INCENTIVE THEORY AND

IN REWARD1
Frank A . Logan
UNIVERSITY OF NEW MEXICO
ALBUQUERQUE, NEW BfEXICO

I. Introduction .......................................... 1
A. Incentive Effects in Partial Reinforcement. ............ 2
B. Re rg-sgas the Mechanism of Incentive Motivation. ..... 5
C. Rationale.. ....................................... 8
II. Method .............................................. 10
A. Subjects .......................................... 10
B. Apparatus ........................................ 11
C. Procedure......................................... 11
111. Experimental Designs and Results. ...................... 12
A. Experiment 1. Joint Extinction Following Partial and
Continuous Reinforcement .......................... 12
B. Experiment 2. Reacquisition Following Original Training
with Partial and Continuous Reinforcement. . . . . . . . . . . . 14
C. Experiments 3 and 4. Reversal of Choice Following Partial
and Continuous Reinforcement. ...................... 15
D. The Overtraining Extinction Effect.. ................. 19
E. Experiment 5. Effect of Amount of Reward on Resistance
to Extinction of Incentive Motivation. ................ 19
F. Experiment 6. Incentive Contrast.. ................... 20
G. Experiment 7. Rate of Increase in Incentive Motivation. . 24
IV. Discussion............................................ 25
V. Summary ............................................ 28
References............................................ 28

I. Introduction
Of the basic phenomena encountered in the experimental analysis of
the learning process, perhaps none has proved to be so intractable to an
even temporarily satisfying theoretical description as that of experi-
mental extinction. The closest was probably Pavlov’s (1927) original
postulation of an internal inhibitory process which, as every introductory
psychology student is supposed to know, was suggested by such other
This research waa supported in part by grants from the National Science
Foundation. Louis Gonzelez supervised the laboratory during the time these data
were collected, and ran many of the subjects. I n addition, Christopher Bennett,
Alfred Coscina, Douglas Gibson, Albert Gonzales, and Eli Padilla assisted in
collecting the data.
1
2 Frank A. Logan

phenomena as spontaneous recovery and disinhibition. But when Hull

(1943) attempted to formalize this approach within a more general and
systematic behavior theory, difficulties became apparent. These are
generally familiar, having been recited extensively in the literature (e.g.,
Gleitman, Nachmias, & Neisser, 1954; Koch, 1954), and while defenses
and adjustments could be made, the superstructure required to support
Hull’s original account of experimental extinction would certainly lack
parsimony.
Furthermore, there has been a trend among many contemporary
learning theorists to reanalyze the distinction between learning and
performance, particularly with reference to the concept of reinforcement.
The emerging view is that learning, per se, results from a simple contiguity
of the events being associated while reinforcement affects performance
in a motivational manner. The contiguity principle was presaged by
Guthrie (1934),but his system did not include the motivational flavor of
more recent approaches ; Tolman’s (1932) theories were highly motiva-
tional, but were also heavily burdened with cognitive-sounding con-
structs. Intellectual descendents of both of these approaches, together
with representatives from the several branches of the Hullian family
tree, are more or less independently converging on a theoretical
rapprochement.
One direction of convergence is apparent in some of my own work
(e.g., Logan, 1960; Logan & Wagner, 1965)and is clearly evident in the
writings of Cofer and Appley (1964),. Miller (1963), Mowrer (1960),
Seward (1956a, 1956b), F. D. Sheffield (1966),and Spence (1956, 1960).
To be sure, there are important differences among these various
approaches and none, in my opinion, has yet hit upon the key which
will start the next round of fruitful theorizing on a grand scale. But the
‘hew look of learning” must imply a new look at extinction; this paper
attempts to outline several features of the problem and presents some
relevant new data.
A. INCENTIVE EFFECTS IN PARTW REINFORCEMENT
Let me begin by dogmatically asserting the language to be followed
here. This language was visible in Hull’s (1952)last version of his theory
and has been elaborated more extensively by Spence (e.g., 1956). The
critical words are “habit,” “incentive,” and “drive,” which are here
symbolized respectively sHr, sINr, and D. These combine multiplica-
tively to determine excitatory potential which, in turn, is reflected in
performance as measured in various ways. Habit refers to a more or less
permanent association between a stimulus and a response acquired
gradually on the basis of contiguous occurrence of the events in question.
Incentive is a learned motivational factor scalable both upward and
 Incentive Theory and Changes in Reward 3

downward to reflect the anticipated reward level based on contiguous

experience of reward following a stimulus-response event. Drive is a
temporary state property of the organism based in primary form upon
conditions of deprivation or noxious stimulation and, in secondary
form, upon stimuli previously associated with aversive states of affairs.
Considering instrumental reward conditioning within this language
system, original acquisition results in at least two learning processes :
habit and incentive. Under some conditions or reinforcement, drive may
also be changed. Correspondingly, extinction could result from loss of
habit, a loss of incentive, or a change in drive; probably all will be
implicated in some situations. The particular purpose of this paper is to
reflect upon the extent to which extinction involves changes in incentive
motivation and the basic contention is that other accounts of extinction
phenomena need to take incentive factors explicitly into consideration.
This last contention, together with the rationale of the present
experimental approach, can be developed in the context of the familiar
partial reinforcement effect on extinction (PREE): e.g., a rat that is
given food reward on only part of the trials that he runs down an alley
persists longer in running when reward is no longer given than a rat
initially rewarded every trial. Let us first consider the early Hull-
Sheffield(V.F. Sheffield, 1949) account of this phenomenon. Nonreward is
assumed to produce a persisting aftereffect which, at least with reason-
ably short intertrial intervals, is present on the succeeding trial; par-
tially rewarded rats encounter these nonreward stimuli during acquisi-
tion and learn to run in the presence of their aftereffects because doing
so is sometimes rewarded. Continuously reinforced rats, in contrast,
encounter these stimuli for the first time during extinction and suffer
thereby a generalization decrement in habit. This approach emphasizes
the rapidity of extinction following continuous reinforcement rather
than the slowness of extinction following partial reinforcement. Par-
ticularly in view of the recent elaborations and evidence provided by
Capaldi (1966), the aftereffects analysis is not only a logically necessary
part of the PREE story but an empirically demonstrable one.
However, the aftereffects analysis does not make any explicit state-
ments about incentive motivation. Presumably, insofar as incentive is
elicited by the instrumental stimuli (including the aftereffects of rein-
forcement and nonreinforcement), a change in them would also produce
a generalization decrement in incentive motivation. That is to say,
continuously reinforced rats should have little incentive to run in the
presence of the aftereffects of nonreinforcement and the performance
decrement should also reflect this motivational change. But the rate at
which incentive is decreased by further nonreinforcements is not
specified nor even considered by an aftereffects analysis.
4 Frank A. Logan

A more recent account of the PREE is provided by frustration theory

(e.g., Amsel, 1962; Spence, 1960). Anecdotally, everyone is familiar with
the fact that failure to receive an anticipated reward results in “frustra-
tion.” The essence of the more formalized version is that frustration
contributes to the state of drive motivation and that, like all drives,
frustration drive has an associated drive stimulus with which overt
responses may be associated. Furthermore, frustration is assumed to be
a learnable drive so that it may occur in anticipatory fashion en route
to a goal where primary frustration has been experienced. These basic
assumptions account for the PREE as follows : continuously rewarded
rats will be frustrated at the outset of extinction and will, on subsequent
trials, begin to anticipate frustration. Although this added frustration
drive increases total motivation, and might thereby increase per-
formance, the initial unconditioned response to the frustration drive
stimuli is assumed to be incompatible with the typical instrumental
response. Extinction following continuous reinforcement reflects the
occurrence of such competing responses in usual performance measures
such as speed of locomotion. Partially reinforced rats, on the other hand,
have encountered nonreward during acquisition, have thus experienced
anticipatory frustration, and have been rewarded for continuing to
respond in the presence of these conditioned frustration drive stimuli ;
hence, they continue to perform even though also frustrated during
extinction.
Two factors are involved in the frustration theory account of the
PREE: extinction produces an increase in drive (at least for the con-
tinuously reinforced rats) but the resulting change in the drive stimulus
results in a loss of hubit and the introduction of competing habits.
Interestingly enough, frustration is assumed to occur only when and
only so long as incentive motivation as an expectation of reward exists,
yet explication of how partial reinforcement affects the level of incentive
and the persistence of incentive during extinction have not yet been
explicitly formulated. Inclusion of these rules is necessary for a complete
frustration theory.
One might attempt to derive the properties of incentive motivation
by assuming i t to be mediated by a classically conditioned response
mechanism such as the fractional anticipatory goal response (rB-sg).One
specific version of this assumption is that the instrumental stimuli and
the feedback stimuli from the instrumental response serve to elicit rg
and that the dynamism properties of so feed into the total motivational
complex. On such an assumption, the laws of classical conditioning
should provide the appropriate rules since rg is assumed to be conditioned
according to Pavlovian principles. Some comments on this approach are
given in the next section.
 Incentive Theory and Changes in Reward 5

B. RE rg-sgAS THE MECHANISMO F INCENTIVE MOTIVATION

I n spite of the apparent appeal in employingthe fractional anticipatory
goal response as the mediating mechanism for incentive, there are
several reasons for questioning the adequacy and generality of this
approach. It may be of heuristic value to outline these reservations, and
they also provide conceptual justification for the empirical analysis to
be reported.
1. The first problem is strictly logical within the theory. Drive
motivation (D) should affect the performance of any response, including
rg. Hence, D would be embedded within incentive motivation if the rg-sg
analysis is followed. Accordingly, D would enter twice in determining
instrumental performance (once in energizing the habit for r6, and again
in energizing the habit for the instrumental response itself). Of course,
the theory could be changed in these respects so as to restrict the ener-
gizing role of drive to anticipatory goal responses, but an analysis would
have to indicate how drive effects can be handled entirely through
incentive motivation. (For an attempt at such an analysis, see Black,
1965.) For example, adding an irrelevant drive can increase performance
(Kendler, 1945), but it is not immediately apparent how this could
happen if re is selectively energized by drives (Spence, Bergmann, &
Lippitt, 1950).
2. A second problem concerns the fact' that r6-s6is also used, appro-
priately I believe, as the mediating mechanism for secondary reinforce-
ment. A dual role for r6-s, in both incentive motivation and secondary
reinforcement is not necessarily bad in its own right, but there appear
to be asymmetries in secondary reinforcement and secondary motivation
that would preclude this approach. These asymmetries warrant a brief
aside.
Although less powerful and persistent than one might hope, secondary
reinforcement is a readily demonstrable phenomenon when an originally
neutral stimulus is paired with a positive reinforcement such as food
(e.g., Zimmerman, 1957). In contrast, the weight of evidence indicates
that a neutral stimulus paired with a negative reinforcement such as the
termination of electrical shock, does not acquire learned reinforcing
properties (e.g., Nefzger, 1957). There is an inverse asymmetry in the
case of learned drive : stimuli paired with the onset of an aversive event
acquire secondary motivating properties (e.g., Millcr, 1948) while those
associated with an increase in deprivation drives apparently do not
(e.g., Novin & Miller, 1962). In short, as viewed in the format of a 2 x 2
matrix, hunger-food is effective for secondary reinforcement but not
secondary motivation ; pain-fear is effective for secondary motivation
but not secondary reinforcement.
6 Frank A. Logan

These asymmetries can be rationalized by attending to the response

components of the events in question and assuming that all learning
consists of S-R associations. The deprivation drives are not normally
response-produced (e.g., one simply becomes hungry over time without
making any special drive-inducing response) and hence there is no
learnable response component to the deprivation drives ; primary
reinforcement based on these deprivation drives, however, is response-
produced (e.g., the consummatory responses of eating, drinking, copula-
ting) and hence there is a learnable response mechanism for secondary
reinforcement. On the other hand, the pain-fear drive does contain
learnable response components (arousal in the sympathetic branch of
the autonomic nervous system) which can mediate secondary motiva-
tion, but primary reinforcement based on this drive is not response-
produced (i.e., the aversive state simply terminates) so that no response
mechanism for secondary reinforcement exists.
This analysis is not only consistent with the best available evidence in
each case, but it also offers promise of rationalizing other perplexing
facts. People will fast before Thanksgiving dinner, or eat pretzels while
drinking beer, in each case “suffering” some increase in drive so as better
to enjoy the consummatory reward. But a normal person will not, as
the saying goes, hit his head against a wall because it feels so good to
stop. There is no secondary motivation while there is greater secondary
reinforcement associated with at least moderate increases in deprivation
drives; the converse is true of stimulation drives.
This analysis shows that rg-sg is a useful mechanism for secondary
reinforcement since it can account for the difference in the effects of
positive and negative reinforcement as primary bases for learned
reinforcement in a manner that is consistent with the difference between
appetitional and aversive drives as primary bases of learned motivation.
From the point of view of incentive motivation, however, positive and
negative reinforcement do not appear to show any significant differences.
Both are effective in conditioning and maintaining instrumental
responses (see, e.g., Bower, 1960) and quantitative details concerning
choice behavior as a measure of relative incentive effects appear com-
parable (e.g., Kaplan, Kaplan, & Walker, 1960). Accordingly, we may
reasonably conclude that incentive motivation is controlled by the same
type of process in positive and negative reinforcement. Were this process
mediated via rg-sgin the case of positive reinforcement, an analogous
response would have to be postulated for negative reinforcement (see,
e.g., L‘relie”’ in Mowrer, 1960). The undesirable by-product of such an
approach would be a learnable response mechanism for secondary
reinforcement based on primary negative reinforcement, which would be
contrary to the best available evidence on this issue.
3. An assortment of difficulties for rg-sgas a mediating mechanism
 Incentive Theory and Changes in Reward 7

for incentive arise if one pays strict attention to the data in the relevant
situations. As a specific illustration, the interstimulus-interval function
in classical conditioning shows an optimum a t somewhere between a
half and one second; the delay of reinforcement gradient does not show
such nonmonotonicity. As another example, Egger and Miller (1963)
have recently shown that the later stimuli in a sequential chain of con-
ditioned stimuli preceding an unconditioned stimulus do not come to
control conditioned responses. As secondary reinforcers, only the early
stimuli are effective. This should mean that rg would not be evoked by
the stimuli arising late in the instrumental behqvior chain, although the
common assumption in this regard is that incentive motivation increases
progressively as the goal is approached. The detailed laws of classical
conditioning have not yet been systematically applied to the incentive
construct within a theory and hence the extent to which inconsistencies
would be detected is not fully known. However, special problems would
almost certainly arise from such realizations as that conditioned
responses occur with some probability while incentive motivation must
be present on every trial.
4. An empirical source of disillusionment concerns the attempts to
identify salivation as a component of rg and to obtain supporting
evidence by recording and/or manipulating salivation during instru-
mental conditioning. By and large, the results of these efforts have not
been very encouraging (see, e.g., Lewis & Kent, 1961) although a few
suggestive results have been reported (e.g., Shapiro, 1962). It may be
granted that salivation has been used simply to illustrate the logic of
an rg-sgmechanism ; rg as an intervening variable need not be physio-
logically localized and hence the failures with salivation are indeed not
crucial. But to many observers, the salivation illustration is simply too
good: the necessary features (e.g., a learnable component of the goal
response that is not incompatible with the instrumental response) are
present and it ought to work better.
5. Finally, an approach in which incentive motivation is mediated by
a response mechanism is most appropriate for a theory utilizing a
cybernetic-feedback type of analysis, especially if incentive is assumed
to contribute to the general motivational level (but see Trapold, 1962,
for contrary evidence). Indeed, this is the context in which it is typically
employed (e.g., Miller, 1963; Mowrer, 1960; F. D. Sheffield, 1966;
Spence, 1960). I n these analyses, the organism is assumed somehow to
be behaving; feedback from his behavior is available and is, through
prior conditioning, associated with the incentive-mediating response
mechanism which, in turn, fosters either continuation of the ongoing
behavior or change, presumably toward a more optimal level. Miller and
Sheffield have, somewhat differently, included response habits on which
to base the choice of responses. Although this type of approach has a
8 Frsnk A. Logan

great deal of appeal and most probably indicates the direction of future
theorizing, it does not readily permit incentive motivation to act
selectively at the instant of choice. This difficulty is most obvious at the
moment of initiation of a response, but actually recurs continuously
during a behavior chain.
To emphasize the difficulty in the context of response initiation, one
might postulate that the organism covertly scans the domain of possible
responses, receives internal and external feedback from each of these,
and selects the one with which the most vigorous rg is associated. I n its
simplest and most visible form, this type of behavior is seen in the VTE
behavior of rats at a choice point in a maze (Spence, 1960; Tolman,
1932). Even in that context, however, a complete theory must postulate
additional storage and comparative mechanisms that enable the
organism to select the optimal response. For example, a rat rewarded
differentially in both arms of a T-maze might happen to orient toward
the arm containing the smaller reward. The feedback cues available in
that orientation elicit some rg, but the rat must be assumed to have the
capacity to withhold continuation of that response ;he must be assumed
to store that level of rg while orienting toward the other arm where the
there-appropriate rg is in turn elicited ; following these orientations, the
rg)s must be compared so that their relative strengths can determine
choice. These molecular details have not been explicitly incorporated
into a behavior theory of even this simple situation, much less the more
complex, freer situations where, for example, the rat is choosing a speed
of locomotion. It seems improbable that any organism has the time or
resources to make momentary decisions on the basis of implicit monitor-
ing of all possible courses of action.
It is, of course, possible that a sufficiently molecular analysis will
ultimately show that this approach is realistically tractable. At the
present time, however, it appears preferable to conceptualize incentive
motivation as specific to different S-R events and immediately given as
a basis for choice before the choice is made. Feedback from prior behavior
is certainly a component of the total stimulus complex in which decisions
occur but incentive motivational differences among the responses to be
made must be available in advance. That is to say, the rg associated with
the feedback from the response in the next instant is important as a
secondary reinforcer and to help guide the ensuing behavior, but this
feedback cannot possibly be present at the actual moment of choice
before the response is made.
C. RATIONALE
On the basis of the above arguments, the view adopted here is that
incentive motivation is a fundamental process like habit (i.e., it is not
 Incentive Theory and Changes in Reward 9

mediated) that changes according to mathematical rules as a result of

experience. Each quantitative value of reward is presumably associated
with an incentive value toward which incentive motivation approaches
as a limit. That is to say, incentive motivation has some value before any
trial; if the reward encountered on that trial has an associated incentive
value differing from the existing level, then some change occurs toward
the level appropriate to the reward received. The simplest such model
would assume that the change is a constant fraction of the discrepancy,
and that-the same rule applies whether the change is an increase or a
decrease in incentive motivation.
Existing data suggest that this simple model is not correct, a t least
with respect to the incremental and decremental rates. Were these rates
identical, then the net incentive value of varied reward would always be
equal to the mean of the incentive values of the rewards involved; this
implication has not been confirmed (e.g., Logan, 1960, 1965). The best
interpretation of these data at the moment is that the incremental rate is
greater than the decremental rate, at least under varied reward. That is
to say, if incentive increases toward a higher level on each trial with a
large reward and decreases toward a lower level on each trial with a
small reward, the running average level will exceed the mean of the
separate levels only if the former rate is greater than the latter. In short,
incentive motivation appears to be gained faster than it is lost.
It is still possible, however, that these rates of change in incentive
motivation are state characteristics of the organism and are not subject
to modification by prior experimental conditions. It is this question to
which the present research is addressed. Specifically, is the rate at which
sINr changes modifiable?
The approach begins with a definitional assumption : other things
(such as exposure) being equal, choice between two alternative courses
of action reflects their relative incentive value. I n practice, holding
“other things equal” is difficult: the organism may have a position,
brightness, or similar preference over and above any incentive effects.
In principle, however, choice depends on relative momentary excitatory
potential as mainly determined by incenkive differences based on prior
reward experiences.
Suppose, then, that a rat is given a choice between two alternatives,
one continuously and the other partially rewarded. The rat should learn
to select the continuously rewarded alternative (see e.g., Davenport,
1963). This direct comparison establishes, by definition, the relative
incentive value of the alternatives : sINr is greater following 100 o/o than
50 yoreward.
Suppose further that after the rat has faced these alternatives and has
demonstrated his preference, he is then confronted with extinction
10 Frank A. Logan

conditions in both alternatives. Should incentive motivation decrease at

the same relative rate in both alternatives in spite of their different prior
reinforcement history, preference should decrease monotonically toward
indifference. However, should incentive motivation decrease more
rapidly after continuous reinforcement, preference should temporarily
reverse to the initially 50% alternative. To repeat: after rats have
learned to select a 100% in preference to a 50% alternative, will there
be a (temporary) reversal of preference when both are continuously
nonreinforced during joint extinction? If so, incentive motivation must
persist longer after partial than after continuous reward and thus the
rate of change in sINr must not be invariant.
The approach to be described followed this general rationale. Rate of
change in sINr was evaluated by the rate of change in preference
following a change in the reward conditions. In order to make the con-
clusions as general as possible, several conditions of changed reward
were studied. In each case, however, the same logic of monitoring choice
behavior provided the rationale for inferring incentive motivational
effects.

II. Method
A. SUBJECTS
The Ss were 232 hooded rats predominantly bred in the colony main-
tained by the Department of Psychology of the University of New
Mexico. Additional rats from one replication of several of the studies
were discarded when it subsequently became apparent that the strain
had become so highly inbred at that time that the data were unusually
insensitive to reward effects. Other rats were excluded because they
showed such strong position/brightness preferences that no variability
in their early choice behavior occurred; still others were deleted at
random t o equate N’s for analysis and to ensure appropriate counter-
balancing. The inclusion of these discarded data would, of course, add
noise to the graphic presentation of the results but would not, in fact,
alter any of the conclusions.
Age, sex, and source of the rats were counterbalanced across the
conditions within any study. These are not reported because sub-
analyses failed to indicate any significant interactions with these factors.
During the experimental treatments all Ss were maintained in individual
cages with water freely available. They were maintained on approxi-
mately 12 gm per day of laboratory chow given immediately after each
day’s session.
 Incentive Theory and Changes in Reward 11

B. APPARATUS
Two apparatuses were employed, half the Ss in each study normally
running in each apparatus. These consisted of pairs of parallel straight
alleys, differing principally in that one pair was 4 feet and the other
8 feet long. All doors in the short apparatus were fully automated; the
goal box door in the long apparatus was manually operated. I n each
apparatus, one alley was black and the other white, confounded with
position. Speeds were recorded from reciprocal-reading clocks that
started with the breaking of photobeams located approximately 3 inches
outside the start door and stopped with the breaking of photobeams
located inside the food cups ; these were individually converted into
feet per second for treatment.
Rats were confined in the goal section of the alley after completing the
response. Food reward consisted of 45-mg Noyes pellets. These were
delivered immediately upon breaking of the food-cup photobeam 'by
Davis feeders pulsed at a rate of about six per second. Rats were removed
after consuming the reward or after about 10 seconds on nonreward
trials.
C. PROCEDURE
Ss were preadapted to the appropriate apparatus on an adjusting
shaping schedule. First, groups of four rats were permitted freely to
explore the maze for approximately 30 minutes with all doors open and
with food pellets available in the food cups. This exploration was
followed by individual magazine training in the goal sections of each
alley with CRF programmed on the food-cup photobeam. By the time
training proper began, each S was eating readily from each food cup and
had adapted to the sound of the feeding mechanism.
Rats were run in squads of four on a rotating basis, producing a some-
what variable intertrial interval averaging after the initial trials about
3-4 minutes. Six trials per day were run, the first four trials being forced
according to one of the following patterns rotated over blocks of 4 days :
RRLL, RLLR, LRRL, LLRR. The fifth trial was then a free-choice
trial providing the basis for observation of preference, followed by a
final forced trial opposite in direction to that chosen on the fifth trial.
The assigned reward conditions simply prevailed from the beginning
of training proper. When partial reinforcement was scheduled in any
alley, a preassigned sequence was followed that distributed six reinforce-
ments over the 12 trials in that alley during each block of 4 days. The
sequence was designed to ensure that each ordinal trial position received
equal frequency of reinforcement. Specifically, for example, half of the
choice trials would be programmed for reinforcement if that alternative
were chosen.
 12 Frank A. Logan

The competing reward alternatives were counterbalanced across alleys

* in each study. There was a slight initial preference for the black alley,
but the terminal data did not depend importantly on which alley
received which reward.

III. Experimental Designs and Results

A. EXPERIMENT 1. JOINT EXTINCTION FOLLOWING PARTIAL
AND CONTINUOUS REINFORCEMENT
The firstexperiment followed the design described in the introduction.
Sixteen rats were given a large reward every trial on which they ran
down one alley (100 % 9A) and were given the same large reward on an
irregular half of the trials that they ran down the other alley (50 yo9A).
These conditions prevailed for a relatively large number of training
trials (42 days at six trials per day, totaling 126 trials in each alley before
extinction), the expectation being that the combination of a large
reward and prolonged training would maximize the chance that the
continuously reinforced response would undergo relatively rapid
extinction. It will be recalled that the rats would be expected to show a
temporary reversal of preference if incentive motivation following
partial reinforcement persists longer than that following continuous
reinforcement.
The results are shown in Fig. 1, the wide-lined curve showing the
percent choice of the continuously reinforced alley during the last 12
days of original acquisition and throughout 36 days of joint extinction.
It is apparent that no reversal occurred. The curve became somewhat
unstable during the later stages of nonreinforcement but never dropped
below the 50 yolevel. This result indicates that incentive motivation did
not persist longer in the partial-reward alley.
One possible complication with this interpretation of the results of
this design arises from the concept of the generalized partial reinforce-
ment effect (Brown & Logan, 1965). This effect refers to the fact that an
instrumental response that is, in its own right, reinforced continuously
may nevertheless show increased resistance to extinction if a similar
response has received partial reinforcement. Persistence trained in one
situation tends to generalize to similar situations. Evidence that such a
phenomenon occurred with respect to the instrumental performance
measure in the present design is also depicted in Fig. 1 by the narrow-
lined curves connecting solid points. These show the response speeds
separately for the continuously and partially reinforced responses by
the rats encountering both conditions and given a choice between them.
It is clear that there was little difference in the speed measure during
 Incentive Theory and Changes in Reward 13

extinction, the rats running, if anything, somewhat faster in the alley

that had been continuously reinforced.
This phenomenon is emphasized by observing the solid narrow-lined
curve connecting open points in Fig. 1 which shows the speed of respond-
ing during extinction by a control group given only continuous reinforce-
ment in a single alley. It is clear from comparing the two solid curves

Extmguish both (n.16)

42 days acquisition

nl -3
-P
8
t

---
5
.-
50- -2
-2
B
0)

P
100%9 A vs 50% 9 A n
v)

MChoice
25 - C. Speed (100%) -1
e-* Speed ( 5 0 % )
W C-speed (100%)
o--o C-speed (50%)
I 1 I
X-3 X-2 X-1 1 2 3 4 5 6 7 8 9
Blocks of 12 trials (each response)

FIG.1. Choice behavior during joint extinction following initial training with
partial and continuous reinforcement in two alternative alleys. Graphed are the
last 12 of 42 days of acquisition and 36 days of extinction. Speed data in the two
alleys are also depicted, together with the speeds of control groups that received
only partial or only continuous reinforcement before extinction.

that extinction proceeded more rapidly when no other response was

included that was trained with partial reinforcement. A converse effect
can also be seen in Fig. 1 by referring to the dashed narrow-lined curve
connecting open points, which shows ruining speed during extinction
following partial reinforcement for a control group that did not also
encounter joint extinction of a continuously reinforced response. The
apparent greater resistance to extinction following partial reinforcement
when extinguished separately was evaiuated statistically by determining
for each rat the trial block on which his speed first fell below the midpoint
between his individual asymptotes a t the end of training and extinction.
The difference was highly significant by the Mann-Whitney U Test
(Z = 4.26, df = 49, p < .01).
14 Frank A. Logan

Accordingly, it must be recognized that two generalization effects on

instrumental performance are embedded in the joint extinction design,
the one being increased resistance to extinction following continuous
reinforcement and the other being reduced resistance to extinction
following partial reinforcement. Whether these effects are indeed critical
for the conclusion suggested by the choice data depends upon the
conceptual basis for the generalization effects themselves. If they
represent incentive motivational effects, then the choice data are clearly
biased against finding any differences in preference during joint extinc-
tion. If, however, the generalization effects represent transfer of instru-
mental performance factors, such as running habits in the presence of
cues of nonreinforcement, then the preference data are uniquely sensitive
to changes in incentive motivation. Although this latter interpretation
seems at least equally probable, additional studies were run to search
for possible differential incentive effects.

B. EXPERIMENT FOLLOWING
2. REACQUISITION ORIGINAL ~ R A I N I N U
WITH PARTIAL
AND CONTINUOUS
REINFORCEMENT
A somewhat different approach to the question of whether or not
extinction has differential effects on incentive motivation following
partial and continuous reinforcement arises in the context of reacquisi-
tion. Since performance during extinction differs radically following
these training conditions, the effects on incentive might differ in a way
that could be detected by the reintroduction of reward. Specifically, for
example, if continuously reinforced rats stop instrumental performance
rapidly because of other-than-incentive factors such as a generalization
decrement in habit, then they might have more residual incentive
motivation than partially reinforced rats who persist longer in making
the instrumental response. Hence, reacquisition might progress at
different rates following extinction after initial training with partial or
continuous reinforcement.
To evaluate this possibility, the control groups shown in Fig. 1 were
run for an additional 60 trials with reward again present. Reacquisition
conditions were run under continuous reinforcement for some rats and
under partial reinforcement for others in order to control for possible
generalization decrements due to change from the original training
conditions. The results are plotted in Fig. 2. It is apparent that all
groups reacquired the running response rapidly, and somewhat more
rapidly with continuous than with partial reinforcement during this
stage of the experiment. But no noticeable differences occurred during
reacquisition as a residual from the original acquisition reinforcement
conditions. Accordingly, in spite of dramatic differences in instrumental
 Incentive Theory and Changes in Reward 15

performance during extinction following partial and continuous rein-

forcement, there are no apparent differences in the persisting effects of
nonreinforcement insofar as these can be seen in instrumental per-
formance after reintroduction of reward.

Reacquisition

50% original acquisition

t----4 0 Kx)% reacquisition

0 50% reacquisition

1 , I ,
x-3 x-2 x-1 1 2 3 4 5
Blocks of 12 trials

FIG. 2. Speeds during reacquisition by groups originally trained with either

partial or continuous reinforcement and split after extinction to receive reacquisi-
tion training with either partial or continuous reinforcement. Graphed are the last
12 trials of extinction (the earlier data being shown as the control groups in Fig. 1)
and 60 trials of reacquisition.

C. EXPERIMENTS OF CHOICEFOLLOWING
3 AND 4. REVERSAL
PARTIAL
AND CONTINUOUS REINFORCEMENT
I n an attempt to minimize the possible role of generalization effects,
a between-groups design was employed to compare the rates of loss of
incentive motivation following partial and continuous reward. Two
groups of rats received a large reward in one alley, one group receiving
the reward on every trial (100 yo9A) and the other group receiving the
large reward on an irregular half of their trials in that alley (50 yo 9A).
Both groups received a small reward continuously (1OOyo1A) in the
other alley. The selection of these amounts of reward was intended to
ensure that both groups would initially prefer the large-reward alterna-
tive, although presumably the incentive basis for this preference would
be somewhat greater with continuous than with partial reinforcement.
16 Frank A. Logan

After clear evidence of such preference for the larger reward was
apparent (28 days at six trials per day), extinction conditions were
initiated in the large-reward alley for both groups. The small reward was
still given to both groups on every trial in the alternative alley. Accord-
ingly, both groups would be expected ultimately to come to prefer the

'
(-3 X-2 X-1 I 2 3 4 5

Blocks of 4 days (24 trials)

6 7 8 9

FIG. 3. Choice behavior during reversal by groups initially given partial or

continuous large reward opposed to continuous small reward, and then extin-
guished in the large-reward alternative. Graphed are the last 12 of 28 days of
acquisition and 36 days following the change. Speeds for each group in each
alternative are also depicted.

small-reward alley since it was now pitted against no reward. By using

this procedure, the incentive motivation that persisted in the large-
reward alley during extinction could be monitored relative to that
maintained in the small reward alley. If incentive motivation persists
longer following partial reinforcement, reversal toward a small reward
should be slower following partial large reward than following continuous
large reward.
The results of this experiment are shown graphically in Fig. 3. The
wide-lined curves again show percent choice of the originally large
reward alley, beginning during the last 12 days of exposure to the
original conditions and proceeding through 36 days during which
extinction prevailed in that alley. It is apparent that reversal was not
 Incentive Theory and Changes in Reward 17

particularly rapid for either group and that there were no noticeable
differences in the rate at which reversal took place. Accordingly, even
in this between-groups design, there is no evidence that incentive moti-
vation persisted longer following partial than continuous reinforcement.
Because some possible generalized partial reinforcement effects may
also arise in this differential-reinforcement procedure, especially follow-
ing a relatively small amount of discrimination training (Brown & Logan,

loo -
Extinguish 9 A wntinw I A
42 days ocquiwtion

75 -

0 50% 9 A vs W % I A (17.22)

Choice
Speed (9A)
Speed ( I A )
I .
X-3 X-2 X-l 1 2 3 4 5 6 7 8 9
Blocks of 4 days (24 trials)

FIG.4. Choice behavior followinga replication of the conditions shown in Fig. 3

except that 50% overtraining was given before extinction of the large-reward
alternative. Graphed are the last 12 of 42 days of acquisition and 36 days following
the change. Speeds for each group in each alternative are also depicted.

1965), the procedure was repeated with new rats who were given suffi-
cient overtraining to minimize further any generalization effects.
Specifically, 42 days under the original reward conditions were given
before the large-reward alley underwenb extinction. This represents
50 yoovertraining beyond that given the previous groups in this design;
otherwise, the procedures were identical.
The results are shown in Fig. 4. The choice data are depicted by the
wide-lined curves and, although the partial group remained somewhat
above the continuous group during the later portions of the extinction
phase, this difference was absolutely quite small and did not approach
statistical significance. Again, the evidence suggests that incentive
motivation did not persist longer following partial reinforcement than
following continuous reinforcement, at least insofar as this is reflected
18 Frank A. Logan

in the rate at which this persisting incentive is able to compete success-

fully with the incentive produced by a continuous small reward.
Attention to the speed data also shown in Fig. 4 reveals several
relevant results. Comparing the two solid curves, it can be seen that
asymptotic running speed was greater with partial large reward than
with continuous large reward, and that this difference persisted during
extinction. Thus, although perhaps attenuated in this design, the
partial reinforcement effects on both acquisition and extinction were
evident in the instrumental response speed measures even though no
differential incentive-motivational effects were found in choice. A new
finding of potential significance is seen by comparison of the two dashed
curves in Fig. 4,which show the running speed of the two groups to the
small reward. There was a clear and statistically significant difference in
their running speeds at the end of original training, the group receiving
partial reinforcement in the other alley running faster for the small
reward than the group receiving continuous reinforcement in the other
alley ( t = 5.62, df = 43,p == .01).
There are two alternative interpretations of this speed difference. On
the one hand, it could be argued that a small reward suffers more by
contrast with continuous large reward than with partial large reward.
Fortunately, this interpretation can be ruled out by attending to
performance in the small-reward alley during extinction in the large-
reward alley. The group that had received continuous large reward
maintained a relatively steady speed in the small-reward alley while the
group that had received partial large reward gradually slowed down in
the small-reward alley. This suggests that the faster speed to the small
reward by the latter group resulted from some facilitating performance
effect generalizing from the partial-reward alley.
A reasonable account may be referred to as the generalized frustration
eflect. The presumption would be that being forced t o run in the small-
reward alley is somewhat frustrating because of the experience of a
large reward in the other alley (i.e., the reward for entering the start box
varies between large and small eventual reward). If the large reward is
given on only part of the trials, the rat also experiences anticipatory
frustration in that alley where he sometimes receives a large reward for
running. The resulting tendency to run in spite of anticipatsry frustra-
tion should generalize to the small-reward alley and the added frustration
drive should contribute to instrumental performance there. I n contrast,
the group receiving continuous large reward experiences anticipatory
frustration only when forced to run in the small-reward alley where
running is never reinforced with a large reward. Their instrumental
performance therefore reflects a more nearly perfect discrimination of
the reward conditions.
 Incentive Theory and Changes in Reward 19

D. THEOVERTRAININGEXTINCTION
EFFECT
A number of studies (e.g., North & Stimmel, 1960) have recently
appeared suggesting that overtraining of an instrumental response may
lead to reduced resistance to extinction of that response compared with
smaller amounts of training. This phenomenon, then, is another instance
in which differential rates of change in incentive motivation might
possibly be detected. Evidence against such an interpretation can be
seen by comparison of Figs. 3 and 4. It will be recalled that the only
difference between these studies concerned the number of training trials
given prior to extinction of the large-reward alley. Hence, if incentive
motivation decreased more rapidly following overtraining (with either
partial or continuous reward) then the reversal in Fig. 4 should have
been faster than that in Fig. 3. It is clear that no such difference occurred,
reversal of preference being essentially equally gradual in both instances.
Examination of the speed curves, however, does show a more precipitous
decline after overtraining. But there is no evidence that the overtraining
extinction effect represents changes in the rate of extinction of incentive
motivation.
E. EXPERIMENT5 . EFFECT
OF AMOUNTOF REWARDON
TO EXTINCTION
RESISTANCE MOTIVATION
OF INCENTIVE
Recent evidence (e.g., Hulse, 1958) indicates that, a t least after a
substantial amount of training, instrumental extinction proceeds more
rapidly with large as compared with small rewards. Experiment 5 was
designed to determine whether this phenomenon might reflect an effect
on the rate of change in incentive motivation. Furthermore, since the
amount-of-reward effect on extinction appears to be specific to con-
tinuous reinforcement, partial reward conditions were also included.
Two groups of rats received a large reward in one alley and a small
reward in the other alley, one group receiving both of these rewards
continuously and the other group receiving both of these rewards on an
irregular 50 yopartial reinforcement schedule. After 42 days of acquisi-
tion under these conditions, both alleys were subjected to joint extinc-
tion. Consistent with the logic developed previously, if extinction of
incentive motivation proceeded more rapidly following the large as
compared with the small reward, preference should show a temporary
reversal toward the small-reward alley.
The results are shown in Fig. 5. The heavy-lined curves show the
choice data for the two groups separately, and it is clear that no reversal
occurred. Instead, both curves drifted gradually toward indifference in
response to continued nonreinforcement. Attention to the speed curves
does reveal that the rate of extinction as measured by instrumental
performance was greater following continuous large reward than
20 Frank A. Logan

following continuous small reward, and that this differential rate of

extinction did not occur comparing partial-large against partial-small
rewards. The speed data are thus consistent with those obtained in
between-group designs, but again the choice data do not indicate that
these effects are based on differentia1 rates of change in incentive
motivation.
100-
42 days acquisition
- *

75 -

0
._
0)

0
5
E 50-

B
0)

25 -

I 1 I
X-3 X-2 X-l 1 2 3 4 5 6 7 8 9
Blocks of 4 days (24 trials)
FIQ.5. Choice behavior during joint extinction following initial training with a
lmge reward in one alternative and a small reward in the other. One group received
both rewards continuously and the other group received both rewards on an
irregular half of their trials in each alley. Graphed are the last 12 of 42 days of
acquisition followed by 36 days of joint extinction. Speeds for m h group in each
alternative are also depicted.

The data in Fig. 5, incidentally, afford an additional confirmation of

the comparison of extinction rates following partial and continuous
reward as studied in Experiments 1, 3, and 4. I n this case, if differential
incentive persisted longer for the partial group than the continuous
group, choice should have dropped more slowly toward indifference.
The close overlap of the choice curves in Fig. 5 indicates that no such
rate changes were involved.

F. EXPERIMENT 6. INCENTIVE CONTRAST

The phenomenon of incentive contrast has had a somewhat rocky
history. The term refers to the possibility that the effective incentive
value of a particular amount of reward depends in part upon the context
 Incentive Theory and Changes in Reward 21

in which it is embedded. Specifically, a small reward might have less

incentive value if the subject is accustomed also to receiving a larger
reward in the same or a similar situation and, conversely, a large reward
might have still greater incentive value if given in a context containing
smaller rewards. I n short, the incentive motivation produced by a
reward may be relative rather than absolute.
There are reasonable a priori grounds for anticipating such a pheno-
menon. A reward, after all, is a stimulus which must be perceived by the
subject, In principle, therefore, the laws of perception should apply as
much to reward stimuli as to any other, and one pervasive perceptual
principle is that of contrast. The size, weight, intensity, etc. of stimuli
are known to be perceived in relation to the context (Sherif, Taub, &
Hovland, 1958), including events occurring somewhat previously in
time. This argument would lead one to suspect that some contrast
effects might be found in the area of incentive motivation.
It is, however, important to distinguish clearly between a reward as a
stimulus event and the hypothetical construct of incentive motivation.
Scaling the incentive value of different rewards might, at first thought,
appear to be a psychophysical problem and indeed, certain gross
similarities are evident. For example, a constant absolute increase in the
amount of reward produces a progressively smaller increase in incentive
motivation the larger the amounts involved (Logan, 1965). However,
the “perceived size” of the reward, while perhaps a necessary precurser
of its incentive value, is not necessarily equivalent to incentive motiva-
tion. For example, a rat might be able to detect the difference between
20 and 30 pellets of reward when viewed as stimuli, but the latter might
not produce greater incentive motivation than the former. I n short,
incentive value is a property of some stimuli, namely rewards, which
should be distinguished from the cue value of such stimuli.
Early evidence suggesting incentive contrast was provided inde-
pendently by Crespi (1944) and Zeaman (1949). These investigators
trained hungry rats in a simple instrumental running response first with
either a large or a small amount of food reward. Subsequently, the reward
sizes were switched, and performance npt only changed in the appro-
priate directions but appeared to overshoot the appropriate levels. That
is to say, the rats that were shifted from a small to a large reward not only
increased in running speed, but ran faster than the r&tsthat had been
trained initially with the larger reward; conversely, the rats that were
shifted to the bmaller reward a t least temporarily ran slower than the
rats initially trained with the smaller reward. These data suggested that
the new value of reward contrasted with the earlier value that the rats
had been accustomed to receiving.
These results have been criticized (Spence, 1956) on the grounds that
22 Frank A. Logan

the original training may not have been carried fully to asymptote and
that the group shifted to the larger reward may have run faster not
because of any incentive contrast effect but because additional habit
was being acquired during the postshift trials. Subsequent studies have
tended to favor this interpretation ;a number of studies are now available
(e.g., Ehrenfreund and Badia, 1962) indicating that the upward shift in
amount of reward does not always produce the overshoot seen in the
earlier data. These latter studies, however, may also be criticized
because the larger reward employed has probably been near or at the
upper limit of potential incentive value. That is to say, a large reward
might appear still larger by contrast with concurrent or preceding
smaller rewards, but if the incentive value of the large reward is already
maximal, any such contrast would not appear in instrumental
performance.
Other studies that might reveal an incentive contrast phenomenon
have employed a concurrent exposure to different amounts of reward.
For example, Bower (1961) and Goldstein and Spence (1963) gave rats
a small reward in an alley of one color and a large reward in another
alley. The performances of these rats were separated for the two alleys
and compared with control rats receiving all of their training with either
the large or small reward. These results have been consistent in showing
no contrast effect with respect to the large reward although some of the
data suggest that performance with a small reward may be inferior if
the rats are concurrently encountering a large reward in a similar
situation. The same possible criticism with respect to the upper limit of
incentive applies and, in addition, there is the question of the extent to
which incentive effects generalize between the situations so as to provide
a basis for contrast. The unresolved issue is the degree to which the
situations would have to be similar for different rewards to show any
effective contrast as opposed to complete discrimination of the rewards
appropriate to the different situations.
The limiting case of similarity, of course, is to give different rewards
in the same situation. This procedure of varied reward has been studied
most systematically by Beier (see Logan, 1960) but the inability to infer
the incentive value of the rewards separately makes any interpretation
based on contrast highly speculative. However, suggestive analyses in
favor of such a phenomenon were reported.
Some data bearing on this issue were presented in Fig. 4. Rats
received a large reward in one alley and a small reward in the other alley
followed by extinction in the large-reward alley while the small reward
continued. If one refers to the running speed in the small-reward alley
when it was initially paired with a large reward and subsequently paired
with no reward, it is apparent that no differences appeared. That is to
 Incentive Theory and Changes in Reward 23

say, the rats ran as well toward one pellet reward when they were
receiving nine pellets in the other alley as when they were receiving
nothing in the other alley. This presumably indicates that incentive was
more or less perfectly discriminated and that no contrast effects, upward
or downward, occur in this situation.

loo -

75 -
.-
5
'c 50-
e
&!
0 5A vs 2 A then 5 A vs 5A h = O )
0 5 A vs 2 A then 2 A vs 2 A (n=IO)
Change DmowIt
42 days acquisition

-
-
Choice
25 - M Speed ( 5 A then 5 A )
*-* Speed ( 2 A then 5A)
Speed ( 2 A then 2 A )
Speed ( 5 8 then 2 A )

Blacks of 24 trials

FIG.6. Choice behavior after reward wm equalized following initial training with
differential reinforcement. Equalization was achieved by increasing the small
reward for one group and by decreasing the large reward for the other group.
Graphed are the last 12 of 42 days of acquisition followed by 36 days with equal
rewards. Speeds for each group in each alternative are also depicted.

Experiment 6 was intended to study the possibility of an incentive

contrast effect within the context using the within-groups design
permitting observation of choice behavior. Two groups were initially
given a relatively large reward (5A) in one alley and a relatively small
reward (2A) in the other alley. The choice of these amounts was intended
to ensure some initial preference for the larger reward but still leaving
room both above and below the initial values for .possible incentive
contrast effects to be detected. Following 42 days of acquisition under
these conditions, the amount of reward was equated in the two alleys.
This was accomplished for one group by increasing the reward in the
one alley from two to five pellets ; for the other group, the reward in the
other alley was decreased from five to two pellets. If the alleys subjected
to a change in amount of reward should show incentive contrast effects,
24 Frank A. Logan

preference should at least temporarily reverse to the alley originally

containing the small reward. For the one group, that alley was shifted
to five pellets which could contrast above the five pellets continued in
the other alley; for the other group, that alley continued at two pellets
while the other alley could be contrasted below that level because of
reduction to two pellets.
The results of this study are shown in Fig. 6. The wide-lined curves
show that choice behavior continued to reveal a quite strong preference
for the alley originally containing the larger reward, even though the
rewards were now equated. Clearly, no tendency for reversal of preference
occurred. Appropriate changes did occur in the speed measures corre-
sponding to the changes in amount of reward, and small but statistically
insignificant elation and depression effects were obtained. The changes
observed in the instrumental response measure, however, apparently do
not reflect incentive motivational factors.
G . EXPERIMENT 7. RATEOF INCREASE IN INCENTIVE MOTIVATION
The previous studies in this series have concentrated primarily on
possible changes in the rate at which incentive motivation is lost.
Possible effects on the incremental rate are, of course, of equal import-
ance. Instrumental studies showing such an effect are not available, but
a likely candidate would be partial reinforcement. It is known that
classically conditioned responses shift gradually following a change
from partial to continuous reinforcement (Spence & Trapold, 1961) and
a similar phenomenon might occur with respect to incentive motivation.
Experiment 7 sought to evaluate this possibility. Two groups of rats
received an intermediate-sized reward (3A) continuously in one alley ;
one group received nothing in the other alley and the second group
received partial large reward (50% 5A). The choice of these values was
based on our earlier work (Logan, 1965) on quantifying incentive
motivation, since 3A would be expected to be preferred, a t least to some
extent, to 60% 5A.
Following 42 days of acquisition under these conditions, both groups
were given the large reward (5A) continuously in the one alley while 3A
was continued in the other alley. Thus, preference should reverse toward
the alley now containing five pellets (although the degree of preference
over three pellets would not be expected to be complete), and interest
resides in the rate at which this reversal takes place. If the group in-
creased from nothing to five pellets acquired that level of incentive
motivation more rapidly than the group accustomed to partial-five
pellets, then reversal could occur more rapidly in the former group.
The results are plotted in Fig. 7. It is clear that the groups attained
different levels of preference during the original 42 days of acquisition,
 Incentive Theory and Changes in Reward 25

thus making the rate comparisons somewhat difficult. Even so, however,
there is no indication that the group shifted from partial to continuous
took longer to reverse than the group shifted from nothing to continuous
large reward. If anything, the latter obtained, presumably reflecting the
initial differencein preference. Hence, these data do not suggest that the
incremental rate parameter for incentive motivation is modified by
prior reinforcement history.

-
- Choice
Speed 3A
o--* Speed nil then 100%5A
O----o Speed 50X5A then K)o%5P
(-3 x-2 x-l 1 2 3 4 5 6 7 8 9
Blocks of 24 trials

FIG.7. Choice behavior during reversal when a relatively small reward was
initially pitted against either nothing or partial reinforcement with a relatively
large reward and was subsequently pitted against continuous reinforcement.
Graphed are the last 12 of 42 days of acquisition followed by 36 days for reversal.
Speeds for each group in each alternative are also depicted.

IV. Discussion
Changes in incentive motivation are essential aspects of contemporary
theories of learning, but little explicit attention has been paid to the
rate at which such changes take place. Were incentive motivation
mediated by a response mechanism such as rg-sg, then the laws of
classical conditioning would suffice to describe incentive change effects.
However, since reasonable questions can be raised about a medietional
approach to incentive motivation, the present research program
attempted to evaluate the question empirically.
26 Frank A. Logan

Incentive motivation is only one of the conceptual determinants of

instrumental performance, and hence cannot be measured directly.
Nevertheless, it is assumed that choice between two alternatives that
are otherwise equivalent can be used to index their relative incentive
values. Accordingly, changes in choice behavior were monitored follow-
ing a variety of changes in reward, including partial reinforcement,
overtraining, and amount of reward. I n each case, the choice data failed
to reveal any effect of these procedures on the rate a t which incentive
motivation changed. Simultaneous measurements of the speed of
instrumental responding showed effects which, while somewhat attenu-
ated, were consistent with those found in between-group designs. They
are also consistent with the results of the extensive series of studies by
Amsel (1967) using the speed measure in closely related within-subject
experiments.
The implication, then, is that the rate of change in incentive motiva-
tion is a state parameter of the subject and is not affected by prior
reinforcement history.
As for any conclusion of this potential significance, other types of
procedures and other species of subjects should be studied before
generality can be assumed. Indeed, several reservations might be noted
in the results of the present program. Most obviously, generalization
effects between the two alleys may have been sufficiently great that
evaluation of their independent properties could not be assessed. This
possibility has been mentioned earlier and data suggesting this problem
were noted. However, the fact that appropriate changes in response
speed did occur in many of the studies combined with the recognition
that consistent differences have been obtained in earlier studies using
these procedures make this reservation about the conclusion somewhat
doubtful.
Perhaps a more critical problem concerns the sensitivity of the choice
measure to transitory changes in incentive motivation. The gradual
reversals of preference obtained in all of the studies suggests that the
choice response includes some habit factors over and above any innate
brightness/position preferences and the prevailing reward conditions.
This is most evident in Experiment 6 (Fig. 6) where choice behavior
continued with little change even after the rewards had been equalized
in the two alleys. That is to say, choice habits acquired on the basis of
initial differences in reward tended to persist when no other basis for
preference existed. It is possible, therefore, that these habits cannot be
overcome rapidly enough to show the temporary effects that changed
rewards are having on incentive motivation. I n short, changes in
preference may lag behind changes in incentive motivation.
This problem cannot be resolved within the context of these studies,
 Incentive Theory and Changes in Reward 27

and must temper the conclusion accordingly. However, it would seem

reasonable to expect that, were conceptually significant effects involved,
some suggestion of these would have appeared somewhere in the total
pattern of results. Thus, if rats indeed lose incentive motivation faster
following continuous than partial reinforcement, one would expect that
they would reach a stage in which incentive had extinguished in the one
alley while some incentive remained in the other and that their choice
of behavior would reflect this fact. I n no case were there indications of
such phenomena.
Less significant effects on incentive motivation may have occurred.
Specifically, for example, the initial incentive value following con-
tinuous reinforcement is greater than that following partial reinforce-
ment. It would be possible for the relative rate of loss of incentive
motivation to be somewhat faster in the former procedure, but not
sufficiently faster to cross the latter in order to motivate a reversal of
preference. This possibility can be tested by using conditions which
initially produce more nearly equal incentive levels but which differ in
the pattern of reinforcement experienced.
It is also possible that more radical initial conditions may affect the
rate of change in incentive motivation. For example, Solomon has found
that dogs given preliminary exposure to unavoidable, inescapable
electric shocks subsequently may have difficulty learning to escape from
shock in another situation. This could mean that the development of
incentive motivation based on primary negative reinforcement is
affected by prior experimental history, although it is also possible that
such training has simply changed the initial response hierarchy to
primary motivation. In any event, the researches reported covered
a wide range of effects that are familiar in the learning literature
and failed to implicate incentive motivation as a major factor in the ob-
served phenomena. Alternative analyses of these, for example, the frus-
tration account of the partial reinforcement effect, are thus indirectly
supported.
The general conclusion suggested by these results is that the assump-
tion of invariant rates of change in incentive motivation is a tenable one.
When the incentive value of the reward received after a response differs
from the preexisting incentive motivation for that response, an appro-
priake change occurs a t a rate that is more or less independent of prior
experimental history. With specific reference to extinction, continued
nonreinforcement results in a loss of incentive to respond, but the rate
at which this decrease occurs does not differ with prior reinforcement
conditions. Confidence in this conclusion will depend, of course, on the
success enjoyed by further development of a quantitative model of
incentive motivation incorporating this assumption.
28 Frank A. Logan

V. Summary
The rate at which incentive motivation changes in response to a
change in the reward conditions was evaluated empirically by use of a
choice procedure with rats as subjects. Two alternatives initially
received different conditions of reinforcement followed by a change in
the conditions for one or both alternatives. The procedures studied were
selected from effects familiar in the literature of learning, including the
effects on extinction of partial reinforcement, amount of reward, and
overtraining. Changes in both amount and probability of reward were
also studied. I n no case was choice behavior affected by these changes
in the manner that would be expected if the rate at which incentive
motivation changes had been significantly affected.
During forced trials, running speeds reflected most of the familiar
phenomena of simple instrumental performance measures. Several novel
results were also obtained in the speed data, notably a reduction in the
partial reinforcement effect when extinction is conducted jointly with a
continuously reinforced response, and a generalized frustration effect
reflected in faster running for small reward when combined with partial
as compared with continuous large reward, In general, the speed data
confirmed that the rats were performing appropriately to the conditions
of reinforcement in each alternative.
Reservations about the generality of the choice procedure were noted,
especially the possible generalization effects between the alternatives and
the observable persisting habits of choice. Nevertheless, the consistency
of the results suggests the conclusion that the rate of change in incentive
motivation is an invariant state parameter of the subject and is not
significantly aRected by prior reinforcement history.

REFERENCES
Amsel, A. Frustrative nonreward in partial reinforcement and discrimination
learning : Some recent history and a theoretical extension. Psychological Review,
1962, 69, 306-328.
Amsel, A. Partial reinforcement effects on vigor and persistence. In K. W. Spence
and J. T. Spence (Eds.,),The psychology of learning and motivation: Advances i n
research and theory, Vol. 1. New York: Academic Press, 1967. Pp. 1-65.
Black,R. W. On the combination of drive and incentive motivktion.Psychological
Review, 1965, 72, 310-317.
Bower, G. H. Partial and correlated reward in escape learning. Journal of Experi-
mental Psychology, 1960,59, 126-130.
Bower, G. H. A contrast effect in differential conditioning.Journal of Experimental
Psychology, 1961,62, 196-199.
Brown, R. T., t Logan, F. A. Generalized partial reinforcement effect. Journal of
Comparative and Physiological Psychology, 1965, 60, 64-69.
Capaldi, E. J. Partial reinforcement:An hypothesis on sequential effects. Psycho-
logical Review, 1966,73,459-477.
 Incentive Theory and Changes in Reward 29

Cofer, C. N., & Appley, M. H. Motivation: Theory and research. New York: Wiley,
1964.
Crespi, L. P. Amount of reinforcement and the level of performance. Psychological
Review, 1944,51, 341-357.
Davenport, J. W. Spatial discrimination and reversal learning based upon dif-
ferential percentage of reinforcement. Journal of Comparative and Physiological
Psychology, 1963,56, 1038-1043.
Egger, M. D., & Miller, N. E. When is a reward reinforcing? : An experimental
study of the information hypothesis. Journal of Comparative and Physiological
Psychobgy, 1963,56, 132-137.
Ehrenfreund, D., & Badia, P. Response strength as a function of drive level and
pre- and post-shift incentive magnitude. Journal of Experimental Psychology,
1962,63,468-471.
Gleitman, H., Nachmias, J., & Neisser, U. The S-R reinforcement theory of
extinction. Psychological Review, 1954,61, 23-33.
Goldstein, H., & Spence, K. W. Performance in differential conditioning as a
function of variation in magnitude of reward. Journal of Experimental Pay-
choology, 1963, 65, 86-93.
Guthrie, E. R. Reward and punishment. Psychological Review, 1934,41,450-460.
Hull, C. L. Principles of behavior. New York: Appleton, 1943.
Hull, C. L. A behavior system. New Haven: Yale University Press, 1952.
Hulse, S. H., Jr. Amount and percentage of reinforcement and duration of goal
confinement in conditioning and extinction. Journal of Experimental Psychology,
1958,56,48-57.
Kaplan, R., Kaplan, S., & Walker, E. L. Individual differences in learning aa a
function of shock level. Journal of Experimental Psychology, 1960, 60, 404-407.
Kendler, H. H. Drive interaction: I. Learning as a function of the simultaneous
presence of the hunger and thirst drives. Journal of Experimental Psychology,
1945, 35, 96-109.
Koch, S. Clark L. Hull. In W. K. Estes, et al., Modern learning theory. New York:
Appleton, 1954. Pp. 1-176.
Lewis, D. J., & Kent, N. D. Attempted direct activation and deactivation of the
fractional anticipatory goal response. Psychological Reports, 1961, 8 , 107-1 10.
Logan, F. A. Incentive. New Haven: Yale University Press, 1960.
Logan, F. A. Decision making by rats. Journal of Comparative and Physiological
Pwchlogy, 1965,599 1-12 and 246-251.
Logan, F. A., & Wagner, A. R. Reward and punishment. Boston: Allyn & Bacon,
1965.
Miller, N. E. Studies of fear as an acquirable drive: I. Fear as motivation and fear-
reduction as reinforcement in the learning of new responses. Journal of Experi-
mental Psychology, 1948, 38, 89-101.
Miller, N. E. Some reflections on the law of effect produce a new alternative to
drive reduction. In M. R. Jones (Ed.), Nebrmka qmposium on motivation.
Lincoln, Nebr.: University of Nebraska Press, 1963. Pp. 65-112.
Mowrer, 0. H. Learning theory and behavior'. New York: Wiley, 1960.
Nefzger, M. D. The properties of stimuli associated with shock reduction. Journal
of Experimental Psychology, 1957, 53, 184-188.
North, A. J., & Stimmel, D. T. Extinction of an instrumental response following
a large number of reinforcements. Psychological Reports, 1960, 6 , 227-234.
Novin, D., & Miller, N. E. Failure to condition thirst induced by feeding drv food
to hungry rats. J o u d of Comparative and Phyaiologieal P&chologyy 1962, 55,
373-374.
30 Frank A. Logan

Pavlov, I. P. Conditioned reflexes. (Translated by G. V. Anrep.) London and New

York: Oxford University Press, 1927.
Seward, J. P. Reinforcement and expectancy: Two theories in search of a con-
troversy. Psychological Review, 1956, 63, 105-113. (a)
Seward, J. P. Drive, incentive, and reinforcement. Psychological Review, 1956, 63,
195-203. (b)
Shapiro, M. M. Temporal relationship between salivation and lever pressing with
differential reinforcement of low rates. Journal of Comparative and Physiological
Psychology, 1962,55, 567-571.
Sheffield, F. D. A drive induction theory of reinforcement. I n R. N. Haber (Ed.),
Current research i n motivation. New York: Holt, 1966. Pp. 98-121.
Sheffield, V. F. Extinction as a function of partial reinforcement and distribution
of practice. Journal of Experimental Psychology, 1949, 39, 511-526.
Sherif, M., Taub, D., & Hovland, C. I. Assimilation and contrast effects of anchoring
stimuli on judgments. Journal of Experimental Psychology, 1958, 55, 150-155.
Spence, K. W. Behavior theory and conditioning. New Haven: Yale University
Press, 1956.
Spence, K. W. Behavior theory and learning. Englewood Cliffs, N.J. : Prentice-Hall,
1960.
Spence, K. W., Bergmann, G., & Lippitt, R. A study of simple learning under
irrelevant motivational-reward conditions. Journal of Experimental Psychology,
1950, 40, 539-551.
Spence, K. W., & Trapold, M. A. Performance in eyelid conditioning as a function
of reinforcement schedules and change in them. Proceedings of the National
Academy of Sciences of the U.S., 1961,47, 1860-1868.
Tolman, E. C. Purposive behavior in animals and men. New York: Century, 1932.
Trapold, M. A. The effect of incentive motivation on an unrelated reflex response.
Journal of Comparative and Physiological Psychology, 1962, 55, 1034-1039.
Zeaman, D. Response latency as a function of the amount of reinforcement.
Journal of Experimental Psychology, 1949,39, 466-483.
Zimmerman, D. W. Durable secondary reinforcement :. Method and theory.
Psychological Review, 1957,64, 373-383.