator of the anxiety level of an individual. The precise manner in which anxiety effects CNS function
and the resultant level of performance is not yet clear, but is obviously a significant contributing factor
in some strccnful situations.

In a fatigue study at the USAF School of Aerospace Medicine, four pilots were required to complete
a 24 hour simulator flight with only a two hour refueling stop in the middle of the run. An activation
analysis of a continuous recorded EEG obtained on one of the flights showed a sustained high leval of high
frequency, low amplitude activity during the first several hours. This corresponded to the period of
expressed anxiety on the part of the pilot as to his ability to perform adequately on the simulator.
Interestingly, his first landing rated as one of the poorest of the eleven made during the 24 hours.
Toward the end of the flight a generally lower level of activation level was observed with a marked
tendency to fluctuate erratically between moderate and low levels. Th'2s, it is seea that this method of
EEG analysis promises to contribute useful information regarding the general level and fluctuations of CNS
arousal.

vi As we have indicated, it is probably a fair statement to make that at the moment there is little
promise of new and exciting use of ongoing EEG material for the enhancement of pilot performance. In
general, we can tell when a subject is getting drowsy, has gone to sleep, or, to a lesser degree of
certainty, is simply inattentive. So we are left with inferring general state changes and its usefulness
for monitoring the state of the organism. However, the event related potential called ERP (or Cortical
Evoked Response) is another matter. Before discussing the EPP, the work of Donchin, et. al. has identified
several interesting electronic signatures indicative of cortical activity (51). T1e first of these is
called N'00 and this electronic component is elicited whenever a rare or unexpected event oczurs. Another
of these is P300 and this endogenous component is seen in association with task revelant, rare stimuli.
Another component is the contingent negative variatiou (CNV). This wave form is a slow negative shift of
potential tjat occurs during the warned fore-period preceeding a motor or mental task. It begins very
shortly after the warning stimulus and terminates after a response decinton by the subject or the occurence
of a stimulus which demands a response. The final, easily identified wave form is a readiness potential,
the RP. This is similar to the CNV in that it is an event-proceeding negative shift. It is distinct from
the CNV in a sense that ýt appears prior to self-paced voluntary responses. It's occurrence is independent
of the presence of an eliciting or command stimulus. These endogenous components of the brain wave have
been stvdied in connection with arousal, attention, selective attention, emotional valence, assessment of
novelty, time estimation, uncertainty, detection of targets, differential identification of stimuli inde-
pendent of size and shpae, and the semantic classification of linguistic symbols (52).

.ii.•
--
 157

Electrocardiogram and Respiratory Indicators: In similar studies of performance the average values of
ECG and respiration have consistently correlated well with general arousal level. The highest values are
seen when performance demands and/or external stressors, such as threat or emergencies are introduced.

This -elationship between av gZch--rt rate the level anxiety was nicely demonstrated in the highly
significan. --ries of experiments by Walter (53). Over a period of four years, he performed a series of
-oapiete deft sive-4voidance ,.uai''o..o'- procedurea with 58 subjects, 37 normal and 21 psychiatric
Ile rzollecd suffi_'ent evidence e experiments to distinguish two types of relations
3h
beLween averzgz e ratz 'vkj b4c'-I -indicated by pulse-wave velocity. One relationship
showed a :.,tLa n p : asure in the initial stages of excitement in normal
subjects and in ti~e tt-- -' t : ,eri"ental stress in disturbed patients. This response was
linked with other signs of generalized tmnsion and anxiety and was associated with adaptive failure or
confusion. The other type of response showed an inverse relationship, that is when heart rate increased,
blood pressure fell. This was a transient effect showing blood pressure changes of about one-half the
magnitude of the first type of response. This secozd type of response was frequently elicited by the
penalty tone which was indictative of an erroneous response. This is an excellent example of increased
resolution and reliability of interpretation afforded by the observation of simultaneous changes in two
related physiological variables.

Ax (54) has reported a steady decrease in the mean value of the ratio of respiratory to nonrespiratory
coupling in five subjects undergoing 123 hours of sleeplessness. This serves to indicate that the ratio of
the length of time that the record is characterized by undisturbed respiration-coupled heart rhythm com-
pared to cardioaccelerator reflex rhythm relates to CNS function under the stress of sleep deprevation.
This is also an example of change in the peripheral expression of a central nervous reflex activity related
to changes anlevels of arousal.

Electrodermal Response Indicators: We have discussed two measures of electrodermal response, the BSR and
the GSR, which are observed to change with the general level of CNS arousal. Levy (55) found that BSR
compressed on a five centimeter per hour write-out was particularily valuable in monitoring states of
consciousness. Under standard conditions he found the pattern of one individual's skin response to be
consistently similar. However, the patterns of different subjects varied from an almost straight line to
a wildly fluctuating one. The flat stable line which he obtained was consistently of low resistance due
to frequent mall amplitude GSRs. The more variable tracings were of higher average resistance showing
less frequent, often large GSRs awich tended to occur in groups. He also observed that persons who
exhibited the low flat type of basic waking pattern seem to be able to maintain a more continuous and
higher level of involvement in their environment than those persons showing a morc variable tracing. In
general, then, he reports a relatively stable, low value of BSR during aroused wakefulness, a more variable
saw-tooth pattern drini drowsiness, and a high resistance pattern during sleep.

Similar changes in BSR have been noted while monitoring pilots during flight. Here, resistance is
initially low when the pilot starts flying the aircraft and gradually increases as he relaxes. His
resistance drops if the co-pilot takes control of the plane and is lowest when the co-pilot is active in
stall-type approach for landing. I suspect we would see the same response in a husband as his wife takes
over driving down the turnpike.

In his series of conditioning procedures, Walter observed that an abundance of nonspecific GSRs was
associated with muscular tension, slight tachycardia, raised blood pressure, and some EEC irregularities
making up the familiar syndrome of tension/anxiety which constitntes one form of CNS arousal. These and
other studies all report similar findings which indicate that the BSR, the number of specific GSRs, and
the amplitude of specific responses when properly interpreted can indicate the general level of CNS arousal.

Individual Central Nervous System Responses: Paving considered indicators of general CNS activity, we
need to turn to individual CNS responses, since a signtficiant portion of CNS activity concerns reflex
responses to stimuli. Many reflex responses are sufficiently complex to involve a major portion of the
suprasegmental CNS. The qualitative and quantitive identification of ongoing reflex response patterns
should contribute greatly to an understand.ng of the !unctional status of the CNS at a particular time.
Those CNS reflexes which have been identified include the adaptive reflex connected with the direction of
a change of stimulus, the defensive reflex :.nresponse to a stimulus too strong for normal functioning,
and the reflex responses per Be, much evide.ce conce-ning central nervous system function can be gardnered
from patterns of evoked responses and contingent effects. These latter responses require a specific
applied stimulus of which the subject is aware and which tends to be distracting or alerting. Evoked
responses may provide valuable guidelines for the interpretation of reflex response patterns observed in
stressful situations or response patterns disturbed by specific activity in the environment.

The Orienting Reflex: The orientating reflex is of particular interest. This reflex, first identified
by Pavlov, has been the focus of an extensive research program in the Soviet Union and has been the
subject of many annual conferences. This reflex is characterized as an unspecific response initiated by
any increase, decrease, or qualitative change of a stimulus independent of its modality. It is really
the "what is it?" reflex of the central nervous system. It only acts to alert and prepare the individual
for action. It does not itself initiate any action and is subject to extinction or habituation quite
easily by repeated presentation of the same stimuli.

Two forr 3 of the orienting reflex have been identified: (1) a generalized orienting and, (2; a
localized orienting. For example, the initial presentation of a tactile stimulus produces a generalired
response including an alpha block in the occipital and motor regions of the cortex, a GSR, an increase
in muscle tension via ENG measurement, an eye movement, and a respiratory pause. After a feO dozen
representations, the only response which may be observed would be a transient alpha block in the motor
region of the ocrtex. Here, the other components of the reaction have been inhibited, transforming the
original reflex picture to a localized or more specific reflex. The total general orienting reflex
picture alsc includes increase in heart rate, vasoconstriction of finger vessels, and vasodilation of the
hand vessels. It is interesting to note that when we change the total sensory input, by adding an
 -pop) -a1 1-IMM.,--- mp

S~158

additional stimulus to the now habituated specific response, the generalized response is once again
elicited. This demonstrates the preadaptive, rather than the adaptive nature of the orienting reflex.

One component of judgement and alertness includes the degree to which the individual is asking
questions of, and interacting with his environment. While the frequency and magnitude of orienting
reflexes may provide valuable indications of this interaction, it is fai.r to state that it is presently
difficult to differentiate these effects from the general level of CNS arousal. Theoretical considerations
and some preliminary reports suggest that it may be possible to demonstrate greater specificity of indi-
vidual CNS reflexes. At least this is the desired direction for further research which is aimed at
permitting clear distinction of CNS arousal to fear, anger, curiosity, and so forth.

EEG Indicators: As we have noted, the identification of the cortical components of central nervous system
reflexes have depended primarily upon observation of alpha block indicating cortical arousal. We have
also noted the presence of distinctive frequency shifts with cortical arousal even when the initial
cortical rhythm is other than alpha, However, unaided visual interruption, or other gross measures of
the EEG, do not permit easy tdentificsý 'l; )F these changes, and thus, increasing attention has be',
focused upon the various methods of examin the EEC in a more microscopic fashion.

The use of toposcopical analysis of EEG records shows interesting contingent effects of so-called
"social" versus "defensive" conditioning of alpha rhythms. Walter was able to demonstrate divergent
changes in alpha when a subject was performing a task in cooperation with the experimenter's instructions,
that is social conditioning, as compared to being thrown on his own resources to solve problems p.ised by
the experimeuter, that is defensive conditioning.

We have explored the cardiac, respiratory and electrodermal indicators of individual CNS arousal and
other activities, and their relationships to each other, and can now turn to a more specific type of
electrocortical activity which promises to give us a great deal of information in assessing human mental
processing activity.

While it is well known and accepted that the task of pilotage and airborne systems controllers has
nhanged dramatically from "seat-of-the-pants" type flying to sophisticated monitoring, pattern recognition
and decision making, we are yet unable to identify, much less quantify, such mental processes. Neverthe-
less, as we have indicated, recent research shows that certain mental acts are related to specific
electronically identifiable wave forms as well as to changes in related physiological parameters. Since
such factors as fatigue, workload and stress (physiologic as well as psychologic) aftect mental performance,
it is highly desirable to be able to identify and quantify Rucla measures. The main thrust of this research
is the identification of specific cortical responses or response patterns evoked by specific stimuli.
These event-related potentials can be characterized as an EEG response wave form, having both positive
and negative values, with certain amplitudes and specific latencies and duration times. In studying these
potentials, a series of positive and negative deflections is averaged for a group of trials. This
characteristic wave-form signature, elicited by a specific stimulus, can be conceptually and empirically
divided into two categories. The earlier components, those occuring in the first 100 milliseconds or so,
subsequent to the stimulus, are referred to as exogenous. These exogenous components, reflect charac-
teristics intrinsic to the stimulus event itself, such as loudness, brightness, intensity or other
psychophysical attributes. This activity is considered to represent the processing of sensory information.
The latter components, up to perhaps 600 milliseconds beyond the stimulus, are considered to •e endogenous.
These endogenous components reflect cognitive processes and attributes of the stimulus deriving not from
its physical properties, but from its task-revelent context. As Lawrence (56) in an unpublished paper
states, "it is these latter components, reflecting aspects of performance potentially applicable to cockpit
or crew station situations which are of primary interest."

Lawrence speculates on a computer controlled workload allocation in a pilotage situation wherein

"the use of brainwaves for the automatic enhancement of warning effectiveness could occur in two ways. A
computer could sense some deficit in an operator's state of being, or potential deficit (anticipating a
crisis) by making inferences about operator state from a set of physiological information channels. The
computer could also observe a lack of attentior to a warning display, os other performance deficit, and
take action to somehow stimulate the human operator. Here, it would make inferences about observed
deficits in operatr performance, probably from &ssessment of ERPs, or their lack, in response to warning
signal displays used functionally as probes. There exists the potential for use of brain wave indicators
of dangerous operator state or behavior in that the presence of theta can predict drowsiness and deteri-
oration of performace and of course the sleeping state can be readily detected. The detection of
undesirable levels of arousal, that is, inappropriately high or low, or undesirable emotional states can
probably be enhanced through the use of EEG information in conjunction with that obtained through other
physiological or behavioral channels. Attention to a display could be assessed by a steady state ERP
technique. ERPs could also be used to distinguish nonresponse to a warning signal resulting from a
purposeful decision to ignore it, from accidental nonrecognition. This way, the hypothesized computer-
based system could refrain from repetition or intensification of information which the operator has
already processed and to which he presumably will eventually respond. Using brain wave information, a
computer could also determine the occurance of an event like target acquisition and utilize this informa-
tion better by employing a built-in control loop than a human observer could by employing a gross skeletal
response such as pushing a button." It must be recognized with the advances of high speed, high perror-
mance aircraft there are circumstances which exist or will soon exist where a few milliseconds could
provide an important advantage. This is especially true if one considers the small single savings in
time and effort that would be accumulated over a rapidly successive series of events in a continuing
recycling context of swift decision making and swift response.

As Lawrence points out, a more proximal goal would be the development of machine ability to sense
such general intagibles as operator uncertainty and the need foxr more information or a need to maintain
certain decision options and an upgrading of information relative to a particular pilot's role in an
X overall mission. Here, instantaneous, qualitative feedback to the, machine could be given in the same way
that varying intensities of temperature guide a miasile toward a heat source. The ability to sense these
variables continuously and sensitively would provide the basis for i-he very fine control of machine by
 159

man, perhaps even along the line of the creation of an artifically intelligent servomechanism so closely
responsive in real time ot the operator's cognitions and perceptions that it could serve virtually as a
functional extension of his own nervous system. It would seem that as we computer assist the functional
machine we must also arrange to computer assist the functioning human being as the operator of that
man-machine system. With this development, the problems of workload, performance, and stress would
undoubtedly be resolved and laid to rest for once and for all.

kEFERENCES

"1. Chiles, W. D. Objective methods for developing indicies of pilot workload, FAA Report (FAA-AN-77-
15), July 1977.

2. ,' R. E. The effect of binaural beats on performance, J. of Auditory Research, 3, 1961,

Pp.. _ 85.

3. ?k:.&.t, R. E., White, D. D. and Hartman, B. 0. Neptune: a multielement task system for evaluating
hum•' performance, USAF School of Aerospace Medicine Technical Report (S.AM-TR 69-25), Brooks AFI,
TX, October 1969.

4. Trumbo, D. Instrumentation in Motor Skills Research, Amer. Psychol., 1969, 24(3), pp. 289-292.

5. Bahrick, H. P., Fitts, P. M. and Briggs, G. E. Learning curves--facts or artilacts, Psychol.

Bull., 1957, 54, pp. 256-268.

6. Kennedy, R. S. and Coulter, X. B. Research note: the interactions among stress, vigilance, and
task cosplexity, Human Factors, 1975, 17(1), pp. 106-109.

7. Demajo, J., Parkinson, S., Leshowitz, B. and Crosby, T. Visual scanning: comparisons between
student and instructor pilots, USAFHRL Technical Report, 1976, June, No. 76-10.
8. O'Donnell, R. D. Handbook of human performance neasures. Unpublished working draft, USAF Institute
"of Technology, Wright-Patterson AFB, OH 1972.

9. Wood, C. A. Neutomuscular correlates of sensori-motor pe-formance in normal and fatigund states,

Dissert. Abs.,, order no. 75-16, 616, 230 pp.

10. Wilkinson, R. T. and Houghton, D, Portable four-choice reaction time test with magnetic tape
memory, Behav. Research Methods and Instrument, 1975, 7(5), pp. 441-446.

11. Gaillard, A. W. and Sanders, A. F. Some effect of ACTH 4-10 or performance during a serial reaction
task, Psychopharmacologia, 1975, 42(2), pp. 201-208.

12. Bartz, A. E. Peripheral detection and central task complexity, H,.man Factors, 1976, 18(1),
pp. 63-70.

13. Salzman, L. F. and Jaques, NON, Heart rate and cardiac cycle effects in reaction time, Percep. and
Motor Skills, 1976, 43(3, pt. 2), pp. 1315-1321.

14. Thackray, R. I., Bailey, J. P. and Touchstone, R. M. Physiological subjective, and performance
correlates of reported boredom and monotony while performing a simulated radar control task, FAA
Office of Aviation Medicine Reports, 1975, HO. 75-8.

15. Holt, W. R. and Brainard, E. C. Selective hyperthermia and reaction time, Percept. and Motor
Skills, 1976, 43(2), pp. 375-382.

16. Singleton, W. T. Deterioration of performance on a short-term perceptual-motor task, Floyd aud

Welford, Symposium on Fatigue, 28:511, pp. 163-172.

17. Sharkey, B. J., McDonald, J. F. and Corbridge, L. G. Pulse rate and pulmonary ventilation as
predictors of human energy cost, Ergonomics, 1966, 9(3), pp. 223-227.

18. Schwarz, J. J. and Ekkers, C. L. Task and load difficulties in directing and regulating a complex
technical system, Mens en Orderneming, 1976, March-April Vol., 30(2), pp. 85-108.

19. Frankenhaeuser, M. and Jchansson, G. Task demand is reflected in catecholomine excretion and
heart rate, J. of Human Stress, 1976, 2(1), pp. 15-23.

20. Montgomery, G. K. Effect of performance evaluation and anxiety on cardiac response in anticipation
of difficult problem solving, Psychophysiologia, 1977, 14(3), pp. 251-257.

21. Nicholson, A. N., Hill, L. E., Borland, R. G. and Ferres, N. M. Activity of the nervous system
during the let-down, approach and landing: a study of short durution high workload, Aerosp. Ned.,
April 1970.

22. Kahneman, D., Tursky, B., Shapiro, D. and Crider, A. Pupillary, heart-rate, and &kin resistance
changes during a mental task, J. of Exp. Psychol., 1969, 79(1), pp. 164-167.

23. Kahneman, D. and Beatty, J. Pupil diameter and load on memory, Science, 1966, 1Y4, pp. 1583-1585.
160

24. Dukes-Dobos, F. N. Fat:tgue from the point of view of urinary metabolites, Methodology in Human
Fatigue Assessment, Haalimoto, Kogi and Grandjean, ods., Taylor and Francis, London, 1971.

25. Von Euler, U. S. Adrenal in and nonadrenalin in various kinds of stress. Symposium on Stress,
Washingon, D. C. Army Medical Service Graduate School and Walter Reed Army Service Center, 1953.

26. Hale, H. B., Williams, E. W. and Buckley, C. J. Aerospace aspects of the first non-stop transatlantic
helicopter flight, Aeropp. Hed., 1969, 40, pp. 718-723.

27. Selye, H. Stress, Act& Inc. Medical Publishers, Montreal, 1950.

28. Shannon, I. L., Pregmoir, J. R. and Brooks, R. A. Glucose concentrations in parotid fluid and
blood serum following intrasvnous glucose loading, Oral Surg., 13:1010, 1960.

29. Warren, B. H., Ware, R. W., Shannon, I. L. and Leverett, S. D. Determination of inflight biochemical
responses utilizing the parotid fluid collection technic, Aerosp. Med., August 1966, p. 796.

30. Basmalian, J. V. Muscles alive: their function revealed by electromyography, The Williams and
Williams Co., Baltimore, 1974.
31. Lafevers, E. W. Power spectral. density analysis of the electromyogram from a work task performed
in a full pressure suit, Dissazt. Abs., 1974, no. 75-1033, 82 pp.

32. Grandjean, E. and Kozi, K. Introductory remarks, Methodology in Human Fatigue Assessment, Haskimoto,
Kogi and Grandjean, eds., Taylor and Francis, London, 1971.

33. Wolf, G. Construct validation of measures of three kinds of experimental fatigue, Percept. and
Motor Skills, 1967, 24, pp. 1067-1076.

34. Saito, Y., Kogi, K. and Kashiwag:;, S. Fractors underlying subjective feelings of fatigue, J. of
the Science of Labor., 197U, 46, pp. 205-224.

35. Hess, W. R. Die funktionelle organization des vegetativen hervensystems. Basle: Benno Schwabe, 1948,

36. Kcgi, K. and Saito, Y. A factot analytic study of phase discrimination in mental fatigue, Methodology
in Human Fatigue Assessment, Haskimoto Kogi and Grandjean, eds., Taylor and Francis, London, 1971.

37. Ettema, J. H. and Zielhuis, R. L. Physiological parameters of mental load, Methodology in Human
Fatigue Assessment, Haskimoto, Kojgi and Grandjean, eds., Taylor and Francis, London, 1971.

38. Kashiwagi, S. Psychological rating of human fatigue, Methodology in Human Fatigue Assessment,
Haskimoto, Kogi and GrandJean, edn., Taylor and Francis, London, 1971.

39. Storm, W. F. Hipenney, J. D. Mis:sion-crew fatigue during rivet joint operations, USAF School of
Aerospace Medicine TR-76-36, 1976.

40. McKenzie, R. E. A systems task used in the stress testing of special mission personnel, Human
Factors, December 1965.

41. Welch, R. B., Longley, E. 0. and Lomaev, 0. The measurement of fatigue in hot working conditions,
Methology in Human Fatigue Assessment, Haskimoto, ' and Grandjean, ads., Taylor and Francis,
London, 1971.

42. Grivel F. The influence of ambient and body heat on human without important physical load: II.
specific heat stress effects evidenced in laboratory studies since 1958, Travail Humain, 1975.
38(2), pp. 223-244.

43. Bartlett, F. Psychological criteria of fatigue, Symposium on Fatigue, W. F. Floyd and A. T. Welford,
eds., H. K. Lewis & Co., London, 1953, pp. 1-5.

44. Burch, N. R. and Greiner, T. H. A biolectric scale of human alertness: concurrent iecordings of the
EEG and GSR, Psychiat. Res. Rep. Amer. Psychiat. Assn,, 12:183-93, January 1960.

45. Luby, E. D., Grisell, J. L., Frohman, C. E., Lees, H., Cohen, B. D. and Gottliebb, J. S. Biochemical,
psychological, and behavioral responses ro sleep deprivation, Am. N.Y. Acad. of Sci., 96:71-9, 13
January 1962.

46. Burch, H. R. Automaatic analysis of the electroencephalogram: a review and classification of systems,
Electroenceph. Clin. Neurophysiol., 11:827-34, November 1959.

47. Riehl, J. L. Analog analysis of EEG activity, Aerosp. Med, 32:1101-8, December 1961.

48. Heymans, C. Reflexogenic areas of cardiovascular system, Perspect. Biol. Med., 3:409-17, Spring
1960.

49. McKenzie, R. E., Buckley, H. P. and Solaranis, K. An exploratory study of psychophysiological

measurements as indicators of air traffic control sectors workload, Federal Aviation Agency
Memorandum Report #157-524-03R., Washington, D. C., June 1966.

50. Johnson, L. C., Ulett, G. A., Sines, J. 0. and Stern, J. A. Cortical activity and cognitive
functioning, USAF School of Aerospace Med., 60:75:1-14, October 1960.
 161

51. Donchin, I., McCarthy, G. and Kutas, M. Electroencephalographic ir.vestigations of hemispheric

specialization, Language and hemispheric specializs':ion in man: cerebral ERP's, Prog. Clin.
Neurophystr., 3, J. E. Desmedt, ed., Karger:Basel:1977.

52. John, S. R. and Schwartz, E. L. The neurophysiology of information processing and cognition, Ann.
Rev. Psychol, 1978, 29, Pato Alto: Annual Reviews Inc.
53. Walter, D. 0., Advances in EEG analysis, Electroencephol. and Clin, Neurophysiology, Supplement
No. 27, 1966.

54. Ax, A. and Luby, K. D. Autonomic responses to sleep deprivation, AMA Arch. Gen. Psychiat.,
4:55-9, January 1961.

55. Levy, E. Z., Johnson, G. E., Serrano, J. Jr., Thaler, V. H. and Ruff, G. E. The use of skin
resistance to monitor states of consciousness, Aerosp. Med. 32:60-6, January 1961.

56. Lawrence, G. H. Brain waves and the enhancement of pilot performance. Unpublished manuscript
prepared for the Environmental Physiology Program, Office of Naval Research, Washington, D. C.,
June 1978 (Submitted to AGARD AMP WG-08).

ACKNOWLEDGEMENTS

I wish to express my appreciation for the cooperation of all of our authors, especiially Doctors
Gartner, Murphy, Buckley and Captain Perelli who gave me permission to abstract from their basic works.
I hope that any shortcomings or criticism of these particular chapters will be directed to the editor
and not the original sources.

I also wish to thxnk Diana L. Deyslc (AlC) and Robin G. Cha"ez (AlC) who did much of the libr'ry
research required for Chapter 12.

Finally a special thank you for the secretarial services of Joyce Keller, Jeanette Jonietz, and
Debra Coronado who all shared in a difficult cask.

I hope our readers and the members of AGARD AMP WG-08 will recognize the outstanding support in
personnel and services given by the United States Air Force School of Aerospace Medicine.

'k,
 163

SUMMARY

The first three chapters provide a conceptual framework for workload, fatigue, and stress, within
which to evaluate the remainder of this report. In each case, the authors attempted to be brief, to
present a "capsule" statement of different definitions and orientations, and to the extent possille to
prevent their own biases from entering into the text. What is the probability that all readers will be
fully satisfied with the contents of the three chapters? Probably minimal, but hopefully few readers
will be grossly dieaatisficd.

The next three chapters, taken as a single unit, give a picture of the workload arena in a broad
sense, partly historical and partly in terms of specific sub-problems and suggestions regarding selected
methods or measurements. These chapters, therefore, augment the conceptual framework provided b"- :he
first three chapters.

Chapters 7, 8, and 9 come to grips in a concrete way with Lhe critical issue of the anatoa.y of work-
load measurement lechnology. Chapter 7 provides a schema (a generalized representation on framework of a
topic or problem area derived through an analytic but pragmatic process) for workload research. Chapter
8 describes a moderately less encompassing bit still global program design appliel to workload problems
by one laboratory. Chapter 9 presents one modelling approach to workload--there are others, of course,
as the author points out. As an aside, Chapter 9 is also a "preview" of an AGARDograph which the Aero-
space Medical Panel is considering sponsoring in the near future. These three chapters are recommended
particularly to laboratory directors, program directors, and supervisory scientists as tools for evalua-
tion and goal-setting in their own programs in workload research.

Chapters 10 through 18 deal with selected measures applied to specific problems in specifiedl settings.
The first six are concerned with aircrew studies and the last three with air traffic control 6tudies.
There are many such sets of studies which could have appeared in this part of this report. These appear
because they were offered and because we, the editors, valued both the investigator and the work he
reported. In each case, the reader will be able to see how one investigation approached one specific
problem using his own skills and the resources available to him. The virtue of thin !s that it lets the
reader move from "frameworks," "schemas," etc., to concrete examples.

Chapter 19 stands by itself in this document. It is a modest conpondium in which some measures from
some domains (e.g., psychophysiology) are described and critiqued, the c:ritiques clearly influenced by
the skills, experiences, d biases of the author. The term "modest" is used to make a point. A compen-
..
dium like this could be . probably useful and certainly very long handbook--probably two or three weighty
volumes. The working .roup initially tasked itself with this objectivw, proposing to use a draft hand-
book offered by a US colleague, but it became apparent very early that the tak was beyond the working
group's capabilities (time, level of effort, etc.). This might be a 'ýseful :uu task for Aerospace
Medical Panel sponsorship, though probably not in the conventional wvck1, .8',j p mcde oi operation,

Two points should be made in concluding. First, all papers after 4 c)ntain pieces of
studies, some data, analyses, findings, and so forth. The editors bdli,.,e thiý eyiriches the more global
parts of each chapter. Second, there are references given at the er,n of each chap:ter. Taken together--
as a package--these constitute a highly useful bibliography.

Workload measurement methodology is in a continuing process of unfolding, acquiring new technology

and instrumentation, moving into new measurement domains. This topic should be revisited by AGARD after
a reasonable period of rest and recovery.

h1
 REPORT DOCUMENTATION PAGE
I. Recipient's 2.Orilginatcr's Reference 3. Further Reference 4. Security aa.ifkcation
SofReference Document
•'AGARD-AG-246 ISBN 92-835-1332-0 UNCLASSIFIED

S.Orisinator Advisory Group for Aerospace Researcb and Development

North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuily sur Seine, France
6. Title
SURVEY OF METHODS TO ASSESS WORKLOAD

7. Presented at

8. Author(s)/Editor(s) 9. Date
Edited by B.O. Hartman* and R.E. McKenziet August 1979

1O.Author's/Editor's Address *Crew Technology Division, USAF School of l1.Pages

Aerospace Medicine, Brooks Air Force Base, Texas, 78235, USA
172
tSan Antonio, Texas, USA
12. Distribution Statement This document is distributed in accordance with AGARD
policies and regulations, which are outlined on the
Outside Back Covers of all AGARD publications.
13. Keywords/Descriptors

Military aircraft Work sampling

Pilots (personnel) Motor reactions
Work measurement Senses
Performance evaluation

14.Absact
Military aircraft are becoming increasingly complex, the associated avionics systems more
sophisticated, and the mission profiles more demanding. The problem is to establish if
such an increase in aircrew workload has become a limiting factor in the operational
employment of some aircraft and to select valuable methods to assess it.

The measurement domain has been broken down into sensory threshold function tests,
motor function and responses to psycho, physio and chemical excitation. The methodology
includes a wide range of instrumentation, laboratories, inflight measurement and modelling
methods, with the goal of cor.ipiling systematically and evaluating the multiplicity of
approaches and techniques implied.

Thts preliminary survey is followed by a companion document (AGARD Advisory Report 139
- AR-139) where conclusions are set forth as far as workload measuremeit methodology is
concerned.

This AGARDograph was sponsored by the AGARD Aerospace Medical Panel.

.' .• i
[4 >.0

.0j 0 go O. ýz

'SO i 00
C)

0,~ 0000
r

0. 00 0

00 U,70.

0 0 00

Cd',

5. rva-

A b

0 01 ch C~ ol ~
R o c
CL 0 0
0 0 ja 0
10O (Pa
 ~%)0 0

CoC o c"
E~ E co

0 0.. ~0 40

0 *= - ' -0

04),
-U CU

Cu Cc
= 0CU .....
2 2.

-w 0.

4)

0 V0~UU0.>

05-. UC 0C aU

CS, en iao

(A en

0~ '0
E- e e

0~ CdU

E E
0S0 CE
0.9
VV~C
U2 0

cc Cd
E V

4. CC.0~. 120

-)U
20'-E 0d
0
~SS. U CO 444)
 Aý00

rf .j'd pjb
-12 =I 0 C

0 0t

cC 0f to

00 0,:s A. -3 s -g

c)

0gd~ 0 A
0~~ 0

00
0~ 0L

O O~ e,

t3 VJ

_____________________________. 5 ________________
 K
j
NATO *~OTAN

7 RUE ANCELttt- 022M MJALt$!C'W*UCAE-)flS

Tote-Mona ?45X1W¶&)- ~To~x '

?ýGARD does NOT hold stgvka of 444Z0oMt) >i at fo p# th.Z-dl iMOM

?otaak nS 40 f AtAfi)
p..blicatlona is ,w-le to J$2AR1) Mewihm Nstinf1• Clkrt%ýA h~e, iYaA ~peu e tna,Ptt CoraD' 06r6 4ne
vay'table frrni tl-& Ctntta4,, bill if ifl itoay be r~AKa~in MWka.&e or 1norcreev form frawn f~la jinchas Apmwclti U4s*n tnt'

BELC~I'JM . 1 A4LY
Cnlrdonra~ieu %GARI- " - V$L .Atrlttatirwg k~jiaive It74f
11fIe lt I.W-vo.rýWuPrzondrt $VAC;A~lDI

U
'EftxMjr e Force flt~llflC
o/ritr ace EliuMocWi FL,
iril Adenaaan
Rti 4, Evezr 140 Rruv4lc orz/11
CVM! ADA LUXENICIOUR
Deflwne S-icntific !)tfrmn'tlM¶ S-rMCýt -See ftlcirnu
l~pasnwn~of
Pcfa'zcNElIMERIANDS
Ntb~na
Ottaw, OUUUCA tAAOZ2N~hedantis AGMRD
IWet'tion? to
WK enlADfne eerK I National Atrnouapc Laibomstory, NIR
Deenc ,'Iseech
baubi '"tdP.O. dox 1261
s1tsrbrogadsn Karrn4If
~NORWAY
Cope~j~gn
FKANICE Nt~seegian LOefect Rmiar', Estebls~nicnt
O.N.E.R.A. (D&'ec'tion') MK;n Llbruy
92 CNW~on sous 5apteux N-2007 Kplkir

GERMANY Pl UA
/.en'nr*Itue flur L4111. uttd Rrur3ffiahn. D'rocqwo do. ;CtYlneatI
dokumvnutaion uriC. -infcnrration d%F x Aca rep
c/o Fachinfurrnutio'Gzentrlan EaieT:e, Rue da Ewo.-ia Pvfltcnica 42
Physik, MrdtnmaZii Gmbh Usbvu
KernxvHngsznil~'v .tn: AGARD National Deiegiltr
7itrlil
CUTECE Deparlment of Research anid Eevokpinent (ARCE)
01nic
dI Air Fewe, Gvp~ie "ti Sff Mnistry of National IMPfrioe, Ankaia
heeeurch idJ Devek~rprbnrt Div,Avctotat NT& OGI
1k
reecefdonce
)F55, Atens, Resewrr infi.-aatior. Centre
ICELANO Statieon varc hl-t'se
[imnto .9f Aviation tMLyCa
(;!o Flugrad Otig~akent 8.R5 31W
ReykavikUNY17H) STA T2'S

AIM- Report Discrf itiion llnd. Sco&4rg Unit

THlE LN'NIM NSTATCS ;4AMiQNAL WTIMSr~BUTlON C:ENT!?J-(NAZ.;A) Y.FS WIT H0. C j
STGCX$ GP ACARDT PU BLICATIONS, AND APPLICATIONN FORK COPIE1-S 514042E$C ir
bit DIAW.
DD&i-t.T I0 Th'AYIONAI, TECBýN:1CAL iOA*ONSERVIC E 'NTS 1 A17 THE ADDRESS likiuqw.

National Tecvhetil spaC6I-c um.senal t'i-o ;*rvicc Teichriologv Repels3

Inforlnntion Scrvce (NTEJ) &?,p1nD
10'nplY Cenil,r (Uft)
5285 Ponrt~ Ru&4 1.9,1s,'M\oN j St lug(T i t
Seua
Via~n~ 2 ~ UOr pitngIon. Kuit BRýI3KF

ReqAwat: fu ~~'clh ,~j'ur ot wl' AIRAI.IN.u r!.A . 11wmc

r. A-nA$!Ofj
'e adet
-U)Ie,.
C, iethoi, or editor, ,iJ
ptbVlcatick ditt. Requem {o N4'*4svtS
sh ulIdt-M
!LINiCIt 1AA iaCalftum Wepu!1 nVJmbCIete W17&kdh {! N$,. a Iw ItV'IZ
~44 NARLIptil'cA~osA tCffvent in 6,e f&;y''iRr' pw)xnais

S#6vtinlfi; iwli fcu',.'Iaw .t$,'/ Rcjott't iSTAR; Ghh-enriniL't Reqpciwl , Iujh;newtzin (CIU1,
pVbir''t
IXkv 41SA
67,' cvr7a~n pr~~tx . tcic~

k(ArylnawS 23)1§U~
^;

-wat w"fN