Journal of Experimental Psychology: Copyright 2004 by the American Psychological Association

Animal Behavior Processes 0097-7403/04/$12.00 DOI: 10.1037/0097-7403.30.3.240

2004, Vol. 30, No. 3, 240 –250

A Partial Reinforcement Extinction Effect Despite Equal Rates of

Reinforcement During Pavlovian Conditioning

Mark Haselgrove Aydan Aydin

Cardiff University University of Mersin

John M. Pearce
Cardiff University

In 4 experiments rats received appetitive Pavlovian conditioning followed by extinction. Food accom-
panied every trial with the conditioned stimulus (CS) for the continuously reinforced groups and only half
of the trials for the partially reinforced groups. In contrast to previous experiments that have compared
the effects of partial and continuous reinforcement, the rate at which food was delivered during the CS
was the same for both groups. The strength of the conditioned response during extinction weakened more
rapidly in the continuously than in the partially reinforced groups. The results demonstrate that the partial
reinforcement extinction effect is a consequence of the nonreinforced trials with the CS, rather than the
rate at which the unconditioned stimulus is delivered during the CS.

For many years the partial reinforcement extinction effect ries of learning have been based on the assumption that the
(PREE) was a mainstay of research in the field of animal learning strength of the association between a stimulus and a response, or
and behavior. Within the context of instrumental conditioning, the between a CS and a US, determines the strength of conditioned
PREE can be defined as a resistance to the detrimental effects of responding. The nonreinforced trials that occur with partial rein-
extinction following trials in which responding is intermittently forcement are predicted to result in a weaker association than when
rather than consistently paired with reinforcement (for a review see a continuous reinforcement schedule is used, and it therefore
Mackintosh, 1974). So widespread was the PREE in studies of follows that extinction will progress more rapidly after partial than
instrumental conditioning—and such was the paucity of evidence after continuous reinforcement.
for the effect within Pavlovian conditioning—that for some time To explain the PREE in instrumental conditioning, several the-
the PREE was thought to differentiate these two forms of condi- orists have proposed that nonreinforced trials have an effect in
tioning (Kimble, 1961). However, the possibility that the PREE addition to weakening associative strength. In particular, such
contradicted the single-process rule of learning (Guthrie, 1935; trials are assumed to generate an internal state of frustration (e.g.,
Hull, 1943) has not stood the test of time. Many studies of Amsel, 1958, 1992) or a memory trace of nonreinforcement (e.g.,
Pavlovian conditioning have now demonstrated that a conditioned Capaldi, 1967). On the one hand, when extinction takes place after
stimulus (CS) that is intermittently paired with an unconditioned training with a continuous reinforcement schedule, the introduc-
stimulus (US) will show a resistance to extinction relative to a CS tion of this state or trace for the first time will change the context
that is consistently paired with a US (e.g., Fitzgerald, 1963; in which the response occurs and thereby facilitate extinction
Fitzgerald, Vardaris, & Teyler, 1966; Gibbon, Farrell, Locurto, through a generalization decrement (see Mackintosh, 1974, pp.
Duncan, & Terrace, 1980; Gibbs, Latham, & Gormezano, 1978; 440 – 443). On the other hand, training with a partial reinforcement
Pearce, Redhead, & Aydin, 1997; Rescorla, 1999b; Slivka & schedule provides an opportunity for responses to be reinforced in
Bitterman, 1966). This has not prevented the PREE from continu- the context of nonreinforcement and thereby minimizes its disrup-
ing to be a thorn in the side of theories of learning, however. From tive effect during extinction.
Thorndike (1911) onward (e.g., Rescorla & Wagner, 1972), theo- Pearce et al. (1997) have used a similar line of reasoning to
explain the effects of partial reinforcement on the extinction of
Pavlovian conditioned responding. According to this view, partial
reinforcement will allow the association between the CS and US to
Mark Haselgrove and John M. Pearce, School of Psychology, Cardiff develop within the context of the aftereffects of nonreinforced
University, Cardiff, Wales, United Kingdom; Aydan Aydin, Department of trials and therefore reduce the generalization decrement engen-
Psychology, University of Mersin, Mersin, Turkey.
dered by these aftereffects during extinction. It should be evident
This research was supported by a grant from the Biotechnology and
Biological Sciences Research Council of the United Kingdom.
from this perspective that the presence of nonreinforced trials with
Correspondence concerning this article should be addressed to Mark the CS during acquisition is crucial for a successful demonstration
Haselgrove or John M. Pearce, School of Psychology, Cardiff University, of the PREE in Pavlovian conditioning. Gallistel and Gibbon
Cardiff, Wales CF10 3YG, United Kingdom. E-mail: haselgrovem@ (2000, 2002), however, have argued that another property of
[Link] or pearcejm@[Link] partial reinforcement is responsible for the PREE in Pavlovian

240
 RATE OF REINFORCEMENT AND THE PREE 241

conditioning. According to Gallistel and Gibbon, extinction after The continuously reinforced group also received food at these
Pavlovian conditioning is the consequence of a decision to stop intervals, but on different trials. Experiments conducted in our
responding. This decision is based on the value of two variables: laboratory (Haselgrove & Pearce, 2003) have shown that the latter
IR̂ICS—the average duration of exposure to the CS between schedule is effective in producing conditioned responding. In
successive presentations of the US during conditioning and addition, it has the advantage of ensuring that both groups in the
ÎCS no R —the duration of exposure to the CS since the US was last current experiment are reinforced both at the termination of the CS
delivered during the CS (Gallistel & Gibbon, 2000, pp. 306, 343). and 5 s earlier. The design of the conditioning stage is shown in
When the ratio of these two variables (ÎCS no R/IR̂ICS) exceeds a Figure 1. On the completion of conditioning, both groups received
decision threshold, the animal stops responding and extinction can nonreinforced trials with the 10-s CS. If the PREE is a conse-
be thought to have occurred. As a concrete example of the appli- quence of presenting nonreinforced trials with the CS during
cation of this theory to the PREE, consider two groups of rats conditioning then, in contrast to the prediction of the theory of
undergoing appetitive Pavlovian conditioning. Both groups re- Gallistel and Gibbon (2000), extinction should progress more
ceive the same number of presentations of a 10-s tone within a rapidly in Group Continuous than in Group Partial.
session; food is presented after the CS for half the trials for a
partially reinforced group and all the trials for a continuously Method
reinforced group. In these circumstances, the value of IR̂ICS will be
10 s for the continuously reinforced group and 20 s for the partially Subjects. The subjects were 16 experimentally naive, male, hooded
Lister rats (Rattus norvegicus) supplied by Harlan Olac (Bicester, Oxon,
reinforced group. Therefore, the total duration of nonreinforced
England). Prior to the start of the experiment, at an age of approximately
exposure to the CS that must accumulate before the value of the 3 months, they were gradually reduced to 80% of their free-feeding
decision variable exceeds the decision threshold will be less for the weights. They were maintained at these weights throughout the experiment
former group than the latter. That is, conditioned responding will by being fed a restricted amount after each experimental session. The rats
cease sooner during extinction following continuous reinforcement were housed in pairs in a light-proof room in which the lights were on for
than following partial reinforcement. A prediction that can be 14.5 hr/day. The subjects were tested at the same time on successive days
derived from this account of extinction is that partial reinforcement during the period when the lights were on in their holding room.
will not affect the number of USs that must be omitted before the Apparatus. Eight conditioning chambers (24.5 ⫻ 23.0 ⫻ 20.0 cm)
subject decides to stop responding. In support of this prediction, were housed in separate light- and sound-attenuating chests. Exhaust fans
Gibbon et al. (1980) showed that the PREE was not evident when in each chest provided a background masking noise of 72 dB (C scale).
Three walls of each chamber were constructed from aluminum, one side
examined in terms of the omission of expected USs.
was clear acrylic, and the ceiling was translucent white acrylic. A 5-⍀
Unlike associative theories, the account of extinction offered by speaker located on the center of the ceiling delivered a 10-Hz click at an
Gallistel and Gibbon (2000, 2002) does not need to appeal to any intensity of 83 dB (C scale). There was a 5.0 ⫻ 6.0 cm recessed food
additional processes to explain the PREE. For this reason alone, magazine in the front wall into which 45-mg food pellets (traditional
the theory merits serious attention. Accordingly, the four experi- formula, P. J. Noyes, Lancaster, NH) could be delivered, its base was
ments reported in this article examine whether the rate at which the located 0.5 cm above the grid floor. Subjects were required to push open
US is presented during the CS is responsible for determining a clear acrylic flap in order to gain access to the magazine. Three pairs of
whether a PREE will be observed. In each experiment, at least two photodiode sensors were set into a rectangular frame 1-cm deep that
groups of rats received appetitive Pavlovian conditioning with surrounded the magazine entrance in such a manner that horizontal beams,
schedules that ensured they received the same rate of food delivery 5 mm in front of the closed magazine flap, were located 10, 20, and 30 mm
above the grid floor. Appropriate circuitry permitted an Archimedes mi-
during the CS. This was achieved either by manipulating the
crocomputer (Acorn Computers Ltd., Cambridge, England), which was
number of USs delivered during each trial or by manipulating the programmed in Arachnid (Paul Fray Ltd., Cambridge, England) to record
duration of the trials. For one group, but not the other, the schedule the interruption of these beams and to control the experimental events.
also contained nonreinforced trials with the CS. If the proposals of Procedure. All rats initially received a single session of magazine
Gallistel and Gibbon are correct, then the performance of the two training. During this 40-min session, one food pellet was delivered at
groups during extinction will be identical. However, if the PREE regular 1-min intervals. For this session only, the acrylic flap of the
depends on the presence of nonreinforced trials during acquisition, magazine was taped open. The subjects were then divided into two groups
then a difference between the groups should be observed during of equal size. In each of 12 sessions of conditioning there were 20
extinction. presentations of a 10-s clicker. The interval between the end of one

Experiment 1
In Experiment 1, two groups of rats received the same duration
of exposure to a CS during conditioning and also the same number
of presentations of the US. Every trial with the CS was accompa-
nied by food for the continuously reinforced group, whereas for the
partially reinforced group food was delivered on half of the trials
with the CS. The duration of each trial was 10 s for both groups.
To equate the groups in terms of the rate of reinforcement during
the CS, food was presented once on every trial to the continuously Figure 1. Procedure and design of Experiment 1. Time is indicated left to
reinforced group and twice on every reinforced trial to the partially right. Open boxes represent conditioned stimulus presentations. Black
reinforced group— once after 5 s and once at the end of the trial. boxes represent unconditioned stimulus presentations.
242 HASELGROVE, AYDIN, AND PEARCE

presentation of the clicker and the start of the next trial, the intertrial 76.50, but no effect of group (F ⬍ 1), and no Group ⫻ Session
interval (ITI), had a mean duration of 120 s (range ⫽ 90 –150 s). For Group interaction, F(11, 143) ⫽ 1.42. The mean durations of magazine
Continuous, the termination of half of the conditioning trials was paired activity for Groups Partial and Continuous across the 12 sessions
with the delivery of a single food pellet. For the remaining trials, a single of conditioning during the 10-s intervals prior to the conditioning
food pellet was delivered 5 s after the onset of the clicker. The sequence of
trials were 1.21 and 1.65 s, respectively. This difference was not
these trials was random with the constraint that no more than two trials of
the same type could occur in succession. Half of the presentations of the CS significant, t(13) ⫽ 2.12.
for Group Partial in each session were not paired with food. These trials The group mean difference scores are shown, in two-trial
occurred at the same point within the session as the trials with food blocks, for the two sessions of extinction in the right panel of
presented at the end of the clicker in Group Continuous. On each of the Figure 2. In both extinction sessions the mean duration of maga-
remaining trials, two food pellets were delivered. The first occurred 5 s zine activity declined more rapidly in Group Continuous than in
after the onset of the clicker and the second immediately after the end of Group Partial. A two-way ANOVA of difference scores, in two-
the clicker. Following the final session of conditioning, the rats received 2 trial blocks, revealed a significant effect of group, F(1, 13) ⫽ 8.72,
sessions of extinction. In each of these sessions there were 18 nonrein- and of two-trial block, F(17, 221) ⫽ 8.11, but no Group ⫻ Block
forced trials with a 10-s clicker. The ITI for these trials was again 120 s
interaction (F ⬍ 1). The mean durations of magazine activity
(range ⫽ 90 –150 s).
The measure of conditioning was magazine activity that was deemed to during the 10-s periods prior to the onset of the clicker for Groups
have taken place whenever the infrared beams in front of the magazine Partial and Continuous across the two sessions of extinction were
were broken. The duration of magazine activity was recorded for 10 s prior 0.36 and 0.22 s, respectively. The difference between these means
to every trial in both stages of the experiment. The duration of this activity was not significant, t(13) ⫽ 1.67.
was also recorded in the presence of the CS on nonreinforced trials in the Gallistel and Gibbon (2000) stated that extinction occurs when
partial group and on trials where food was delivered at the termination of subjects make a decision to stop responding. To identify when this
the CS in Group Continuous. Magazine activity was also recorded during decision was made in the present experiment, we used the extinc-
each extinction trial. A difference score was calculated by subtracting the tion criterion adopted by Gallistel and Gibbon (p. 306) of a decline
duration of magazine activity during the pre-CS period from the duration
to 50% of the preextinction rate of responding. Extinction was
of magazine activity during the CS period.
therefore deemed to be effective when the difference score was, for
two consecutive two-trial blocks, less than half that measured
Results and Discussion during the final session of conditioning. The number of trial blocks
A Type I error rate of p ⬍ .05 was adopted for all the statistical that elapsed before this criterion was met was 13.29 for Group
tests. One subject in Group Partial failed to engage in any maga- Partial and 8.25 for Group Continuous. This difference was sig-
zine activity and was discarded from the experiment. nificant, t(13) ⫽ 2.50.
The mean difference scores for Groups Partial and Continuous Given that both groups received the same rate of reinforcement
throughout the 12 sessions of conditioning can be seen in the left during the CS, the theory of Gallistel and Gibbon (2000) predicts
panel of Figure 2. There is an indication that conditioning was that they should perform similarly during the extinction stage of
initially more effective in Group Continuous than in Group Partial. the experiment. Thus the greater resistance to extinction that was
However, this observation was not supported by statistical analy- observed in Group Partial than in Group Continuous poses a
sis. A two-way analysis of variance (ANOVA) of individual dif- challenge to the explanation for the PREE that can be derived from
ference scores for the two groups for each of the 12 sessions of this theory. In contrast, the results are consistent with the sugges-
conditioning revealed a significant effect of session, F(11, 143) ⫽ tion that the PREE is a consequence of the nonreinforced trials that

Figure 2. The mean difference scores for Group Partial and Group Continuous across the 12 sessions of
conditioning (left panel) and across the 2 sessions of extinction (right panel) of Experiment 1.
 RATE OF REINFORCEMENT AND THE PREE 243

occur intermittently throughout conditioning (Pearce et al., 1997). If presenting food twice during each reinforced trial was respon-
There is, however, another way in which the results can be ex- sible for the subsequent slow rate at which responding weakened
plained. The groups differed not only in their reinforcement sched- during extinction with Group Partial of Experiment 1, then extinc-
ules but also with respect to the frequency with which food was tion should be more rapid in Group Continuous-1 than in Group
delivered within reinforced trials. Reinforcement for Group Con- Partial-2. In addition, extinction should be more rapid in Group
tinuous consisted of a single food pellet on a trial, whereas for Partial-1 than in Group Continuous-2. Alternatively, if the pres-
Group Partial it consisted of a single food pellet followed 5 s later ence of nonreinforced trials during conditioning determines the
by another pellet. It is conceivable that it was this difference, rather rate of extinction, then the strength of conditioned responding
than the different reinforcement schedules, that was responsible for should decline more rapidly in both of the continuously reinforced
the slower rate of extinction in the partial than the continuous groups than in either of the partially reinforced groups. Finally, the
group. A measure of support for this suggestion can be found in account of extinction provided by the theory of Gallistel and
experiments which have shown that extinction progresses more Gibbon (2000) leads to the prediction that extinction will progress
slowly after Pavlovian conditioning has taken place with a large most rapidly in Group Continuous-2, somewhat more slowly in
rather than a small US. For example, Wagner, Siegel, Thomas and Groups Partial-2 and Continuous-1 (which will not differ from
Ellison (1964) reported that conditioned salivation extinguished each other), and most slowly in Group Partial-1. The justification
more slowly if the CS had been paired with six pellets of food on for these predictions can be appreciated by an examination of
each trial instead of one pellet. If two food pellets separated by a Figure 3. According to the theory of Gallistel and Gibbon, extinc-
short interval constitute a larger US than one pellet, then whatever tion will progress most rapidly in a group whose CS has the
explanation is deemed appropriate for the results reported by highest rate of reinforcement. Each group in the current experi-
Wagner et al. may also apply to the present results. This possibility ment receives the same duration of exposure to the CS. Therefore
was tested in Experiment 2. the rate of extinction should be directly related to the number of
USs delivered during conditioning, with higher numbers of USs
resulting in faster extinction.
Experiment 2
There were four groups in this experiment, two of which re- Method
ceived the same treatment during conditioning as Groups Contin-
uous and Partial from Experiment 1 (now labeled Group Subjects and apparatus. The subjects were 32 male rats from the same
Continuous-1 and Group Partial-2, respectively). In addition, two stock and maintained in the same manner as for Experiment 1. The
apparatus was the same as for Experiment 1.
further groups were included that were again conditioned with
Procedure. All subjects were first magazine trained in the same man-
either a continuous or partial reinforcement schedule. For Group ner as Experiment 1. They were then divided into four groups of equal size,
Continuous-2, the CS was paired with two food pellets on each and given 13 sessions of conditioning in each of which there were 20
trial, one delivered 5 s after the start of the CS and another 5 s later. presentations of a 10-s clicker CS. For Groups Partial-2 and Continuous-1
For Group Partial-1, half of the conditioning trials were nonrein- the procedure was the same, respectively, as for Groups Partial and
forced, and for the remainder a single pellet of food was presented Continuous from Experiment 1. For Group Continuous-2, two food pellets
either 5 s after the onset of the clicker or at its termination. The were delivered on every trial in the same manner as the reinforced trials for
design of the conditioning stage for Experiment 2 is shown in Group Partial-2. For Group Partial-1, food was presented on only half the
Figure 3. trials, either 5 s after the onset of the CS or immediately after its termi-
nation. Following the final session of conditioning, the rats received two
sessions of extinction. The first 8 trials of the first extinction session
proceeded as in the previous conditioning session. The remainder of this
session comprised 14 nonreinforced trials with the 10-s clicker, which were
then repeated for the second extinction session. The mean ITI throughout
the experiment, for all groups, was 120 s (range ⫽ 90 –150 s).
Food was always presented 5 s after the start of each trial for Group
Continuous-2, which made it impossible to measure performance through-
out the presentation of the clicker without contamination from responding
brought about by the delivery of this food pellet. To circumvent this
problem, magazine activity during conditioning was measured only during
the first 5 s of the clicker in each group. For extinction, magazine activity
was measured throughout each trial with the clicker. Procedural details that
have been omitted were the same as for Experiment 1.

Results and Discussion

The mean difference scores for the four groups throughout the
13 sessions of conditioning can be seen in the left panel of
Figure 4. There were no statistically significant differences among
Figure 3. Procedure and design of Experiment 2. Time is indicated left to the groups. A three-way ANOVA of individual difference scores
right. Open boxes represent conditioned stimulus presentations. Black with the factors of number of food pellets delivered on a trial,
boxes represent unconditioned stimulus presentations. conditioning schedule, and session revealed no effect of number of
244 HASELGROVE, AYDIN, AND PEARCE

Figure 4. The mean difference scores for Groups Continuous-1, Partial-1, Continuous-2, and Partial-2 across
the 13 sessions of conditioning (left panel) and across the 2 sessions of extinction (right panel) of Experi-
ment 2.

food pellets, F(1, 28) ⫽ 3.33, and no effect of schedule (F ⬍ 1). The mean durations of magazine activity during the 5-s periods
There was however a significant effect of session, F(12, 36) ⫽ prior to the onset of the clicker, for all the extinction trials, were
60.14. The Number of Food Pellets ⫻ Schedule interaction was 0.15 s for Group Continuous-1, 0.18 s for Group Partial-1, 0.16 s
not significant (F ⬍ 1), and neither was the Number of Food for Group Continuous-2, and 0.19 s for Group Partial-2. The
Pellets ⫻ Session interaction, or the Schedule ⫻ Session interac- differences among these means were not significant (Fs ⬍ 1).
tion, Fs(12, 336) ⬍ 1.56. The three-way interaction also fell short The number of two-trial blocks that elapsed before the extinc-
of significance (F ⬍ 1). The mean durations of magazine activity tion criterion was met were 6.50 for Group Continuous-1, 13.88
during the 5-s period prior to the onset of the CS for all condi- for Group Partial-1, 7.00 for Group Continuous-2, and 11.13 for
tioning sessions were 0.60 s for Group Continuous-1, 0.57 s for Group Partial-2. A two-way ANOVA of criterion scores revealed
Group Partial-1, 0.69 s for Group Continuous-2, and 0.69 s for a significant effect of schedule, F(1, 28) ⫽ 25.21, but no effect of
Group Partial-2. A two-way ANOVA of mean durations of mag- number of food pellets (F ⬍ 1), and no Number of Food Pellets ⫻
azine activity revealed no effect of number of reinforcers, no effect Schedule interaction, F(1, 28) ⫽ 2.01.
of schedule, and no interaction, Fs(1, 28) ⬍ 3.35. In keeping with the results from Experiment 1, the present
The group mean difference scores are shown, in two-trial results again show the importance of nonreinforced trials during
blocks, for the two sessions of extinction in the right panel of training for the PREE in Pavlovian conditioning. At the same time,
Figure 4. In both sessions the strength of conditioned responding they lend no support to the explanation for this effect put forward
declined more rapidly in the two continuously reinforced groups by Gallistel and Gibbon (2000). It follows from this theory that
than in either of the two partially reinforced groups. A three-way extinction should have progressed at the same rate in Group
ANOVA of difference scores, in two-trial blocks, revealed no Partial-2 and Continuous-1, whereas the duration of magazine
effect of number of food pellets (F ⬍ 1), but significant effects of activity during the test trials was generally longer in the partially
schedule, F(1, 28) ⫽ 16.65, and trial block, F(13, 364) ⫽ 25.38. reinforced group. The possibility was raised earlier that the differ-
The Number of Food Pellets ⫻ Schedule interaction was not ence between the results for these two groups might have occurred
significant (F ⬍ 1), but the Number of Food Pellets ⫻ Trial Block because the manner in which the US was delivered made it a more
interaction and the Schedule ⫻ Trial Block interactions were both effective reinforcer for Group Partial-2 than for Group
significant, Fs(13, 354) ⬎ 1.78. The three-way interaction was Continuous-1. If this were the case, then extinction should have
also significant, F(13, 364) ⫽ 2.39. Subsequent tests of simple progressed more slowly in Group Continuous-2 than in Group
main effects revealed a significant difference between Groups Partial-1. In fact, the opposite outcome was observed.
Partial-1 and Continuous-1 on Trial Blocks 4 to 7 and 10 to 14,
Fs(1, 392) ⬎ 4.62; and significant differences between Groups Experiment 3
Partial-2 and Continuous-2 on Trial Blocks 3, 4, 5, 7 and 9 to 12,
Fs(1, 392) ⬎ 4.74. On Trial Block 14 there was a significant If the PREE that was demonstrated in Experiment 1 was not a
difference between Groups Partial-1 and Partial-2, F(1, 392) ⫽ consequence of two food pellets presented in close succession
5.77. being a more effective reinforcer than when they are presented
 RATE OF REINFORCEMENT AND THE PREE 245

separately, then it may be possible to explain the results by ing to the theory of Gallistel and Gibbon (2000, 2002), they should
assuming that these different ways of presenting the US had the perform similarly during extinction. In addition, both groups re-
opposite effect. That is, delivering one pellet shortly after the other ceived only one pellet of food on each reinforced trial, and the
may have rendered food a less effective reinforcer than when the trials were ordered and spaced in such a fashion that the delivery
pellets were presented individually. For instance, it is possible that of food for these groups occurred at the same point within a
the effectiveness of the second food pellet to act as a reinforcer session. The two groups were thus matched not only on their rate
was diminished by the presentation of an identical food pellet 5 s of reinforcement during the CS but also on their distribution of
earlier. Carew, Pinsker, and Kandel (1972) and Pfautz (1980) have reinforcement during conditioning. The effectiveness of food as a
both reported that the capacity for a US to evoke an unconditioned reinforcer should therefore be the same for the two groups, which
response can be diminished if it is preceded by the same US. means it would not be reasonable to explain any difference be-
Furthermore, Terry (1976) has shown how unsignaled pretrial US tween them during testing in terms of the explanation that was
exposure renders that US less effective as a reinforcer when it is offered for Experiment 2. On the other hand, if prior experience of
subsequently paired with a CS in Pavlovian conditioning. An a partial reinforcement schedule should weaken the effects of
implication of these findings is that presenting two food pellets extinction, then conditioned responding on the test trials will be
with an interval of 5 s between them might render the second food greater in Group Partial-10 than in Group Continuous-20.
pellet ineffective as a US for Pavlovian conditioning. Once this A third group, Group Continuous-10, that was included in the
possibility is acknowledged, then it follows that the rate at which experiment received the same sequence and duration of trials as
the US was delivered during Group Partial of Experiment 1 was Group Partial-10, but food was presented at the end of every trial.
effectively less than for Group Continuous, and the theory of The rate of reinforcement during the CS for this group was greater
Gallistel and Gibbon (2000) predicts that a PREE should be than for Group Continuous-20, so that according to the theory of
observed. Gallistel and Gibbon (2000), the effects of extinction should be-
The purpose of Experiment 3 was to determine whether it is come evident more rapidly for the group trained with the CS of
possible to demonstrate a PREE in Pavlovian conditioning when shorter duration. Once again, a different prediction is made if the
the rate of delivery of the US during the CS is the same for the principal determinant of the effects of extinction is whether sub-
partial and continuous schedules and when the interval between jects have experienced partial or continuous reinforcement. On this
successive presentations of the US is also the same. To this end, basis, little difference between the groups would be expected. Of
Group Partial-10 received conditioning with a 10-s CS that was course, responding during the test trials should weaken more
followed immediately by food on half the trials. In order to slowly in Group Partial-10 than Group Continuous-10.
reproduce with continuous reinforcement the same rate of food Although the groups received different durations of the CS for
delivery during the CS that was created by this intermittent sched- conditioning, to simplify the interpretation of the results from the
ule, the duration of the CS was extended from 10 to 20 s for Group experiment they all received extinction test trials with a 20-s CS.
Continuous-20. If food was delivered at the end of each of these We chose to use this value because the likely effect of increasing
trials, then it is possible that inhibition of delay (Pavlov, 1927) the duration of the CS for the extinction stage in Group Partial-10
would weaken responding at the start of the CS to a greater extent would be to weaken responding through a generalization decre-
than for Group Partial-10 and would make it difficult to interpret ment (see Haselgrove & Pearce, 2003). Accordingly, if the exper-
any difference between the groups that might be observed during iment should reveal a PREE with these groups, it will be difficult
extinction. Accordingly, for Group Continuous-20, food was pre- to attribute it to the change in duration of the CS in Group
sented 10 s after the start of the CS on half the trials and at the end Partial-10.
of the CS on the remaining trials (see Figure 5). It was hoped that
this schedule would result in similar performance by both groups
during the initial 10 s of each trial. Method
The treatments just described ensured that the two groups re-
Subjects and apparatus. The subjects were 24 male rats from the same
ceived the same rate of reinforcement during the CS and, accord-
stock and maintained in the same manner as for Experiment 1. The
apparatus was the same as for Experiment 1.
Procedure. All subjects were first magazine trained in the same man-
ner as Experiment 1. The subjects were then divided into three groups of
equal size. For Groups Continuous-10 and Partial-10, each of 13 sessions
of conditioning contained 20 presentations of a 10-s clicker separated by an
ITI of 120 s (range ⫽ 90 –150 s). For Group Continuous-10, the termina-
tion of each presentation of the clicker was paired with the delivery of a
single food pellet. For Group Partial-10, the termination of a randomly
selected half of the trials was paired with the delivery of a single food
pellet. For Group Continuous-20, each conditioning session contained 10
presentations of a 20-s clicker. The first trial began at the same point within
the session as that for Group Partial-10. Subsequent trials were separated
by an ITI of 240 s (range ⫽ 180 –300 s). For a randomly selected half of
the trials food was presented 10 s after the start of the clicker, and for the
Figure 5. Procedure and design of Experiment 3. Time is indicated left to remaining trials it was presented at the end of the clicker. The timing of
right. Open boxes represent conditioned stimulus presentations. Black events ensured that each food pellet was delivered at an identical point
boxes represent unconditioned stimulus presentations. within a session for Group Partial-10 and Continous-20. To compare the
246 HASELGROVE, AYDIN, AND PEARCE

performance of the three groups during conditioning, the duration of for Group Partial-20. These differences were not significant
magazine activity was recorded for the first 10 s of each trial in each group. (F ⬍ 1).
The first 8 trials of the first extinction session for Groups Continuous-10 The group mean difference scores are shown, in two-trial
and Partial-10 and the first 4 trials for Group Continuous-20 proceeded as
blocks, for the two sessions of extinction in the right panel of
in the previous conditioning session. The remainder of this session com-
Figure 6. In both extinction sessions the strength of conditioned
prised 14 nonreinforced trials with a 20-s clicker, which were repeated for
the second extinction session. The duration of magazine activity was responding declined more rapidly in the two continuously rein-
recorded for the entire duration of each trial during extinction. The mean forced groups than in Group Partial-10. These observations were
ITI for the trials conducted in extinction for each group was 180 s (range ⫽ supported by a two-way ANOVA of difference scores, in two-trial
120 –240 s). Procedural details that have been omitted were the same as for blocks, which revealed no effect of group, F(2, 21) ⫽ 1.26, but a
Experiment 1. significant effect of trial block, F(13, 273) ⫽ 42.95, and a signif-
icant Group ⫻ Trial Block interaction, F(26, 273) ⫽ 2.05. Tests of
Results and Discussion simple effects revealed differences among the groups on Trial
Blocks 7, 10, and 11, Fs(2, 294) ⬎ 3.35. Tests conducted accord-
The mean difference scores for the three groups for the 13
ing to the Newman–Keuls procedure revealed that on each of these
sessions of conditioning can be seen in the left panel of Figure 6.
trial blocks the mean difference score was smaller in Group
Acquisition progressed most rapidly in Group Continuous-10, at
Continuous-10 than in Group Partial-10. In addition, on Blocks 10
an intermediate rate in Group Continuous-20, and most slowly in
Group Partial-10. These observations were supported by a two- and 11, the mean difference score was smaller in Group
way ANOVA of individual difference scores which revealed no Continuous-20 than in Group Partial-10. The mean durations of
effect of group, F(2, 21) ⫽ 1.71, but a significant effect of session, magazine activity during the 10-s period prior to the onset of the
F(12, 252) ⫽ 51.56, and a significant Group ⫻ Session interaction, clicker for all 14 two-trial blocks were 0.78 s for Group
F(24, 252) ⫽ 1.61. Tests of simple main effects revealed that there Continuous-10, 0.74 s for Group Partial-10, and 0.50 s for Group
were differences among the groups on Sessions 2, 3 and 8, Fs(2, Continuous-20. The differences among these means were not
273) ⬎ 3.36. Subsequent tests conducted according to the significant (Fs ⬍ 1).
Newman–Keuls procedure (Kirk, 1968) revealed that on these The number of two-trial blocks that elapsed before the extinc-
three sessions, the mean difference score was greater in Group tion criterion was met were 8.50 for Group Continuous-10, 12.00
Continuous-10 than in Group Partial-10. On Session 3, the mean for Group Partial-10, and 9.25 for Group Continuous-20. A one-
difference score was also significantly greater for Group way ANOVA of individual criterion scores revealed the difference
Continuous-10 than Group Continuous-20, and on Session 8, the among the groups was significant, F(2, 21) ⫽ 6.52. Newman–
performance of Group Continuous-20 was superior to that of Keuls tests revealed that the mean criterion scores were greater in
Group Partial-10. The fact that performance during the first half of Group Partial-10 than in either Group Continuous-20 or Group
the clicker was superior in Group Continuous-20 to Group Continuous-10.
Partial-10 demonstrates that the steps taken to prevent inhibition of The important results from the experiment concern Groups
delay from weakening responding during the first 10 s of the CS Partial-10 and Continuous-20. Both groups received the same
for Group Continuous-20 were effective. The mean durations of duration of exposure to the CS in each conditioning session, and
magazine activity during the 10-s period prior to the onset of the both groups received the US delivered at the same rate in the
clicker averaged across all the sessions of conditioning were 0.92 s presence of the CS. According to the theory of Gallistel and
for Group Continuous-10, 0.74 s for Group Partial-10, and 0.78 s Gibbon (2000), therefore, there should have been no difference

Figure 6. The mean difference scores for Groups Continuous-10, Partial-10, and Continuous-20 across the 13
sessions of conditioning (left panel) and across the 2 sessions of extinction (right panel) of Experiment 3.
 RATE OF REINFORCEMENT AND THE PREE 247

between these groups during the extinction trials, yet there was allowing the effects of partial reinforcement to be examined in
clear evidence of a PREE. In keeping with the results from the groups that received identical rates of reinforcement during the CS.
previous experiments, the present findings point to the nonrein- Figure 7 shows the design for Experiment 4. For Group Con-
forced trials with the CS during conditioning as being responsible tinuous-10/20, the termination of each presentation of the CS,
for this effect. It might be argued that the slow rate of extinction which had a duration of either 10 or 20 s, was paired with the
that was observed in Group Partial-10, relative to Group delivery of a single food pellet. The duration of the CS for Group
Continuous-20, occurred because of the increase in the duration of Partial-10/5 was either 10 or 5 s, and a randomly selected half of
the CS that took place between the conditioning and extinction these presentations was followed by a single pellet of food. These
stages of the experiment for the first of these groups. Although it values ensure that the rate of reinforcement during the CS was the
is not possible to reject this explanation with complete confidence, same for both groups, and the theory of Gallistel and Gibbon
there are at least two reasons for treating it lightly. First, several (2000, 2002) predicts that they should perform similarly during
experiments have shown that instead of retarding the rate of extinction. To test this prediction, both groups received test trials
extinction, an increase in the duration of the CS has the opposite with a 10-s CS. Because food was always presented at the end of
effect of facilitating extinction (Haselgrove & Pearce, 2003; Polin, the CS for both groups, the type of explanation just considered for
1959). Second, Group Continuous-10 also experienced an increase Experiment 3 is not appropriate for the present study.
in the duration of the CS for the extinction phase, yet there was no
hint that this manipulation strengthened responding relative to Method
Group Continuous-20. Indeed, at the end of the first session there
Subjects and apparatus. The subjects were 16 male rats from the same
was an indication that the change in CS duration may have facil- stock and maintained in the same manner as for Experiment 1. The
itated extinction. apparatus was the same as for Experiment 1.
The duration of the ITI was altered from conditioning to extinc- Procedure. All subjects were magazine trained in the same manner as
tion for the three groups. This interval was increased from 120 s to Experiment 1. They were then divided into two groups of equal size. For
180 s for Group Partial-10 and Continuous-10, and decreased from Group Partial-10/5 each of 15 daily sessions of conditioning contained 8
240 s to 180 s for Group Continuous-20. It is hard to predict how trials with a 5-s clicker and 8 trials with a 10-s clicker. The termination of
these changes might have affected the outcome of the experiment. a randomly selected half of each type of trial was followed by the delivery
Teichner (1952) reported that increasing or decreasing the ITI of a single food pellet. The mean ITI was 240 s (range ⫽ 180 –300 s). For
Group Continuous-10/20 every session contained 4 trials with a 20-s
between conditioning and extinction facilitated the rate at which
clicker and 4 trials with a 10-s clicker. Each trial terminated with the
operant bar pressing extinguished, but we are not aware of a
delivery of a single food pellet. Trial sequences in both groups were
similar effect being reported for Pavlovian conditioning. It is worth randomized with the constraint that no more than two of the same type
noting, however, that there was rather little difference between the could occur in succession. Each trial was separated by an ITI that had a
results from the two continuously reinforced groups, despite the mean of 480 s (range ⫽ 280 –740 s). The ITI sequence was ordered so that
different changes to the ITI that they received. On this basis, the Group Continuous-10/20 and Group Partial-10/5 received food at the same
most reasonable conclusion to draw would be that changing the ITI time within a session. The first of two extinction test sessions commenced
for the extinction phase had little impact on the outcome of the on the day following the final session of conditioning. The first 4 trials for
results for all three groups. Group Partial-10/5 and the first 2 trials for Group Continuous-10/20 of the
first extinction session proceeded as in the previous conditioning session.
The remainder of this session comprised 14 nonreinforced trials with a 10-s
Experiment 4 clicker, which were repeated in the following session. The mean ITI for the
trials conducted in extinction for each group was 360 s (range ⫽ 240 – 480
On the reinforced trials of Experiment 3, food was always s). Procedural details that have been omitted were the same as for Exper-
iment 1.
presented at the end of a 10-s CS for Group Partial-10, whereas it
was presented at the midpoint or the end of a 20-s CS for Group
Continuous-20. The reason for presenting food in the middle of Results and Discussion
some trials for Group Continuous-20 was to prevent the develop- The mean difference scores for the two groups throughout the
ment of inhibition of delay. Although the results suggest this 15 sessions of conditioning can be seen in the left panel of
manipulation was successful, by adopting it we may have paved Figure 8. There is an indication that conditioning was more effec-
the way for our findings to be explained by the theory of Gallistel
and Gibbon (2000). On those trials when food was delivered
before the end of the CS, rats in Group Continuous-20 may have
paid so much attention to the food that they ignored the fact that
the CS was still present. This diversion of their attention would
then result in the CS having an effective duration on these trials of
10 rather than 20 s, and the rate of reinforcement during the CS
would be less for Group Continuous-20 than for Group Partial-10.
Once this possibility is acknowledged then it follows from the
theory of Gallistel and Gibbon (2000) that extinction will progress
more slowly in the partially than in the continuously reinforced Figure 7. Procedure and design of Experiment 4. Time is indicated left to
group. In view of this possible explanation for the results from right. Open boxes represent conditioned stimulus presentations. Black
Experiment 3, another method was adopted in Experiment 4 for boxes represent unconditioned stimulus presentations.
248 HASELGROVE, AYDIN, AND PEARCE

Figure 8. The mean difference scores for Groups Continuous-10/20 and Partial-10/5 across the 15 sessions of
conditioning (left panel) and across the 2 sessions of extinction (right panel) of Experiment 4.

tive with Group Partial-10/5 than Group Continuous-10/20. How- Figure 8. In both extinction sessions the strength of conditioned
ever this observation was not supported by a two-way ANOVA of responding declined more rapidly in Group Continuous-10/20 than
individual difference scores, which revealed a significant effect of in Group Partial-10/5. A two-way ANOVA of individual differ-
session, F(14, 196) ⫽ 8.49, but the effect of group, F(1, 14) ⫽ ence scores revealed a significant effect of group, F(1, 14) ⫽ 6.61,
1.79, and the interaction (F ⬍ 1), were not significant. The mean of trial block, F(13, 182) ⫽ 10.78, and a significant Group ⫻ Trial
durations of magazine activity during the 10-s periods prior to the Block interaction, F(13, 182) ⫽ 3.99. An analysis of simple effects
onset of the clicker, averaged across all the sessions of condition- revealed differences between the groups on Trial Blocks 3 to 7 and
ing, were 0.50 s for Group Continuous-10/20 and 0.74 s for Group 10, Fs(1, 196) ⬎ 7.04. The numerical difference between the
Partial-10/5. The difference between these means was not signif- groups evident on the first two-trial block of extinction was largely
icant, t(14) ⫽ 1.06. It is possible that using intermixed trial a consequence of the change in performance that occurred on Trial
durations of 10 s and 20 s in the Group Continuous-10/20 resulted 2. The mean difference scores on Trial 1 of the extinction stage
in inhibition of delay that would weaken responding at the start of was 5.01 s for Group Continuous-10/20 and 5.33 s for Group
the CS and perhaps facilitate extinction with the 10-s CS in the test Partial-10/5. This difference between the groups was not signifi-
phase. Inspection of the left panel of Figure 8 shows that by the cant, t(14) ⫽ 0.27. The mean number of two-trial blocks that
end of conditioning the performance of the two groups was very elapsed before the extinction criterion was met were 8.25 in Group
similar, which reduces concerns about the influence of inhibition Partial-10/5 and 3.00 in Group Continuous-10/20. This difference
of delay on the outcome of the experiment. Furthermore, the mean between the groups was significant, t(14) ⫽ 3.30. The mean
difference scores were calculated for the first 5 s of the clicker on duration of magazine activity for the pre-CS periods averaged
the 10-s trials during the final session of conditioning for Group across all the extinction trials was 0.25 s for Group Continuous-
Continuous-10/20 and Group Partial-10/5. These difference scores 10/20 and 0.24 s for Group Partial-10/5. This difference was not
were 2.30 s and 1.93 s for Group Continuous-10/20 and Group significant, t(14) ⫽ 0.10.
Partial-10/5, respectively. The difference between these means In keeping with the results from the previous three experiments,
was not significant, t(14) ⫽ 0.96. Nonetheless, to evaluate the extinction progressed more rapidly in a group that received a
potential influence of inhibition of delay on the course of extinc- continuous rather than a partial reinforcement schedule during
tion further, the latency to enter the magazine after the start of each conditioning. The occurrence of this effect is difficult to explain
trial during the first two and final two sessions of conditioning with the theory of Gallistel and Gibbon (2000) because the rate at
were recorded for both groups. The mean latencies for the first two which the US was presented during the CS was the same for both
sessions were 2.97 s for Group Continuous-10/20 and 3.16 s for groups.
Group Partial-10/5. The mean latencies for the final two sessions
were 1.98 s for Group Continuous-10/20 and 1.88 s for Group General Discussion
Partial-10/5. A two-way ANOVA of individual latencies revealed
a significant effect of session, F(1, 14) ⫽ 10.83, but no effect of Four experiments compared the rates of extinction in groups that
group and no interaction (Fs ⬍ 1). That the mean latency to enter had previously received either a partial or a continuous schedule of
the magazine reduced for both groups between the start and end of reinforcement in Pavlovian conditioning. Three different tech-
conditioning suggests that inhibition of delay did not develop niques were used to equate, between groups, the rate of reinforce-
within the CS for either group. ment during conditioning. Extinction progressed more rapidly in
The group mean difference scores are shown, in two-trial every experiment when conditioning had previously been con-
blocks, for the two sessions of extinction in the right panel of ducted with a continuous rather than a partial reinforcement sched-
 RATE OF REINFORCEMENT AND THE PREE 249

ule. These results are problematic for any account of the PREE that served the opposite of a PREE because the difference between the
attributes the effect to differences in the rate of reinforcement associative strengths of the two stimuli was the dominant factor
(Gallistel & Gibbon, 2000, 2002). that determined the course of extinction, whereas in the study by
The results also pose a problem for the majority of associative Rescorla (1999b), it may have been the associability of the CS that
theories of Pavlovian conditioning, because partial reinforcement was principally responsible for the outcome of the experiment.
should result in weaker associative strength, and hence more rapid In several of the experiments reported here, a continuous rein-
extinction, than continuous reinforcement. Given this shortcom- forcement schedule was used in which food was presented either at
ing, Pearce et al. (1997) proposed that the PREE in Pavlovian the midpoint or at the end of the CS. It is conceivable that animals
conditioning might be explained in the same manner that has would come to anticipate the delivery of food at both points during
proved relatively successful for instrumental conditioning. That is, the CS, so that on each trial there was an occasion when an
the nonreinforced trials during conditioning with a partial rein- anticipated US was omitted. If the absence of the US were to
forcement schedule might leave an aftereffect that persists to the generate the aftereffects that are assumed to be produced by the
next trial. If this trial is reinforced, then the aftereffects of nonre- nonreinforced trials of a partial reinforcement schedule, then a
inforcement will serve as a component of the context for the PREE might also be expected with this schedule of reinforcement.
reinforced trial, and thereby serve as a cue for conditioned re- The experiments did not include the necessary control groups to
sponding. The occurrence of this cue during extinction will then permit a satisfactory evaluation of this prediction, but the results
encourage responding to a greater extent by subjects trained with from Experiment 1 suggest it is unlikely to be confirmed. In that
a partial rather than a continuous reinforcement schedule. experiment, both groups received a reinforcement schedule that
An obvious implication of this explanation is that a PREE resulted in food being unexpectedly omitted at the midpoint and
should not be observed in Pavlovian conditioning when the effects the end of a 10-s CS. However, responding weakened during
of the different reinforcement schedules are compared using the extinction considerably more slowly in the group that received
same subjects. Provided that the trials are intermixed, conditioning nonreinforced trials during conditioning. On the basis of this
with the stimuli that are consistently and intermittently paired with finding it appears that it is the aftereffects of a nonreinforced trial
food will take place in the context of the aftereffects of previously that are critical for a demonstration of the PREE, not the unex-
nonreinforced trials. The occurrence of this context during extinc- pected omission of a US within a reinforced trial.
tion will then promote responding to the same degree for both In contrast to the complexity of the associative analysis, the
stimuli, and the principal determinant of the rate of extinction will account of extinction and the PREE offered by Gallistel and
be their associative strengths. In support of this analysis, Pearce et Gibbon (2000) is attractively simple. According to Gallistel and
al. (1997) have reported a reversed PREE in Pavlovian appetitive Gibbon, responding ceases during extinction when the animal
conditioning with rats using a within-subject methodology (see makes a decision to stop responding. This decision is determined
also Bouton & Sunsay, 2001). by the ratio of (a) the duration of exposure to the CS since the last
At least one theory of associative learning is capable of provid- reinforced trial to (b) the average duration of exposure to the CS
ing an account of the PREE without appealing to the influence of between successive presentations of the US during conditioning. In
aftereffects of nonreinforced trials. Mackintosh (1975) proposed the current experiments, the value of this ratio was held constant
that the associability of a CS (that is, the ease with which its between different groups of rats. Nevertheless, a reliable PREE
associative strength can be changed) is governed by its history of was still observed. These results complement those of Haselgrove
reinforcement. Of particular relevance to the present discussion is and Pearce (2003) who showed that extinction was facilitated by
the prediction from the theory that the associability of a CS will be an increase or a decrease in the duration of the CS following
greater after it has been consistently rather than intermittently conditioning, even though the value of this ratio was again held
paired with a US. If this is correct, then the effects of extinction constant between groups. Taken together, the results from both
would be expected to be more pronounced after conditioning with sets of experiments suggest that the theory of Gallistel and Gibbon
a continuous than a partial reinforcement schedule. Support for this (2000) falls short of providing a satisfactory account of the effects
explanation of the PREE can be found in an experiment on of nonreinforcement in Pavlovian conditioning. We suggest that
autoshaping in pigeons by Rescorla (1999a), which demonstrated the effects of nonreinforcement are best understood from an asso-
that the changes in associative strength during extinction are less ciative perspective (e.g., Mackintosh, 1975; Pearce, 1987, 1994;
after partial than continuous reinforcement. A problem with this Rescorla & Wagner, 1972) that takes into account the effects that
explanation for the PREE, is that it fails to explain the reverse the omission of an expected US will have on the context in which
effect which has been reported when the same group of subjects it is presented (Pearce et al., 1997).
are trained with both reinforcement schedules (Pearce et al., 1997).
Perhaps, therefore, at least three different processes influence References
extinction after partial and continuous reinforcement: the associa-
tive strength of the CS, its associability, and the presence of the Amsel, A. (1958). The role of frustrative nonreward in noncontinuous
aftereffects of nonreinforcement. Given that the influence of these reward situations. Psychological Bulletin, 55, 102–119.
Amsel, A. (1992). Frustration theory. Cambridge, England: Cambridge
processes is likely to vary in different tasks, it should not be
University Press.
surprising if experimental tests for the PREE occasionally reveal Bouton, M. E., & Sunsay, C. (2001). Contextual control of appetitive
conflicting outcomes. Thus in contrast to the reversed PREE that conditioning: Influence of a contextual stimulus generated by a partial
was reported by Pearce et al. (1997) using a within-subject meth- reinforcement procedure. Quarterly Journal of Experimental Psychol-
odology, Rescorla (1999b) has demonstrated a conventional PREE ogy: Comparative and Physiological Psychology, 54B, 109 –125.
using pigeons and autoshaping. Perhaps Pearce et al. (1997) ob- Capaldi, E. J. (1967). A sequential theory of instrumental training. In K. W.
250 HASELGROVE, AYDIN, AND PEARCE

Spence & J. T. Spence (Eds.), The psychology of learning and motiva- Pearce, J. M. (1987). A model for stimulus generalization in Pavlovian
tion (pp. 67–156). New York: Academic Press. conditioning. Psychological Review, 94, 61–73.
Carew, T. J., Pinsker, H. M., & Kandel, E. R. (1972, January 28). Long- Pearce, J. M. (1994). Similarity and discrimination: A selective review and
term habituation of a defensive withdrawal reflex in aplysia. Science, a connectionist model. Psychological Review, 94, 587– 607.
175, 451– 454. Pearce, J. M., Redhead, E. S., & Aydin, A. (1997). Partial reinforcement in
Fitzgerald, R. D. (1963). Effects of partial reinforcement with acid on the appetitive Pavlovian conditioning with rats. Quarterly Journal of Ex-
classically conditioned salivary response in dogs. Journal of Compara- perimental Psychology, 50B, 274 –294.
tive and Physiological Psychology, 56, 1056 –1060. Pfautz, P. L. (1980). Unconditioned facilitation and diminution of the
Fitzgerald, R. D., Vardaris, R. M., & Teyler, T. J. (1966). Effects of partial unconditioned response. Unpublished doctoral thesis, Yale University.
reinforcement followed by continuous reinforcement on classically con- Polin, A. T. (1959). The effects of flooding and physical suppression as
ditioned heart-rate in the dog. Journal of Comparative and Physiological extinction techniques on an anxiety motivated avoidance locomotor
Psychology, 62, 483– 486. response. Journal of Psychology, 47, 235–245.
Gallistel, C. R., & Gibbon, J. (2000). Time, rate, and conditioning. Psy- Rescorla, R. A. (1999a). Partial reinforcement reduces the associative
chological Review, 107, 289 –344. change produced by nonreinforcement. Journal of Experimental Psy-
Gallistel, C. R., & Gibbon, J. (2002). The symbolic foundations of condi- chology: Animal Behavior Processes, 25, 403– 414.
tioned behavior. Mahwah, NJ: Erlbaum. Rescorla, R. A. (1999b). Within-subject partial reinforcement extinction
Gibbon, J., Farrell, L., Locurto, C. M., Duncan, H. J., & Terrace, H. S. effect in autoshaping. Quarterly Journal of Experimental Psychology,
(1980). Partial reinforcement in autoshaping with pigeons. Animal 52B, 75– 87.
Learning and Behavior, 8, 45–59. Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian condi-
Gibbs, C. M., Latham, S. B., & Gormezano, I. (1978). Classical condi- tioning: Variations in the effectiveness of reinforcement and nonrein-
tioning of the rabbit nictitating membrane response: Effects of reinforce- forcement. In A. H. Black & W. F. Prokasy (Eds.), Classical condition-
ment schedule on response maintenance and resistance to extinction. ing II (pp. 64 –99). New York: Appleton-Century-Crofts.
Animal Learning and Behavior, 6, 209 –215. Slivka, R. M., & Bitterman, M. E. (1966). Classical appetitive conditioning
Guthrie, E. R. (1935). The psychology of learning. New York: Harper. in the pigeon: Partial reinforcement. Psychonomic Science, 4, 181–182.
Haselgrove, M., & Pearce, J. M. (2003). Facilitation of extinction by an Teichner, W. H. (1952). Experimental extinction as function of the inter-
increase or a decrease in trial duration. Journal of Experimental Psy- trial intervals during conditioning and extinction. Journal of Experimen-
chology: Animal Behavior Processes, 29, 153–166. tal Psychology, 44, 170 –178.
Hull, C. L. (1943). Principles of behaviour. New York: Appleton-Century- Terry, W. S. (1976). Effects of priming unconditioned stimulus represen-
Crofts. tation in short-term memory on Pavlovian conditioning. Journal of
Kimble, G. A. (1961). Hilgard and Marquis’ conditioning and learning. Experimental Psychology: Animal Behavior Processes, 2, 354 –369.
New York: Appleton-Century-Crofts. Thorndike, E. L. (1911). Animal intelligence. New York: Macmillan.
Kirk, R. E. (1968). Experimental design: Procedures for the behavioural Wagner, A. R., Siegel, S., Thomas, E., & Ellison, G. D. (1964). Reinforce-
sciences. Belmont, CA: Brooks/Cole. ment history and the extinction of a conditioned salivary response.
Mackintosh, N. J. (1974). The psychology of animal learning. London: Journal of Comparative and Physiological Psychology, 58, 354 –358.
Academic Press.
Mackintosh, N. J. (1975). A theory of attention: Variations in the associa-
bility of stimuli with reinforcement. Psychological Review, 82, 276 – Received December 22, 2003
298. Revision received March 12, 2004
Pavlov, I. P. (1927). Conditioned reflexes. New York: Dover. Accepted March 12, 2004 䡲