Cognition, Technology & Work (2002) 4:229–239

Ownership and Copyright

Cognition
# 2002 Springer-Verlag London Limited Technology &
Work

The Effect of Workload and Workshift on

Air Traffic Control: A Taxonomy of
Communicative Problems
P. Corradini and C. Cacciari
Department of Biomedical Sciences, University of Modena and Reggio Emilia, Modena, Italy

Abstract: Communication is crucial for air navigation safety: communicative problems are implied in 70% of aviation accidents and incidents. This study
investigated the influence of workshift (backward rapid rotation) and workload on air traffic controller (ATCo) communications: a taxonomy of possible
communicative errors and incorrectness was designed and a specific grid proposed to analyse the communicative exchanges taking place during the
workshifts and under different workloads. The corpus we used to design and test our taxonomy and obtain measures of the communicative performance of
ATCos consisted of 10 hours of radio exchanges between tower and approach controllers and pilots in an Italian airport. Results showed that the taxonomy
was indeed apt to capture a variety of communicative problems: controllers widely employed a linguistic code strongly deviating from standard phraseology,
with a widespread presence of Italian language, of non-standard expressions, ellipses and redundancies. Shiftwork and workload significantly affected the
ATCos’ communicative performance: linguistic deviations significantly increased during the nightshift with a low workload, while the most correct
exchanges occurred in the morning shift.

Keywords: Air traffic control (ATC); Aviation psychology; Communication; Human factors; Workload; Workshift

1. INTRODUCTION communicative processes were reported in more than 70%

of air accidents and incidents (Billings and Cheaney 1981;
Communication is crucial in air navigation, being essential Kanki and Palmer 1993; NASA-ASRS 1994; Cushing
for accomplishing at least two types of tasks (Kanki and 1994).
Palmer 1993; Andriessen 1995): Air traffic control (hence forth ATC) activity is
organised in shifts that cover the 24 hours with a volume
1. Interpersonal tasks related to a variety of different areas:
of air traffic that can massively vary in time. Workshift and
team formation and resource management, co-ordina-
workload affect the well-being and the cognitive perfor-
tion, tasks and workload distribution, decision making,
mances of human operators in a variety of organisational
situation awareness maintenance, development and
settings (Andlauer et al 1977; Reinberg et al 1981; Beugnet-
updating of mental models and of predictable beha-
Lambert et al 1998; Stein 1988; Costa et al 1995). In a
vioural patterns (Orasanu and Fischer 1991, 1992;
recent study carried out on the same population of air traffic
Orasanu and Salas 1993; Kanki and Foushee 1989).
controllers (henceforth, ATCos) that we tested, Cicogna et
2. Technical/operational tasks related to the execution of al (1998) confirmed that workshift organisation influences
procedures and to the interactions with technical tools the chronopsychological characteristics of ATCos, namely
(Foushee 1984; Ginnett 1993).
their level of vigilance, their sleep cycles and other
Work team efficiency is highly dependent on commu- functions that show cyclic variations in time. ATCos
nicative abilities: studies on work teams in air navigation vigilance level during the workshifts varied differently from
environments showed a relationship between co-operative the rest days: while during the rest days the vigilance
processes and team performance from the one side, and the followed a physiological cycle with periods of decrease (a
quantity and the quality of communicative exchanges slight one between 1 and 3 p.m. and a heavy decrease
among the team members, from the other (Kanki and between 1 and 7 a.m.), during the workshifts the ATCos
Foushee 1989; Kanki et al 1991; Andriessen 1995; Bellorini almost always maintained a high level of vigilance, with the
and Vanderhaegen 1995). Problems in co-operative and only exception of the night hours (from 2 a.m. to 5 a.m.).
230 P. Corradini and C. Cacciari

The effect of workload on ATCos, according to about the traffic situation and to update their mental
incidents and accidents databases, can be represented by representation.
an inverted U relationship (EATCHIP 1997). In fact, The language used is a highly formalised code defined by
controllers’ performance level decreases: ICAO norms, an operative standard phraseology based on the
English language (a technical jargon) and defined by
1. during periods of high workload;
specific messages, sequences, formats, terminology and
2. in the periods of time following a high workload phase, pronunciation rules. The phraseology has been elaborated
because of a decrease in vigilance after a stressing in order to possibly reduce ambiguities and misunderstand-
activity, and when a high workload period follows a ings, to allow a strict control over message length and
light workload one because it requires a sudden re- structure and to reduce radio frequency congestion and
concentration on the task; ‘noise’. In fact, it has been observed that frequency
3. during periods of very low workload, because of a congestion is one of the major problems in ATC
decrease of attention or of controllers’ overconfidence. (NASA-ASRS 1994), also because of the ever increasing
amount of air traffic.
Despite the evidence suggesting a severe impact of work-
According to the phraseology, the controller’s message
shift and workload on the cognitive performances of
(see Example (1)) begins with a callsign (1) (an alphanu-
individuals, only a few studies investigated the ways in
merical string used to identify any specific aircraft) of the
which these two factors affected one of the most crucial
aircraft contacted followed by the name of the Control
aspect of ATC, namely the communicative exchanges
Centre (2). A set of instructions follows, needed by pilots
between ATCos and pilots (Corradini and Cacciari 2000,
to organise the flight and to effect specific manoeuvres
2001).
(clearances (3)), and information (4). The pilot acknowl-
In this study we proposed a taxonomy for analysing the
edges the reception of the message (acknowledgement (5))
communication performance of ATCos and applied such a
and repeats it (readback (6)). The controller verifies the
taxonomy to a large sample (10 hours) of communicative
correctness of the readback (this practice is called hear-
exchanges that took place in different workshifts and with
back), possibly corrects errors and finally gives an acknowl-
diverse levels of workload (operationalised in terms of the
edgement of it. Each message should contain the callsign of
volume of traffic managed by a controller). We aimed to
the aircraft contacted. For instance,
test whether these two organisational factors had an effect
on the quality of the communicative performance of (1)
ATCos.
C: Alitalia 1186 (1), Ciampino Tower (2), cleared to take-
off runway 12 (3), wind 40 degrees 7 knots (4).
1.1. The Structure of Air Traffic Control
P: Roger (5), cleared to take-off runway 12, wind 40 degrees
7 knots (6), Alitalia 1186 (1).
ATC has two objectives: to ensure the safety of flights and C: Correct (5) Alitalia 1186.
the efficiency of traffic. Therefore, it provides aircraft
assistance to take-off, navigate and land according to
Deviations from this phraseology and from this system of
ICAO norms, by means of information and instructions
cross-repetitions weaken the defences against misunder-
given orally via radio and automatically via technical tools.
standings and represent a possible source of miscommuni-
The airspace is organised in different areas controlled by
cation and ambiguity.
different Control Centres: the ‘control zone’ is the area
surrounding an airport and is managed by the Control
1.2. Miscommunications and Ambiguities in Air
Tower, mostly responsible for take-offs and landings. Above
Traffic Control
this, and with a wider extension, is the ‘control area’, under
the jurisdiction of the Approach Control Centre; in the
Ambiguity is a major cause of miscommunication in ATC
upper space there are ‘airways’, i.e. defined air routes,
(Cushing 1994; Corradini and Cacciari, 2000, 2001;
controlled by the Route Control Centre. While operators
Bellorini et al 1998). Ambiguities in pilot–controller
usually perform a visual control in the control tower, they
communications can occur at different levels (Cushing
are supported by a radar in the other Centres.
1994):
Controllers and pilots communicate via radio (each
Centre is assigned a different radio frequency) creating a 1. Phonetic/perceptual ambiguity: words, nouns and numbers
communicative network called ‘party-line’: there all aircraft homophone or nearly-homophone can cause semantic
can listen to all verbal exchanges with the control centre misunderstanding or confusion relative to the identity
while only one pilot can speak with the controller in any of the addressee (Grayson and Billings 1981). In this
given moment. This allows the pilots to monitor all the latter case, a similarity between the callsigns of two
messages exchanged and so to collect more information different aircraft might produce an interference be-
The Effect of Workload and Workshift on Air Traffic Control 231

tween that actually uttered by the controller and that miles, level at 6000, to pass under you’, meant to
relative to another aircraft. As a consequence, a pilot communicate information, was interpreted as a com-
might consider to be the addressee of a message that was mand. The controller meant ‘level at 6000’ as a
instead addressed to another pilot. In fact, in many of declarative statement, while the pilot interpreted it as
the errors we observed the confused callsign shared at an imperative one (Cushing 1994).
least the company name or the numerical string, when
not both of them. Carrier companies usually attribute 1.3. A Taxonomy of ATC Miscommunications and
callsigns only differing for the last letter to the aircraft Ambiguities
following the same route (e.g. KLM254a and
KLM254b). Controllers often shorten the callsigns The taxonomy forming our grid (see Appendix 1) was
eliminating either the company name or the numerical designed to detect the communicative phenomena relevant
string and the readbacks are not always effective in for ATC. We identified a set of basic categories on the basis
disambiguating the miscommunication. Confusion be- of the following sources: (a) the vast literature on ATC
tween numbers in parameters can happen as well, also communication (Morrow et al 1991, 1993; Seamster et al
because the numerical ranges of different parameters 1992; Cushing 1994; NASA-ASRS 1994); (b) a previous
such as heading, speed, radial, flight level might study where we interviewed both controllers and pilots
partially overlap (e.g. ‘heading 225, climb to 325’ can (Cacciari et al 1997; Depolo at al 1997); (c) observations
be understood by the pilot as ‘heading 325, climb to on the field. A description of our taxonomy follows, with
225’). These confusions are more frequent when the examples relative to each category presented in the Results
controller’s messages are long (Morrow et al 1993) and section. We pooled together some of the subcategories of
when the messages contain many parameters, presum- the main classes described below, because too few cases
ably because of an increase in the memory load. Such were found for the statistical treatment of the data. We will
confusion can also be enhanced by the practice, specify the categories that were pooled together.
common for both controllers and pilots, to omit words The taxonomy was structured in the following cate-
that specify the parameters (in the previous example, gories:
the words ‘flight level’). 1. Organisation of the transition: as suggested by
2. Lexical/semantic ambiguity: this kind of ambiguity is Morrow et al (1993), this category is used to
essentially due to the absence of the standard describe the structure of the communicative
phraseology or to deviations from it. Controllers and exchange with respect to what is predicted by
pilots often do not use a homogeneous language in their ICAO norms. It includes two subcategories:
communications: although they mostly employ the 1.1 Routine transitions where no misunderstanding or
formal phraseology, there are frequent intrusions of linguistic deviation from ICAO norms occurs;
informal expressions and of the natural language that is,
1.2 Non-routine transitions that contain some form of
by definition, ambiguous (Falzon 1982). This is
misunderstanding and are followed by speech turns
particularly true in two cases: (a) non-routine situations
of clarification, not predicted by the standard
that require the operators to update and revise their
procedure.
individual and collective cognitive models (Orasanu
This category does not include the transactions
and Salas 1993; Orasanu and Fischer 1991, 1992); and
containing linguistic expressions violating the
(b) when the speakers share a different amount of
correct phraseology but not producing misunder-
knowledge not being aware of this knowledge disequili-
standings.
brium. Misuse of the phraseology and code switches can
also be due to an insufficient knowledge of the 2. Problems with the medium (radio): difficulties in
technical jargon, or to cross-cultural communicative communicating due to disturbances in the radio
problems (e.g. prosodic errors, poor knowledge of contact. As no relevant radio problems were
English). Miscommunications of this kind were among encountered in our corpus, this category will not
the causal factors of several air disasters, for instance be discussed further.
those of Tenerife in 1979 and of Cove Neck in 1970 3. Lack or misuse of the ICAO phraseology: this is a very
(Cushing 1994). large category including all the cases where the
3. Pragmatic ambiguity: qualitative aspects of spoken operators do not use, or only partially use, the
language, such as the pauses and the intonation, phraseology. We distinguished four subcategories:
might produce interpretative ambiguities. The illocu- 3.1 Cross-linguistic factors: the controllers do not always
tionary force of a message, for instance, might be employ the English language, as prescribed, but
misunderstood as in the following example: a message instead use:
uttered by a controller, ‘Traffic at ten o’clock, three 3.1.1 The Italian language;
232 P. Corradini and C. Cacciari

3.1.2 A mixed language, where they shift from Italian to firming the ‘hearback’ is not strictly prescribed after
English and vice versa. each pilot’s turn).
Since we did not find any language shift within a Here is an example of different kinds of
speech turn (i.e. all the speech turns were uttered omissions (the omitted words are in brackets):
in one language), we eliminated this category from
(4)
our analysis.
3.2 Use of non-standard expressions: the expressions do C: [callsign] [wind] One hundred degrees, fourteen
not belong to the standard phraseology and their knots, runway one two; visibility ten kilometres,
meaning is differently shared by the ATCos; more [temperatures] twenty with four, H one zero
precisely, we distinguish between: twelve, transition [level] seventy-five.
P: [QNH] one zero twelve, [transition level] seventy-
3.2.1 Collective non-standard expressions: these are five, [runway one two], all copied, thank you
broadly diffused and almost conventionalised [wind] [callsign]
because of their frequency; e.g.:
(2) 3.3.3 Callsigns: they can either be completely absent, or
shortened, for both the station sending and the
Collective non-standard Standard phraseology station receiving the message. For the Control
routine
Centres, the geographical name or the station
OK Roger name (‘Tower’ or ‘Radar’) can be omitted; for the
Yes Affirmative aircraft, the omission can concern the company
No Negative name or the numerical string. Nicknames are often
Down to 220 / Descend to flight employed, as well as in the following examples:
Go down 220 level 220
Ten eighty / One zero eight zero (5)
One o eight o / (1080)
Shortened callsign Correct callsign
Ten eight zero
1225, cleared to . . . KLM 1225, cleared
3.2.2 Individual non-standard expressions: expressions to . . .
recurrently employed only by a single operator, i.e. AirFrance, climb FL 320 AirFrance 309, climb
individual ‘reinterpretations’ of the standard forms FL 320
or common language phrases tout court; e.g.:
(6)
(3)
Nickname Real name
Individual non-standard Standard phraseology
routine MerAir Meridiana
Speedbird British Airway
May I have 220? Request flight level 220

The name of the Control Centre is considered as

These different subcategories of non-standard absent only if it is not uttered in the opening turn
expressions were pooled together for the statistical of a transaction both by the pilot and by the
analysis. controller. The aircraft callsign, instead, is scored
3.3 Omissions: ellipse of words of phrases that should be as absent each time it is not repeated in each turn,
uttered. They can concern different elements: according to the standard procedure meant to avoid
3.3.1 (a) Clearances, (b) information (about meteorolo- ambiguities and confusion among different air-
gical conditions, take-off/landing time, aircraft crafts.
position), (c) parameters (such as altitude or In the final analysis, we pooled together the
speed), (d) words or numbers (such as ‘QNH’, omissions of clearance, information and acknowl-
‘flight level’, ‘temperature’ and so on). When the edgement. We kept, instead, a separate category for
omissions are routine we classify them as non- incorrectness in callsign because of its relevance
standard expressions as well. and incidence; we only pooled together complete
3.3.2 Acknowledgement: we distinguish between cases and partial omissions of callsign.
in which a controller does not overtly correct a 3.4 Redundancies: these are represented by all the
communicative error made by the pilot and cases expressions not strictly necessary, as follows:
where he or she does not confirm the readback of a (a) repetitions of information or instructions;
previous correction (the acknowledgement con- (b) politeness formula (e.g., thanks, greetings);
The Effect of Workload and Workshift on Air Traffic Control 233

(c) negotiations (i.e. more or less explicit requests 1. The work position, i.e. tower vs. approach (radar)
such as changes in the timetables or in the flight controllers.
plans). 2. The workshift. The work was organised in a shift system
Examples of redundancies are shown in the within a 6-day cycle: afternoon (1 p.m.–8 p.m.);
following box (the words in italic should have been morning (7 a.m.–1 p.m.); night (8 p.m.–7 a.m.).
omitted):
3. The workload, i.e. maximum vs. minimum number of
(7) aircraft per hour controlled by an operator in one shift,
on the basis of the traffic scheduled at the airport (e.g.:
Politeness formula Good morning to you, X105, the sample ‘tower-morning shift-maximum workload’
Useless locution for information you’re number
stands for 10 minutes of tower controller–pilot com-
four in sequence
munications recorded during the hour of maximum
Negotiation Tower, X1255, we start previewed expected workload in the shift from 3 a.m. to
pushback at our own discretion 11 a.m.). The mean number of aircraft scheduled and
the standard deviation (in parentheses) per hour in
These different kinds of redundancies were pooled each condition are shown in Table 1.
together in the statistical analysis.
4. Confusions of homophones/interferences: these are Table 1. Mean number of aircraft handled per hour in each condition
alphanumerical strings confused because of their Workshift Workload
phonetic similarity. We consider three cases of
Minimum Maximum
confusion:
Morning shift 7 (0.7) 14.6 (2.5)
4.1 Confusion of numbers (flight parameters, meteor- Afternoon shift 7.6 (1.7) 13 (1.9)
ological information, distances); Night shift 1.2 (0.5) 11.6 (2.7)
4.2 Confusion of callsign or of part of them;
4.3 Confusion of words.
We recorded 12 samples of 10 minutes for each of the 5
We included in this category also confusion due to a days, one sample for each condition (position 6 workshift
phonetic interference rather than to homophony, such as 6 workload), so that we eventually obtained five samples
the transposition of numbers inside a string. for each condition (50 minutes) and 10 hours communica-
We turn now to describing the results of the study in tions sampled in total; the number of speech turns analysed
which we tested this taxonomy. As previously reported, the in each condition is reported in Table 2.
general aim of this study was to investigate the role played
by workshifts and workload on the quality and appropriate-
Table 2. Total number of controllers’ speech turns analysed for each
ness of the communicative exchanges of ATCos. condition

Workshift Workload Position Total

2. METHOD Tower Radar

2.1. Subjects Morning shift Minimum workload 67 49 116

Maximum workload 101 135 236
Afternoon shift Minimum workload 73 53 126
The subjects were 18 male ATCos (11 radar controllers, Maximum workload 62 116 178
hence forth RCs, and 7 tower controllers, hence forth TCs) Night shift Minimum workload 16 49 65
who volunteered to participate in this study. They were Maximum workload 94 97 191
aged from 22 to 49 (mean RC age: 38.7 years, 1 RC under Total 413 499 912
30; TCs: 31 years, 6 TCs under 30); the mean length of
service was 18 years 10 months for RCs and 5 years 5
months for TCs. The samples of communications were first transcribed
verbatim and then revised with an expert (a long-
2.2. Procedure experienced controller). They were analysed using the
taxonomy we have just presented (see Paragraph 1.2.3. and
The pilot–controller radio communications forming our Appendix 1). We calculated the frequency of occurrence of
sample were recorded at the Control Centre of a medium each category in each of the conditions over the 5 days,
sized airport in northern Italy over 5 days considered as separately for controller and pilots. Unfortunately it was
representative of the average traffic of the airport. They not possible to assess the workshift schedule and the
were sampled according to the following factors: workload level of pilots. Hence we concentrate only on the
234 P. Corradini and C. Cacciari

relation among shift work, workload and communicative and incorrect callsigns in 67% of them, F(2,897) = 9.470, p
characteristics relative to ATCos. We calculated the < 0.0001). (see Table 1).
number of speech turns that each operator uttered in In particular, these results show a peak in linguistic
each condition and used the total number of speech turns deviations in the communications produced during the
as the basis for calculating the mean frequencies for each nightshift in the period of minimum workload that
category (frequency/total number of speech turn produced occurred between 2 a.m. and 4 a.m.. This corresponds to
in the relative workshift, separately for controller and the maximum decrease of controllers’ vigilance and to the
pilots). For instance, we found 13 redundancies produced highest level of mental fatigue. In this condition, messages
by a tower controller in the morning shift with a low were more incorrect in that they presented more Italian
workload, with the total number of speech turns he uttered words (in 78% of the speech turns), incorrect callsigns (in
in that condition of 67. Hence, the mean value for the 73% of them) and collective and individual non-standard
category ‘Redundancies’ was obtained by dividing 13 by 67 expressions (each of the speech turns presented on average
(that is, 0.19). When transformed in percentage, this value 1.23 non-standard expressions), and were highly elliptic,
represents the percentage of speech turns in which such i.e., controllers omitted more elements (in 69% of the
particular deviation was present. Results are presented as turns) than in other periods (see also Table 3), a
percentages expressing the ratio of speech turns in which a communicative pattern that might be due both to a
specific category occurred. decrease in vigilance and an increase of mental fatigue.
Multivariate analyses of variance were performed on the
Communications produced in the morning shifts were,
data relative to each of our main categories.
in general, the most correct (even if they still presented a
high rate incorrectness): all the linguistic deviations,
3. RESULTS AND DISCUSSION except redundancies, were significantly less frequent than
in the other shifts (Italian was used in 37% of the speech
Our results suggest that the taxonomy we designed was turns, non-standard expressions were present in 48% of
appropriate for capturing a variety of miscommunications them, omissions in 18% and incorrect callsigns in 41% of
and ambiguities in ATCo communications. As to the them; see the statistics mentioned above). Afternoon shifts
organisational factors we considered, the workshift organi- showed an intermediate level of correctness (Italian was
sation and level of workload indeed affected the character- used in 46% of the speech turns, non-standard expressions
istics of communications: deviations from the prescribed in 85% of them, omissions in 29% and incorrect callsigns in
phraseology were produced significantly more frequently 49%).
during the nightshifts (F (5,897) = 13.225, p < 0.001) both Linguistic deviations were significantly more frequent
by RCs and TCs (Italian was used in 53% of the speech during the low workload periods than in the high workload
turns, F(2,897) = 6.923, p < 0.001; non-standard expres- ones (F(5,896) = 12.109, p < 0.000). In particular, the
sions in 99% of the cases, F(2,897) = 19.157, p < 0.0001; difference was significant for the use of Italian (58% with
omissions in 51% of them, F(2,897) = 19.263, p < 0.0001 minimum workload vs. 34% with maximum workload;

Table 3. Percentages of speech turns containing a linguistic deviation in the different conditions

Workshift Workload Position Linguistic deviations (percentage of speech turns containing a deviation)

Use of Italian Non-standard Omissions Incorrect/omitted Redundancies

expressions callsigns

Morning shift Minimum workload Tower 63 54 9 37 19

Radar 16 27 16 43 53
Maximum workload Tower 43 60 28 27 47
Radar 27 52 60 57 55
Afternoon shift Minimum workload Tower 51 74 24 23 35
Radar 59 83 38 83 34
Maximum workload Tower 36 82 19 21 42
Radar 39 100* (1.02) 33 70 57
Night shift Minimum workload Tower 60 73 47 53 67
Radar 96 100* (1.74) 92 94 27
Maximum workload Tower 52 88 35 64 56
Radar 5 62 34 57 47

* In this condition the mean of deviation per speech turn was >1 (see values in parentheses), i.e. each speech turn contained more than one deviation on
average.
The Effect of Workload and Workshift on Air Traffic Control 235

F(1,897) = 46.980, p < 0.0001) and for omissions (b) TCs employed Italian more frequently than RCs
(respectively, 38% vs. 28%; F(1,897) = 5.097, p < 0.05). (50% vs. 35% of speech turns); this could be due to
Non-standard expressions (81% vs. 74%) and incorrect the kind of traffic they had to manage: in the
callsigns (56% vs. 49%) presented a similar pattern aerodrome area more local and tourism traffic
although not statistically significant. transit than in the approach area, and it is not rare
The communications during the morning shift showed a for pilots to be personally known by the controllers.
partly different pattern, since they were less correct in
correspondence with a high workload level, except for the Despite the norms prescribing a strict adherence to
use of Italian. This was probably due to the fact that the standard phraseology, controllers produced communica-
low workload during the morning was higher than the low tions that deviated from that phraseology diffusely and in
workload in the nightshift (respectively, seven vs. one to all conditions, presumably not only as a consequence of
two aircraft); while one or two aircraft represented the organisational and environmental factors (Corradini and
condition of very low workload that impaired performance, Cacciari 2001). They employed an ‘alternative code’,
seven aircraft could be a load that sustained controllers’ which is analysed in detail in the following sections.
arousal and vigilance. In the afternoon shift the workload
level did not systematically affect the linguistic deviations. 3.1. Organisation of the Transition
The work position also influenced the communicative
patterns of the controllers. All kinds of linguistic deviations Non-standard transition, i.e. the exchanges presenting
were produced more frequently by radar controllers than by misunderstandings that subsequently required further ex-
tower ones, a part from the use of Italian: differences planations or corrections, were overall 25 (4% of all the
between the communications of the two operator groups transitions). An example follows where the pilot asks
were statistically significant for omissions (present in 39% of repeatedly for departure instructions probably because he is
the speech turns by RCs and in 27% of them by TCs, not familiar with the airport and with the names employed
F(1,897) = 7.905, p < 0.005) and incorrect callsign by the controller (*** stands for a geographical name and
(employed in 67% of the speech turns by RCs and in 38% ˆ ˆ ˆ for a route identification code. Other examples can be
of them by TCs, F(1,897) = 41.225, p < 0.001). Non- found in Section 3.7, point c.)
standard expressions and redundancies, although not reach- (8)
ing significance, showed the same trend (non-standard
expressions were found in 83% of RCs’ speech turns and in P: And may I have (. . .) departure clearance?
72% of TCs’; redundancies in 45% an 44% of the speech C: Three four zero four, cleared to destination via *** five
turns respectively). Italian, on the contrary, was more often Mike standard departure, Whiskey niner one, initially FL
employed by TCs (in 51% of the speech turns) than by RCs one three zero. Sqwack zero two five two.
(in 40% of the speech turns) (F(1,897) = 8.887, p < 0.003). P: ***, say again, *** . . . five Alfa . . .
Different factors could account for these results: C: *** five Mike standard departure.
P: *** five Mike standard departure, Whiskey niner one,
(a) That RCs produced more phraseology deviations one three zero initially, sqwack zero two five two, XX
than TCs could be due to their higher medium age three four zero four.
and to their longer length of service (see Section C: Three four zero four, all correct.
3). P: Thank you.
(. . .)
(a.1) The age and the longer exposition to shift work P: Tower, XX three four zero four ready for taxi.
could produce a slight decrease in performance C: Stand by.
correctness and in tolerance to shift work (Rein- P: Standing by.
berg et al 1980) and a risk of burn-out, i.e. changes (. . .)
in mood and behaviour that can beome safety C: Three four zero four, taxi holding point runway one two,
relevant (Dell’Erba et al 1994). left, left again, please expedite.
P: Confirm runway one two?
(a.2) It should be noted that the ATC training
C: Confirm Sir, we have the traffic inbound runway one two.
programmes have changed over the years, and in P: So our departure – a standard departure (. . .) which
recent times much more attention has been given departure?
to the correct use of phraseology. C: ˆ ˆ ˆ five Lima standard departure.
(a.3) The differences we found do not anyway seem to P: Five Lima, roger.
depend on vigilance level, as no significant
difference in vigilance was found between the Non-standard transitions were more frequent in tower than
two operative groups of tower vs. radar controllers in the radar approach centre (17 vs. 8) and under the
(Cicogna et al 1998). maximum workload (15 vs. 10), but these differences were
236 P. Corradini and C. Cacciari

not significant; we could not perform other statistical Omissions often became a sort of linguistic routine, as for
analysis given the small amount of cases. example in the case of omission of terms such as ‘QNH’,
‘speed’, ‘turn’, ‘heading’. Given that the numerical range of
3.2. The Use of Italian many flight parameters partially overlaps (as in the ranges
of speed, heading, flight level, QNH), the ellipsis of the
Italian was used in 38% of the messages in the corpus (we term to which the number refers to increases the risk of
counted the number of Italian words uttered over the total misunderstandings.
amount of words excluding the geographical references). The acknowledgement lacked completely only in one
ATCos created a sort of ‘pidgin’ where they mixed controller’s message and in less than 1% of pilots’: a lack of
translations of the English phraseology, shared Italian confirmation might lead to relevant misunderstandings if,
idiomatic expressions, natural language and English words. for instance, a controller assumed that a pilot received a
The use of Italian hinders the advantages of the ‘party line’, message even if he or she did not acknowledge it.
making the monitoring of communications by non-Italian
flight staff impossible. 3.5. Omitted or Incorrect Callsign

3.3. Collective and Individual Non-standard Although the importance of the callsign, i.e. the aircraft
Routines identification string, is well acknowledged (e.g. Billings and
Cheaney 1981; NASA-ASRS 1994), the callsign was
Controllers employed a consistent amount of collective omitted, shortened or replaced by a nickname in 43% of
non-standard routines (57%) that, even if conventionally controllers’ messages (and in 44% of pilots’ messages).
rooted in the aviation operators’ experience, did not While nicknames were clearly shared by operators,
correspond to the criteria leading standard phraseology. partial ellipses of callsign were very variable, also because of
Individual non-standard routines were less frequent but still the different ways in which a number can be pronounced
diffused (17%). In the following examples the controller (see ‘routines’), e.g.:
uttered messages definitely ambiguous: in (96) the message
(11)
contained two alternatives the pilot had to choose
between, instead of a precise instruction (‘X’ replaces the Callsign Possible utterances
name of an airline company and ‘***’ stands for a
Alitalia 1882, climb . . . ‘Alitalia, climb . . .’;
geographical name):
‘One eight eight two, climb . . .’;
(9) ‘One double eight two, climb . . .’;
‘Alitalia eighteen eighty-two,
C: X224, which level do you request to ***? climb . . .’

In (10), the controller employed a ‘conditional sentence’, Linguistic behaviour of this sort could be problematic,
i.e. a sentence containing an instruction effective only if a especially when aircraft have similar callsign and when
previous condition, not depending on the pilot addressed there are more aircraft of the same company in a single
but on another aircraft, was met: airspace: the omission of the number in the callsign could
(10) prevent any discrimination.
In our corpus we found four mistakes relative to the
C: X1310, cleared to line up runway one two following air callsign numbers; they were probably due to interference
carrier X992. with other numbers in the message and were always
detected and corrected.
Conditional sentences could be very risky (ASRS 1999),
since their comprehension and interpretation depend on 3.6. Redundancies
the pilot’s ability and awareness; these sentences could
conceal ambiguities very hard to detect and solve. Despite the fact that one of the goals of standard
phraseology is the control over message length, we observed
3.4. Omissions a relevant presence of redundancies, i.e. of useless locutions
and repetitions (18%), politeness formulae (26%) and
The frequency of the omissions was inversely proportional negotiation (2%).
to the relevance of the elements omitted: clearances were Redundant messages might have negative consequences
never omitted; information and parameters were absent since they unnecessarily keep the radio frequency active,
only in 3% of messages and the information omitted was representing also a cognitive load for the receiver. Useless
not crucial. The ellipsis of words or numbers was much locutions represent ‘noise’ with respect to the information to
more common (26% for controllers and 29% for pilots). be transferred and increase the risk of misinterpretations and
The Effect of Workload and Workshift on Air Traffic Control 237

misunderstandings. Negotiations could also have a negative (d) semantic ambiguity: extremely few cases were found
fall-out on pilot–controller interpersonal interaction. whose treatment exceeds the aim of this paper.
An example of a redundant expression is the following
where the message ‘This is actually a vectoring for a ILS
final one two and you may descend to two thousand feet,
not three.’ could have been better expressed by ‘Negative,
descend two thousand feet’. 4. CONCLUSIONS
(12)
We proposed a taxonomy of communicative problems that
P: *** buongiorno, XX zero five zero, flight level zero niner might produce ambiguities and misunderstandings in ATC
zero. communication. The taxonomy was used to assess whether
C: XX zero five zero buongiorno, ident six three four zero. workshifts and workload had an effect on the commu-
P: Six three four zero ident. nicative side of the cognitive performance of ATCos.
C: Radar contact, [position] twenty-three miles west, left Although a consistent literature on the effects of workshift
[heading] zero three zero, descend two thousand feet, QNH
and workload on humans’ cognitive performances is
one zero zero four, transition level seven five, vectoring
ILS final [runway] one two. Shortly number one.
available, as well as an even more consistent body of
P: OK, left heading zero three zero, descending] three studies on ATC communication, very few studies con-
thousand feet, on QNH one zero . . . eh . . . zero four, sidered the effects of these organisational factors on such a
expecting runway one two. crucial aspect of ATC (none of them on the Italian
C: This is actually a vectoring for an ILS final one two and context).
you may descend to two thousand feet, not three. Our results suggest that indeed the communicative
P: Down two thousand feet. performance of ATCos was affected by these chronopsy-
chological and organisational factors. In fact, an increase in
communicative incorrectness was found in the messages
3.7. Misunderstandings and Confusions produced during nightshifts. Particularly such increase
characterised the period of minimum workload, where a
We found few cases of misunderstandings followed by
decrease in the vigilance level was presumably associated
requests for clarification, i.e. of non-routine transitions (less
with an increase in mental fatigue. That no other peak of
than 4%). Although the misunderstandings were always
linguistic incorrectness was observed supports the hypoth-
detected and corrected, they are still relevant. We detected
esis that vigilance plays an important role in determining
four types of ambiguity that gave raise to such misunder-
operators’ performance efficiency, as also suggested by the
standings and to requests for clarification:
fact that the most correct communications occurred during
(a) instructions that were not clearly understood because a the morning shift. The effect of workload was not so clear
single term or value was not heard; cut; nonetheless the accuracy and correctness of the
(b) words or digits phonetically similar that were confused, communications decreased when the workload level was
as in (133): low, as suggested in the literature.
(13) Controllers and pilots created an alternative linguistic code
instead of using the correct standard phraseology – a code
C: X six two six two nine, now clear of traffic. Report before inspired by standard phraseology but also strongly deviating
leaving this frequency. from it, particularly with respect to: the use of Italian
P: Roger, X six six six two nine. instead of English; the use of collective and individual non-
standard routines; the omission of words and numbers; the
(c) interferences or transpositions that took place between omission and incorrect formulation of callsigns; the
callsigns or numbers presented in different parts of the presence of redundancies as politeness formula, repetitions,
messages and referred to different parameters, for negotiations, expressions, and intrusions of natural lan-
instance: guage. Even if this jargon was generally well understood, it
(14) could potentially produce ambiguities and misunderstand-
ings that might be fatal, especially in non-standard or
C X zero six zero turn left heading three two zero, descend to emergency situations where cognitive, attentional and time
five thousand feet, QNH one o o two. resources are modified and restricted (Cushing 1994;
P Left three fifty, down to five thousand, one o o two, X zero Bellorini et al 1998).
six zero. Unlike natural conversations, even a single case of
C Heading three two zero.
undetected miscommunication or misunderstanding can be
P Correction, three two zero.
fatal in this operational environment.
238 P. Corradini and C. Cacciari

References Kanki BG, Greaud VA, Irwin CM (1991). Communication variations and
aircrew performance. International Journal of Aviation Psychology
1(2):149–162.
Andlauer F, Andlauer P, Carpentier J, Cazamian P (1977). Ergonomie du
travail de nuit et des horaires alternants. CUJAS, Paris. Morrow D, Lee A, Rodlov M (1991). Collaboration in pilot–controller
communication. In Proceedings of the sixth international symposium on
Andriessen EJH (1995). The effectiveness of interaction technology: a four aviation psychology, Ohio State University, Columbus.
level evaluation. AOP paper 95.001. Delft University of Technology,
Delft. Morrow D, Lee A, Rodvold M (1993). Analysis of problems in routine
controller–pilot communication. International Journal of Aviation
ASRS (1999). Conditional clearance confusion. Callback, 238. Psychology 3(4):285–302.
Bellorini A, Vanderhaegen F (1995). Communication and cooperation NASA-ASRS (1994). Synopsis of pilot/controller communications
analysis in air traffic control. In Proceedings of the eighth international reports. ASRS, Mountain View, CA.
symposium on aviation psychology, Ohio State University, Columbus.
Orasanu J, Fischer U (1991). Information transfer and shared mental
Bellorini A, Cacciabue PC, Nanetti N (1998). Ambiguity and problems in models for decision making. In Proceedings of the sixth international
communication in air traffic control. Presented at the ninth European symposium on aviation psychology, Ohio State University, Columbus.
conference on cognitive ergonomy, Limerick, Ireland.
Orasanu J, Fischer U (1992). Team cognition in the cockpit: linguistic
Beugnet-Lambert C, Leconte P, Lancry A (1988). Chronopsychologie: control of shared problem solving. In Proceedings of the 14th annual
rythmes et activités humaines. Presses Universitaires de Lille, Lille. meeting of the Cognitive Science Society. Erlbaum, Hillsdale, NJ.
Billings CE, Cheaney ES (1981). Information transfer problems in the Orasanu J, Salas E (1993). Team decision making in complex
aviation system. NASA TP 1875. NASA–Ames Research Center, environment. In Klein GA, Orasanu J, Calderwood R, Zsambok CE
Moffet Field, CA. (eds). Decision making in action: models and methods. Ablex,
Cacciari C, Giusberti F, Nicoletti R, Scozzari V, Santolini S, Sarchielli M Norwood, NJ.
(1997). La rappresentazione mentale del compito nel controllo del Reinberg A, Andlauer P, Guillet P, Nicolai A, Vieux N, Laporte A
traffico aereo. In La Rosa M (ed). Governo delle tecnologie, efficienza e (1980). Oral temperature, circadian rhythm amplitude, ageing and
creatività. Il contributo dell’ergonomia, Monduzzi, Bologna. tolerance to shiftwork. Ergonomics 23:55–64.
Cicogna P, Nicoletti R, Alzani A, Iani C, Natale V (1998). Livelli di Reinberg A, Vieux N, Andlauer P (eds) (1981). Night and shiftwork:
vigilanza soggettiva e tipologia circadiana nel controllo del traffico biological and social aspects. Pergamon Press, Oxford.
aereo. Presented at the sixth Italian Society of Chronobiology
Seamster T, Cannon JR, Pierce RM, Redding RE (1992). Analysis of en
conference, Chianciano.
route air traffic controller team communication and controller resource
Corradini P, Cacciari C (2000). ‘Speedbird 996, cleared to twenty’. management (CRM). In Proceedings of the Human Factors Society
Linguistic routines and ambiguity in air traffic control communications. 36th annual meeting. Santa Monica, USA.
Presented at the seventh international pragmatics conference, Budapest.
Stein ES (1998). Human operator workload. In Smolensky MW, Stein ES
Corradini P, Cacciari C (2001). Shiftwork and workload: effects on air (eds). Human factors in air traffic control. Academic Press, San Diego,
traffic controllers communications. In Proceedings of the IEE conference CA.
people in control 2001, Manchester.
Costa G, Schallenberg G, Ferracin A, Gaffuri E (1995). Psychophysical
conditions of air traffic controllers evaluated by the standard shiftwork Correspondence and offprint requests to: Cristina Cacciari, Department of
index. Work and Stress 9(2/3):281–288. Biomedical Sciences, Via Campi 287, 41100 Modena, Italy. Email:
Cushing S (1994). Fatal words: communication clashes and aircraft [email protected]
crashes. University of Chicago Press, Chicago.
Dell’Erba G, Venturi P, Rizzo F, Porcu S, Pancheri P (1994). Burnout and
health status in Italian air traffic controllers. Aviation, Space and
Environmental Medicine 65(4):315-3-22.
APPENDIX 1: THE TAXONOMY USED
Depolo M, Cavallero C, Cicogna PC, Contento S, Zanobini F (1997). Il
FOR THE COMMUNICATION ANALYSIS
controllo del traffico aereo: un confronto fra la rappresentazione mentale
del compito nei ontrollori e nei piloti. In La Rosa M (ed). Governo delle 1. Organisation of the transition
tecnologie, efficienza e creatività. Il contributo dell’ergonomia, 1.1. Standard
Monduzzi, Bologna.
EATCHIP (1997). Model of the cognitive aspects of air traffic control
1.2. Non-standard (presenting misunderstanding that
(HUM.ET1.STO1.1000-Rep-02). EUROCONTROL, Brussels. altered the normal communicative formats)
Falzon P (1982). Les communications verbales in situation de travail : 2. Problems with the medium
analyse des restrictions du langage naturel. Rapport technique No. 19. 3. Misuse of standard phraseology
Groupe de Psychologie ergonomique, INRIA.
Foushee CH (1984). Dyads and triads at 35,000 feet: factors affecting
3.1. Cross-linguistic factors
group process and aircrew performance. American Psychologist 3.1.1. Use of Italian
39(8):885–893. 3.1.2. Shift from English to Italian and vice versa
Ginnett RC (1993). Crews as groups: their formation and their leadership. 3.2. Use of non-standard communicative routines
In Wiener EL, Kanki BG, Helmreich RL (eds). Cockpit Resource
Management, Academic Press, San Diego, CA. 3.2.1. Non-standard collective routines (non-tech-
Grayson RL, Billings CE. (1981). Information transfer between air traffic nical expressions shared by the operators)
control and aircraft: communication problems in flight operations. In 3.2.2. Non-standard individual routines (expres-
Billings CE, Cheaney ES (eds). Information transfer problems in the sions that are neither technical nor shared;
aviation system (NASA TP 1875). NASA–Ames Research Center,
Moffet Field, CA. current language)
Kanki BG, Foushee HC (1989). Communication as group process mediator 3.3. Omissions
of aircrew performance. Aviation, Space and Environmental Medicine 3.3.1. Information or clearances (or part of them)
60(5):402–410.
3.3.1.a. Clearances
Kanki BG, Palmer MT (1993). Communication and crew resource
management. In Wiener EL, Kanki BG, Helmreich RL (eds). Cockpit 3.3.1.b. Information or parameters
Resource Management, Academic Press, San Diego, CA. 3.3.1.c. Words or numbers
The Effect of Workload and Workshift on Air Traffic Control 239

3.3.2. Acknowledgements and readbacks 3.4. Redundancies

3.3.2.a. Acknowledgement totally absent 3.4.1. Repetitions
(no response) 3.4.2. Redundant expressions
3.3.2.b. Acknowledgement without read- 3.4.3. Politeness formula
back 3.4.4. Negotiation expressions
3.3.2.c. Acknowledgement with partial
readback 4. Confusion of homophones/interferences
4.1. Numbers
3.3.3. Callsign (name of control station absent or 4.2. Callsigns
partial, both for sender and receiver) 4.3. Words
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.