ArticlePDF Available

Abstract and Figures

Extinction is a very relevant learning phenomenon from a theoretical and applied point of view. One of its most relevant features is that relapse phenomena often take place once the extinction training has been completed. Accordingly, as extinction-based therapies constitute the most widespread empirically validated treatment of anxiety disorders, one of their most important limitations is this potential relapse. We provide the first demonstration of relapse reduction in human contingency learning using mild aversive stimuli. This effect was found after partial extinction (i.e., reinforced trials were occasionally experienced during extinction, Experiment 1) and progressive extinction treatments (Experiment 3), and it was not only because of differences in uncertainty levels between the partial and a standard extinction group (Experiment 2). The theoretical explanation of these results, the potential uses of this strategy in applied situations, and its current limitations are discussed. (PsycINFO Database Record
Content may be subject to copyright.
Journal of Experimental Psychology: Learning,
Memory, and Cognition
Slower Reacquisition After Partial Extinction in Human
Contingency Learning
Joaquín Morís, Itxaso Barberia, Miguel A. Vadillo, Ainhoa Andrades, and Francisco J. López
Online First Publication, June 23, 2016. http://dx.doi.org/10.1037/xlm0000282
CITATION
Morís, J., Barberia, I., Vadillo, M. A., Andrades, A., & López, F. J. (2016, June 23). Slower
Reacquisition After Partial Extinction in Human Contingency Learning. Journal of Experimental
Psychology: Learning, Memory, and Cognition. Advance online publication. http://
dx.doi.org/10.1037/xlm0000282
Slower Reacquisition After Partial Extinction in Human
Contingency Learning
Joaquín Morís
Universidad de Oviedo and Institut d’Investigació Biomèdica de
Bellvitge (IDIBELL)
Itxaso Barberia
Universitat de Barcelona
Miguel A. Vadillo
King’s College London and University College London Ainhoa Andrades and Francisco J. López
Instituto de Investigación Biomédica de Málaga (IBIMA) and
Universidad de Málaga
Extinction is a very relevant learning phenomenon from a theoretical and applied point of view. One of
its most relevant features is that relapse phenomena often take place once the extinction training has been
completed. Accordingly, as extinction-based therapies constitute the most widespread empirically vali-
dated treatment of anxiety disorders, one of their most important limitations is this potential relapse. We
provide the first demonstration of relapse reduction in human contingency learning using mild aversive
stimuli. This effect was found after partial extinction (i.e., reinforced trials were occasionally experienced
during extinction, Experiment 1) and progressive extinction treatments (Experiment 3), and it was not
only because of differences in uncertainty levels between the partial and a standard extinction group
(Experiment 2). The theoretical explanation of these results, the potential uses of this strategy in applied
situations, and its current limitations are discussed.
Keywords: extinction, contingency learning, relapse, aversive learning
Learning relationships between different events relevant to or-
ganisms is ubiquitous, and takes place almost constantly (Clark,
2013; Friston, 2003). Most of the time, this acquired knowledge is
helpful, as it is the basis of useful predictions about future events
or inferences about their relationships. However, it can also be
maladaptive. For example, in many anxiety disorders, seemingly
harmless cues may become associated to strong and disruptive
emotional responses, leading to distress and daily problems for
those suffering them (Beckers, Krypotos, Boddez, Effting, &
Kindt, 2013; Mineka & Zinbarg, 2006).
Fortunately, this type of learning is flexible in the sense that,
once it has been acquired, it may be modified or altered. One of the
ways in which this modification can take place is through extinc-
tion, the repeated presentation of a given cue in the absence of the
consequences with which it had been previously associated (Pav-
lov, 1927). A progressive reduction in the acquired response to this
cue is observed after experiencing a series of presentations of the
cue alone. This flexibility allows the organism not only to adapt to
changes in the cue-outcome relationships as they occur in the
environment but also to ease those maladaptive forms of learning
that lead to disruptive emotional responses. For example, in
cognitive and behavioral therapies of anxiety disorders, expo-
sure therapy—a repeated, systematic, and controlled exposure
to the feared cue, in the absence of any traumatic event—seems
to be a key component to success (Bouton, 2002; Longmore &
Worrell, 2007; Norton & Price, 2007). Unfortunately, extinc-
tion is not such a robust effect as acquisition itself, and several
factors can lead to a recovery of the original fear response, or
relapse. Thus, the current challenge is not so much to achieve
the fear reduction but to prevent its relapse (Vervliet, Craske, &
Hermans, 2013). The return of fear was first documented by
Rachman (1966) in relation to the reemergence of fear follow-
ing systematic desensitization. In the existing literature, esti-
mates of return of fear after exposure therapy range from 19 to
62% according to the review by Craske and Mystkowski (2006).
Thus, relapse may be considered as a serious difficulty for
cognitive–behavioral therapy of anxiety disorders. In the pres-
ent series of experiments we will focus on how certain forms of
relapse can be attenuated in the case of a human contingency
Joaquín Morís, Departamento de Psicología, Universidad de Oviedo, and
Institut d’Investigació Biomèdica de Bellvitge (IDIBELL); Itxaso Barberia,
Departamento de Psicología Básica, Universitat de Barcelona; Miguel A.
Vadillo, Department of Primary Care and Public Health Sciences, King’s
College London, and Department of Experimental Psychology, University
College London; Ainhoa Andrades and Francisco J. López, Instituto de
Investigación Biomédica de Málaga (IBIMA) and Departamento de Psi-
cología Básica, Universidad de Málaga.
This work was supported by grants P11-SEJ-7898 from Junta de Anda-
lucía and PSI2011-24662 from the Spanish Ministry of Science and Inno-
vation, UMA FC14-SEJ-332014 from University of Málaga, and Grant
PSI2013-43516-R from the Spanish Ministry of Economy and Competi-
tiveness. Raw data and analysis scripts are available at the Open Science
Framework (http://osf.io/thrc6).
Correspondence concerning this article should be addressed to Fran-
cisco J. López, Departamento de Psicología Básica, Facultad de Psi-
cología, Universidad de Málaga, Campus de Teatinos, Málaga 29017,
Spain. E-mail: frjlopez@uma.es
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Experimental Psychology:
Learning, Memory, and Cognition © 2016 American Psychological Association
2016, Vol. 42, No. 5, 000 0278-7393/16/$12.00 http://dx.doi.org/10.1037/xlm0000282
1
learning paradigm (see Shanks, 2010, for a review of the
contingency learning literature).
There are various sources of relapse that may take place after
behavioral extinction (Bouton, 2002, 2004, or Vervliet et al., 2013
for reviews), affecting both responses acquired in the laboratory
and fear responses suffered by patients with anxiety disorders. One
of these forms of relapse described in the literature is rapid
reacquisition (Ricker & Bouton, 1996). Rapid reacquisition refers
to the effect that, after extinction, if the cue is again paired with the
outcome with which it was previously associated, this new learn-
ing is faster than the original learning, indicating a carry-over
effect of the original learning.
Interesting to the authors, some other forms of relapse such as
renewal, reinstatement or spontaneous recovery have been evi-
denced in animal as well as human conditioning studies, using
different experimental paradigms, and also after exposure therapy
in clinical settings (Craske, Liao, Brown, & Vervliet, 2012). Re-
lapse phenomena, then, have been regarded as providing a useful
insight into the nature of extinction and its underlying mecha-
nisms. For a start, all the effects show that responding to a
seemingly extinguished cue can reappear under certain conditions,
suggesting that extinction does not erase previous learning. In-
stead, it seems that extinction implies a new learning that may be
expressed or not depending on additional factors. Understanding
both extinction and relapse has been the focus of interest of much
research as it would provide essential information to understand
extinction mechanisms as well as sound basis to improve exposure
based therapies (see Milad & Quirk, 2012 or Vervliet et al., 2013
for recent reviews).
A fruitful research heuristic is that provided by theoretical
accounts of extinction and its related phenomena. Currently, the
model proposed by Mark Bouton (Bouton, 1993, 2002) is the most
widely accepted explanation of these effects. According to Bou-
ton’s model, the extinction treatment produces a new learning,
separate from the original one. A new inhibitory association be-
tween the conditioned cue and the outcome is created during
extinction, while the original excitatory association stays mostly
intact. Furthermore, contextual cues control the expression of this
inhibitory association. The more similar the contextual cues at the
time of retrieval are to those present during extinction, the more
activation of the inhibitory association, producing a higher reduc-
tion of responses to the cue. Following these two ideas, extinction
and its related phenomena of relapse can be easily explained
(Bouton, 2002; Vervliet et al., 2013 for a detailed review).
Bouton et al. (Bouton, Woods, & Pineño, 2004; Woods &
Bouton, 2007) found in animal classical and operant conditioning
that the rate of reacquisition of a previously extinguished response
slowed down when some cue-outcome pairings were included as
part of the extinction treatment. Specifically, Bouton et al. (2004)
showed, in a series of appetitive classical conditioning experi-
ments, that rats were slower to reacquire a previously extinguished
response when the extinction treatment included occasional rein-
forced trials (i.e., trials in which the cue was followed by an
appetitive outcome, food) intermixed among more frequent non-
reinforced trials than when extinction only included nonreinforced
trials. Woods and Bouton (2007) showed a similar effect in operant
conditioning. Rats were slower to reacquire an operant response
after an extinction treatment that included intermixed occasional
reinforced responses than after a standard extinction treatment
including only nonreinforced responses. These results are consis-
tent with the idea that rapid reacquisition may be alleviated in a
seemingly paradoxical way namely, intermixing cue alone trials
with some trials that also include the outcome. According to
Bouton’s model this should occur because in this case reinforced
trials, as one of the features present during the acquisition and the
extinction phase, become part of both contexts. Later, when the
reinforced trials appear during reacquisition, their presence should
help activate not only the excitatory context of acquisition but also
the inhibitory association formed during extinction, causing a
slower reacquisition than in a standard extinction condition where
only the acquisition context is retrieved. Consistently with Capal-
di’s (1967, 1994) sequential learning theory, Ricker and Bouton
(1996) suggested that reacquisition is controlled by a “trial-
signaling” mechanism, whereby the animal learns that a certain
type of trial, either a reinforced or a nonreinforced trial, reliably
signals the type of upcoming trial. As a result, during acquisition
animals not only learn that the cue is paired by the reinforcer, but
also that reinforced trials follow other reinforced trials, whereas
during standard extinction animals learnt that nonreinforced trials
follow other nonreinforced trials. This way, the type of preceding
trial (reinforced or nonreinforced) becomes part of the context that
can control responding. Then, including reinforced trials during
extinction training, might allow reinforced trials to become asso-
ciated not only with the acquisition but also with the extinction
context. Thus, during reacquisition training, reinforced trials
should have also the ability to retrieve the memory of nonrein-
forced trials (i.e., the inhibitory association learned during extinc-
tion training).
These results with nonhuman animals provide an important
starting point, as they show a potential way to reduce relapse and
increase the effectiveness of extinction-based treatments. How-
ever, no previous evidence of this effect with aversive stimuli has
been described in humans to the best of our knowledge.
Thus, the main objective of the present experimental series is to
evaluate whether the rate of reacquisition of a previously learned
cue-outcome relationship after an extinction treatment varies de-
pending on the inclusion of occasional reinforced trials during
extinction (partial extinction). This evaluation was made with
human participants in a contingency learning task including an
aversive outcome (see Method for details). From a theoretical
point of view, changes in the rate of reacquisition because of this
intermixed or partial extinction paradigm would be consistent with
Bouton’s model predictions concerning the contextual control of
the memory stored during extinction. In other words, our objective
is to show a potentiated extinction effect as measured by differ-
ences in the reacquisition rate after intermixing reinforced trials
during extinction and a standard extinction preparation without
any reinforced trial. In Experiment 1, the effect of partial extinc-
tion was evaluated in a deterministic task (all cue-outcome rela-
tionships programmed were deterministic except the for the target
partial extinction treatment) whereas in Experiment 2 the effect
was evaluated in a probabilistic task (all of the cue-outcome
relationships programmed were probabilistic). This way, in Exper-
iment 2, our target evaluation was made in a situation in which
both, partial extinction and standard extinction included probabi-
listic relationships. Experiment 3 tested if a gradual reduction in
the probability of reinforcement during extinction could also pro-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
duce slower reacquisition while having similar levels of respond-
ing at the end of the extinction phase.
Experiment 1
Method
Participants and apparatus. In total, 49 undergraduate Psy-
chology students from University of Málaga completed the task.
They received course credits for their participation. Participants
conducted the experiment in small groups using a computer room
with individual cubicles. Visual stimuli were presented on a 17-
inch monitor with resolution set to 1,024 768 pixels. Auditory
stimuli were presented using individual headphones. The experi-
mental program was written in E-Prime 2.0 (Psychology Software
Tools, Pittsburgh, PA).
Design and procedure. The design of the experiment is sum-
marized in Table 1. During Phase 1, both groups of participants
were exposed to 18 pairings of Cue A with the auditory tone and
18 trials in which Cue B was presented alone. The trials were
presented in a pseudorandom order. Out of each group of six trials,
there were three of each type, with their order randomly selected.
This pseudorandom order was used across the whole experiment.
The experimental manipulation took place during Phase 2. Par-
ticipants in the Intermixed group were exposed to a sequence of
54 trials in which Cue A was followed by the outcome on 22.2%
of the times. These trials were intermixed with an equal number of
trials where Cue B was presented alone. To ensure that the tran-
sition between Phase 1 and Phase 2, and between Phase 2 and
Phase 3 were similar across groups, there was a restriction in the
first two and last two trials of Cue A, which could not be followed
by the outcome. Participants in the Continuous group were ex-
posed to an identical sequence of trials, with the only exception
that Cue A was never followed by the outcome. In the third and
final phase, all participants were exposed to a sequence of 18 trials
with Cue A always followed by the outcome and 18 trials of Cue
B presented on its own.
A blue square with size 359 359 pixels and RGB values of 85,
142, and 213 and a red circle with diameter 359 pixels and RGB
values of 192, 80, and 77 were used as Cues A and B, counter-
balanced across participants. All visual stimuli were presented
against a light gray background (RGB values of 128, 128, and 128)
and centered on the screen. The outcome was a 500 ms white noise
at 95 dB (approximately), presented through a set of headphones.
The experiment began with the following instructions, translated
from Spanish:
Thank you for participating in this experiment! The task you are about
to perform is very simple. During the experiment you will see a series
of figures. These figures will sometimes be followed by a noise. Your
goal is to predict on each occasion whether or not the noise will be
presented. On each trial, you will first see the figure and, below, you
will see a rating scale to enter your predictions. To make your
prediction, click on any point of the scale. Click on values close to the
right extreme (close to the 100 label) if you think that the noise is very
likely to follow. If you think that the noise is unlikely to be presented,
click on values close to the left extreme (close to zero). When you
click on any point of the rating scale, you will see a marker and a label
with the specific value you have chosen. If you want to change this
value, just click again on a different place of the scale. Once you have
chosen the desired value, you can press ENTER to see what happens
and check whether the noise is presented or not. Pay attention to the
figures and to the noise to learn how to make good predictions. If you
have any doubt now or while you are conducting the task, ask the
experimenter for help and make sure that you understand everything.
Good luck!
Each trial began with a black fixation cross (Courier New font,
size 180) presented on the center of the screen for 2,500 to 3,500
ms (uniform distribution with 100 ms step). Afterward, the cue
was presented together with a rating scale on the lower part of the
screen where participants had to click to enter their response follow-
ing the procedure described in the instructions. The cue remained on
screen for 3,000 ms after participants had responded. The outcome,
when present, overlapped with the time period of 2,000 to 2,500 ms
after the onset of the cue. At the end of the trial, all visual and auditory
stimuli were removed and the fixation cross of the next trial was
presented.
Results and Discussion
Preanalysis treatment of data. Before the analysis, we re-
moved the data from those participants whose mean score of the
three last trials of Phase 1 for either Cue A or Cue B was more than
2SDs away from the group mean (Filters 1 and 2). The same
selection criterion was applied to the three last trials of Cue B in
Phase 2 (Filter 3). These filters were chosen because they ensured
an adequate level of similar performance between participants in
situations nonrelated to the effect of interest (Filter 3) and previous
to the experimental manipulation (Filters 1 and 2). Four partici-
pants did not meet the filter criteria described and were removed
from further analyses. The final sample consisted of 45 partici-
pants (20 in the Intermixed group, 25 in the Continuous group).
Statistical analyses. All the statistical analyses were per-
formed using IBM SPSS version 21 (IBM Corp., 2012). In all the
repeated measures analysis of variances (ANOVAs) the sphericity
was tested and the degrees of freedom were corrected using
Greenhouse-Geisser’s epsilon whenever necessary. The curve fit-
ting procedures were carried out using Matlab version 2012b and
its Curve Fitting Toolbox (The MathWorks, Inc., 2012). The effect
size statistics reported are partial eta squared (p
2) for the ANOVAs
and Cohen’s din the case of the ttests (Cohen, 1988).
Acquisition phase. The top panel of Figure 1 depicts the
results for the Acquisition phase. A mixed ANOVA was used to
analyze these data, which included a between-subjects factor
(Group: Intermixed vs. Continuous) and two within-subjects fac-
tors (Cue: A vs. B, and Trial: 1 to 18). The results showed a
significant effect of Cue, Trial, and Cue Trial interaction (re-
Table 1
Design Summary of Experiment 1
Group Phase 1 Phase 2 Phase 3
Intermixed 18 A (100%) 54 A (22.2%) 18 A (100%)
18 B (0%) 54 B (0%) 18 B (0%)
Continuous 18 A (100%) 54 A (0%) 18 A (100%)
18 B (0%) 54 B (0%) 18 B (0%)
Note. The percentage of trials in which Cues A and B were paired with
the outcome is indicated between brackets.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
spectively, F(1, 43) 2314.37, p.001, p
2.981, F(6.74,
289.98) 2.41, p.022, p
2.053, and F(5.23, 224.98)
43.76, p.001, p
2.504). We observed no significant effect of
the Group factor during acquisition, as there was no main effect of
Group (F(1, 43) 0.6, p.442, p
2.014), nor Group Cue
(F(1, 43) 0.3, p.582, p
2.007), Group Trial (F(6.74,
289.98) 1.83, p.083, p
2.04), or Group Cue Trial
interactions (F(5.23, 224.98) 0.41, p.847, p
2.009).
Extinction phase. A similar ANOVA was performed with the
data of the Extinction phase, shown in the middle panel of Figure
1. It showed a significant effect of Group (F(1, 43) 39.88, p
.001, p
2.481), Cue (F(1, 43) 96.19, p.001, p
2.691),
Trial (F(11.27, 484.98) 20.55, p.001, p
2.323), Group
Cue (F(1, 43) 33.13, p.001, p
2.435), Cue Trial (F(9.94,
427.69) 9.32, p.001, p
2.178), and Group Cue Trial
interactions (F(9.94, 427.69) 2.52, p.005, p
2.055). The
Group Trial contrast was not significant (F(11.27, 484.98)
1.57, p.101, p
2.035). As can be seen in Figure 1, this was
because of the fact that responses to cue A differed between
Intermixed and Continuous groups, with that difference varying
across trials. This was confirmed with an additional ANOVA with
two factors, Group (Intermixed vs. Continuous) and Trial (1 to 18)
0
20
40
60
80
100
123456789101112131415161718
tneme
gduJ evitc
iderP naeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
0
20
40
60
80
100
1 4 7 101316192225283134374043464952
tn
em
egduJ evitciderP naeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
0
20
40
60
80
100
123456789101112131415161718
tnemegduJ evitciderP nae
M
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
Figure 1. Mean predictive judgments of each of the three phases in Experiment 1. Acquisition, Extinction, and
Reacquisition are presented in the top, middle, and bottom panels, respectively. The error bars are the SEM.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
of the responses to Cue A. There was a main significant effect of
the two factors (Group, [F(1, 49) 27.9, p.001, p
2.363],
Trial, [F(11.86, 581.38) 16.26, p.001, p
2.249]), and a
significant interaction between them (Group Trial, [F(11.86,
581.38) 2.36, p.006, p
2.046]). In the case of the responses
to Cue B, there was no significant effect of Group, Trial, or
Group Trial interaction (Fs1.37, ps.247).
Reacquisition phase. Data from the Reacquisition phase are
shown in the bottom panel of Figure 1. A three-way ANOVA was
used using the same factors as in the case of the Acquisition phase
analysis. There was no main effect of Group (F(1, 43) 3.07, p
.087, p
2.067), but there were significant effects of Cue (F(1,
43) 974.71, p.001, p
2.957) and Trial factors (F(5.77,
248.5) 39.97, p.001, p
2.481), and significant Group
Trial (F(5.77, 248.5) 7.01, p.001, p
2.14), Cue Trial
(F(6, 258.17) 58.98, p.001, p
2.578), and Group Cue
Trial interactions (F(6, 258.17) 6.15, p.001, p
2.125). The
Group Cue (F(1, 43) 2.54, p.118, p
2.055) interaction
was not significant.
Given the results observed and our previous hypothesis, we
tested the possibility that the reacquisition of Cue A could have
differed between groups. To do so, we analyzed only responses to
Cue A in both groups, using an ANOVA with only the Group and
Trial factors. It showed that there was no main effect of Group
(F(1, 43) 2.86, p.098, p
2.062) but there was a main effect
of Trial (F(5.35, 229.88) 60.83, p.001, p
2.586). Crucially
for our hypothesis, there was a significant interaction between
these two factors, Group Trial (F(5.34, 229.87) 8.09, p
.001, p
2.158). As expected, this pattern of results did not
happen in an equivalent analysis for Cue B, where there were no
significant effects of Group nor a Group Trial interaction (Fs
1.27, ps.283).
Post hoc ttests for independent samples were run on the re-
sponses to Cue A on each trial, finding statistical differences
between groups in Trial 1 (t(28.25) 3.99, p.001, d
s
1.27),
Trial 2 (t(43) 3.06, p.004, d
s
0.92), Trial 3 (t(43) 3.64,
p.001, d
s
1.09), and Trial 4 (t(43) 2.75, p.009, d
s
0.765).
Curve fitting. Given the negatively accelerated shape of the
reacquisition learning curve, it is also possible to carry out a model
fitting analysis. We can try to estimate a curve that describes the
behavior of each group as a function of the number of trials
elapsed. As 100 was the maximum value of participants’ re-
sponses, a very simple curve equation would be the following one:
y100
100
axb
(1)
In this equation, ywould be the response observed to Cue A and
xthe number of the current trial. There are two free parameters a
and brepresenting, respectively, the initial level of responding
right before the beginning of the Reacquisition phase and the slope
or acceleration of the curve. As the value of bincreases, the curve
takes fewer trials to reach its asymptote, while with lower values
it takes a greater number to reach it. We estimated for each group
the values of aand b(see Figure 2). This estimation was done
considering all data points for each trial and used the Curve Fitting
toolbox of Matlab, and its implementation of the Trust Region
nonlinear least squares algorithm (Branch, Coleman, & Li, 1999).
In the Intermixed group, parameters aand bthat produced the best
fitting results were 0.363 (95% confidence interval [CI] [0.184,
0.541]) and 0.759 (95% CI [0.68, 0.838]), respectively, R
2
.332.
In the Continuous group, best-fitting aand bparameter values
were 0.061 (95% CI [0.036, 0.159]) and 1.409 (95% CI [1.243,
1.576]), respectively, R
2
.453. This indicated that there was a
statistically significant difference between the acceleration param-
eters of the reacquisition curves in both groups, as parameter b
values differed across groups.
Experiment 2
The results of Experiment 1 suggest that, in comparison to a
standard extinction procedure, a probabilistic extinction might be
effective at slowing down the rate of reacquisition. However, in
Experiment 1 only participants in the Intermixed group were
exposed to a probabilistic cue-outcome relation at some point of
the experiment. That is, only in this group one of the cues (Cue A)
was not a perfect predictor of the presence or absence of the
outcome (specifically, in Phase 2 it predicted the absence of the
outcome in 77.8% of the cases and its presence in the other 22.2%
of the cases) whereas in the other group all cues within each of the
phases had a deterministic, perfect relationship with the outcome.
It is possible that the slower learning rate observed at test is not
because of the fact that the target cue was extinguished following
a partial reinforcement procedure, but to the simple fact that
participants in the Intermixed group were exposed to a probabi-
listic contingency that increased uncertainty about the task in
general and reduced participants’ confidence about the consistency
of cue-outcome relations. In Experiment 2 all participants were
exposed to probabilistic cue-outcome relations during the three
phases of the experiments and thus, we tried to generalize the
differential reacquisition results obtained to a situation in which
Figure 2. Data points of all the participants from each group (Intermixed
group, blue dots and blue line; Continuous group, red dots and red line) in
Experiment 1, representing predictive judgments for each trial. The line
depicted is the nonlinear function that was fitted to these data. Each data
point has been displaced a small random distance in the horizontal axis to
facilitate its visual display. See the online article for the color version of
this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
uncertainty is increased and not limited to the target partial ex-
tinction phase. As can be seen in the design summary shown in
Table 2, this was achieved by including a third cue in both groups
that was partially reinforced during the whole sequence of trials, as
well as making the acquisition of Cue A also probabilistic.
Method
Participants and apparatus. Sixty-nine undergraduate Psy-
chology students from University of Málaga participated in the
experiment and received course credits for it. All of them were
tested under the same conditions as in Experiment 1.
Design and procedure. Unless stated otherwise, all proce-
dural details and design were as in Experiment 1. The main
difference with respect to the previous experiment was the inclu-
sion of a third cue, C, that held a probabilistic relation with the
outcome throughout all phases of the experiment. The additional
cue used was a yellow triangle (RGB 255, 242, 0) of height and
base of 359 pixels. As in Experiment 1, the role of each of the
stimuli was counterbalanced across participants.
The number of trials of each type was also reduced at variance
with Experiment 1 to compensate for the extra cue included. As
shown in Table 2, during Phase 1 all participants were exposed to
eight trials in which the outcome followed Cue A in seven of those
trials (87.5% of the trials were reinforced). Participants were also
exposed to eight trials of Cue B presented without the outcome,
and eight trials in which Cue C was followed by the outcome
12.5% of the times (1 out of 8 trials). As in Experiment 1, trials
were presented in a pseudorandom order. The experimental ma-
nipulation took place during Phase 2, where participants in the
Intermixed group were exposed to 24 trials of Cue A followed by
the outcome 12.5% of the time, while for participants in the
Continuous group Cue A was never followed by the outcome. In
both groups, these trials were intermixed with 24 trials with Cue B
presented on its own and 24 trials with Cue C followed by the
outcome as in Phase 1, 12.5% of the times. During the third and
final phase, all participants were exposed to 10 trials with cue A
followed by the outcome, 10 trials with B alone, and 10 trials with
the probabilistic Cue C followed by the outcome 10% of the times.
Results and Discussion
Preanalysis treatment of data. Ten participants (4 from the
Intermixed group and 6 from the Continuous group) did not meet
the inclusion criteria described in Experiment 1 and were excluded
from further analyses. After this, the Intermixed group had 30
participants, while the Continuous group had 29 participants.
Statistical analyses. All the statistical analyses were the same
as in Experiment 1, with the only exception of the number of trials,
and the Cue factor, that in this experiment had three levels as it
included also Cue C. As in the previous experiment, whenever the
sphericity assumption was violated, the Greenhouse-Geyser cor-
rection was applied.
Acquisition phase. The top panel of Figure 3 depicts the results
for the Acquisition phase. The Group Cue Trial mixed ANOVA
showed main effects of Cue (F(1.78, 101.24) 567.68, p.001,
p
2.909) and Trial (F(4.97, 283.42) 2.41, p.037, p
2.041),
but not of Group (F(1, 57) 0.08, p.775, p
2.001). The
interactions between Group Cue (F(1.78, 101.24) 0.95, p.38,
p
2.016) and Group Trial (F(4.97, 283.42) 0.4, p.848, p
2
.007) were not significant. As expected, the Cue Trial interaction
(F(7.82, 445.55) 20.77, p.001, p
2.267) was statistically
significant. This interaction together with the Cue main effect indi-
cated that the different contingencies programmed for each cue lead to
different levels of responding, and that the differences between them
were acquired across time. There was no Group Cue Trial
interaction (F(7.82, 445.55) 0.39, p.922, p
2.007). Given that
none of the contrasts including the Group factor were significant, we
can conclude that there were no differences between groups during
acquisition.
Extinction phase. The results for the Extinction phase are
shown in the middle panel of Figure 3. During extinction, we
found significant main effects of Group (F(1, 57) 9.71, p
.003, p
2.145), Cue (F(1.58, 90.3) 81.1, p.001, p
2.587),
and Trial (F(10.95, 624.42) 20.18, p.001, p
2.261), as well
as Group Cue (F(1.58, 90.3) 21.77, p.001, p
2.276) and
Cue Trial interactions (F(16.35, 932.04) 13.13, p.001,
p
2.187). The Group Trial (F(10.95, 624.42) 1.58, p
.101, p
2.027) and Group Cue Trial interactions (F(16.35,
932.04) 0.93, p.534, p
2.016) were not statistically
significant. As in the case of Experiment 1, these results were
because of the differences observed in Cue A. When the ANOVA
analyses were repeated separating across cues, there was a signif-
icant effect of Group and a Trial Group interaction (respec-
tively, F(1, 57) 27.39, p.001, p
2.325 and F(11.42,
651.11) 1.92, p.033, p
2.033) for Cue A, but not for Cues
B or C (all Fs1).
Reacquisition phase. The bottom panel of Figure 3 depicts
the results of the Reacquisition phase. As in Experiment 1, the data
obtained in the Reacquisition phase showed no main effect of
Group (F(1, 57) 1.63, p.206, p
2.028) or Group Cue
interaction (F(1.52, 86.77) 1.37, p.257, p
2.023), but there
was a significant Group Cue Trial interaction (F(9.59,
546.4) 2.27, p.014, p
2.038). Additionally, there were
significant effects of Cue (F(1.52, 86.77) 374.14, p.001,
p
2.868), Trial (F(5.74, 327.18) 34.7, p.001, p
2.378),
Group Trial (F(5.74, 327.18) 5.04, p.001, p
2.081), and
Cue Trial interactions (F(9.59, 546.4) 42.3, p.001, p
2
.426).
To follow up the critical Group Cue Trial interaction, we
ran separate analysis for each of the cues. In the case of Cue A we
found a marginally significant Group effect (F(1, 57) 3.31, p
.074, p
2.055) and significant effects of Trial (F(5.02, 286.33)
81.29, p.001, p
2.588) and a Group Trial interaction
Table 2
Design Summary of Experiment 2
Group Phase 1 Phase 2 Phase 3
Intermixed 8 A (87.5%) 24 A (12.5%) 10 A (100%)
8B (0%) 24 B (0%) 10 B (0%)
8C (12.5%) 24 C (12.5%) 10 C (10%)
Continuous 8 A (87.5%) 24 A (0%) 10 A (100%)
8B (0%) 24 B (0%) 10 B (0%)
8C (12.5%) 24 C (12.5%) 10 C (10%)
Note. The percentage of trials in which Cues A, B, and C were paired
with the outcome is indicated between brackets.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
(F(5.02, 286.33) 5.47, p.001, p
2.088). This indicated that
the differences between the two groups varied significantly across
trials. The main effect of Group and the Group Trial interaction
were nonsignificant for Cues B (Fs1.26, ps.28) and C (Fs
1.15, ps.33).
A series of post hoc ttests were run for differences between
responses to Cue A from both groups across the different trials,
showing that there were significant differences in Trial 1
(t(35.16) 3.64, p.001, d
s
0.681), Trial 4 (t(57) 2.25, p
.028, d
s
0.479), Trial 7 (t(42.03) 3.66, p.001, d
s
0.704),
Trial 9 (t(43.79) 2.77, p.008, d
s
0.537), and Trial 10
(t(43.44) 2.54, p.015, d
s
0.49). As can be seen in Figure
3, differences found in Trial 1 and those in the rest of trials were
of opposite directions. Whereas mean responses for the Intermixed
group were higher in Trial 1, they were significantly lower in the
rest of the trials.
Curve fitting. The same procedure described for Experiment
1 was applied also to the data from Experiment 2. The best-fitting
curves are shown in Figure 4. The parameters aand b that
produced best-fitting results for the Intermixed group were a
0
20
40
60
80
100
12345678
t
ne
me
g
duJ
ev
i
t
c
id
erP
n
ae
M
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Intermixed Group - Cue C
Continuous Group - Cue A
Continuous Group - Cue B
Continuous Group - Cue C
0
20
40
60
80
100
1471013161922
tn
e
m
egd
u
J ev
i
t
c
id
er
P
n
aeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Intermixed Group - Cue C
Continuous Group - Cue A
Continuous Group - Cue B
Continuous Group - Cue C
0
20
40
60
80
100
12345678910
tn
e
megd
u
J
ev
it
c
id
er
P
n
aeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Intermixed Group - Cue C
Continuous Group - Cue A
Continuous Group - Cue B
Continuous Group - Cue C
Figure 3. Mean predictive judgments of each of the three phases in Experiment 2. Acquisition, Extinction, and
Reacquisition are presented in the top, middle, and bottom panels, respectively. The error bars are the SEM.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
0.291 (95% CI [0.142, 0.439] and b0.668 (95% CI [0.586,
0.751]), respectively, R
2
.32 for the model. In the case of the
Continuous group their values were aa0.003 (95% CI [0.08
to 0.085] and b0.996 (95% CI [0.903 to 1.09]), being the
variance that the model explained R
2
.596. As in Experiment 1,
the value of parameter bwas significantly higher in the Continuous
than in the Intermixed Group, indicating that reacquisition was
faster in the former than in the latter group.
Experiment 3
The consistent pattern of results observed in Experiments 1 and
2 provides strong support for the hypothesis that a probabilistic
extinction slows down the rate of reacquisition. However, this
conclusion is potentially undermined by a common problem of
both experiments. In both cases, the experimental manipulation not
only affected the rate of reacquisition, but also the asymptotic level
of performance during extinction. By the end of extinction, par-
ticipants in the Intermixed group were still showing a substantial
amount of responding to the extinguished Cue A. Perhaps it is the
incomplete extinction of A, and not the probabilistic extinction per
se, that explains why reacquisition was slower in the Intermixed
group.
In Experiment 3 we implemented an additional change in the
design to achieve complete extinction of A by the end of the
extinction stage. Specifically, the proportion of reinforced Cue A
trials was reduced progressively during the initial blocks of ex-
tinction so that, by the end of that stage, only nonreinforced Cue A
trials were presented. As can be seen in the design summary shown
in Table 3, the last blocks of extinction included only nonrein-
forced trials in both groups. Because of this, we expected that their
levels of responding would eventually converge to similar values.
However, the predictions derived from Bouton’s model are not
affected by these changes: Even if all participants reach similar
levels of responding during extinction, those in the Intermixed
group should still show slower learning during reacquisition as
reinforced trials were still part of the extinction context, that is,
they still experienced reinforced trials followed by nonreinforced
trials.
Method
Participants and apparatus. Seventy undergraduate Psy-
chology students from University of Málaga participated in the
experiment and received course credits for it. All of them were
tested under the same conditions as in Experiments 1 and 2.
Design and procedure. The design of the experiment was
largely based on Experiment 2, except for two crucial differences.
First, to shorten the experiment as much as possible, the probabi-
listic Cue C was not included in Experiment 3. As in Experiment
1, only two cues were used throughout the sequence of trials. The
assignment of stimuli to cues was counterbalanced across partic-
ipants. Second, to obtain complete extinction we included a longer
sequence of extinction trials. Therefore, participants were exposed
to 56 Cue A trials and 56 nonreinforced Cue B trials during Phase
2. In the Intermixed group, the percentage of reinforced trials of
Cue A decreased progressively, with 5 of the first 8 trials being
reinforced, 3 of the second 8 trials, 1 of the third 8 trials, and none
of the last 32 trials. In the Continuous group, all 56 Cue A trials
were nonreinforced. Apart from these two modifications, the de-
sign and procedure were as in Experiment 2.
Results and Discussion
Preanalysis treatment of data. Given that the objective of
the experiment was to test the effect of intermixed extinction after
reaching an asymptotic level of extinction, we added an additional
exclusion criterion, similar to those used in Experiments 1 and 2.
Data from participants whose mean responses for Cue A in the last
three trials of extinction was more than 2 SDs away from the
overall group mean were excluded from further analyses. Six
participants from the Continuous group did not meet the inclusion
criteria described in Experiment 1 or the additional criterion just
mentioned. After this, the Intermixed group included 35 partici-
pants, while the Continuous group included 29 participants.
Statistical analyses. All the statistical analyses were the same
as in Experiment 1, with the only exception of the number of trials
Table 3
Design Summary of Experiment 3
Group Phase 1 Phase 2 Phase 3
Intermixed 8 A (87.5%) 56 A (16%) 10 A (100%)
8B (0%) 56 B (0%) 10 B (0%)
Continuous 8 A (87.5%) 56 A (0%) 10 A (100%)
8B (0%) 56 B (0%) 10 B (0%)
Note. The percentage of trials in which Cues A and B were paired with
the outcome is indicated between brackets. During Phase 2 in the inter-
mixed group, the percentage of reinforced trials of Cue A decreased
progressively, with 5 of the first 8 trials being reinforced, 3 of the second
8 trials, 1 of the third 8 trials, and none of the last 32 trials.
Figure 4. Data points of all the participants from each group (Intermixed
group, blue dots and blue line; Continuous group, red dots and red line) in
Experiment 2, representing predictive judgments for each trial. The line
depicted is the nonlinear function that was fitted to these data. Each data
point has been displaced a small random distance in the horizontal axis to
facilitate its visual display. See the online article for the color version of
this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
in the acquisition and reacquisition stages. As in previous exper-
iments, the Greenhouse-Geyser correction was applied when nec-
essary.
Acquisition phase. The top panel of Figure 5 depicts the
results for the Acquisition phase. The Group Cue Trial mixed
ANOVA showed main effects of Cue (F(1, 63) 2829.35, p
.001, p
2.979) and Trial (F(1.87, 115.76) 8.45, p.001, p
2
.12). As expected, the Cue Trial interaction was statistically
significant (F(2.88, 178.47) 131.48, p.001, p
2.679). All
other effects and interactions were nonsignificant (largest F
1.34, smallest p.251). These results show that both groups
learned equally the target cue-outcome contingencies.
Extinction phase. The middle panel of Figure 5 shows the
results for the Extinction phase. The Group Cue Trial ANOVA
on predictive judgments yielded main effects of Cue (F(1, 62)
219.74, p.001, p
2.78), Trial (F(14.19, 879.49) 76.41, p
.001, p
2.552), and Group (F(1, 62) 91.82, p.001, p
2
0.597). The Group Cue (F(1, 63) 100.82, p.001, p
2.619),
Group Trial (F(14.185, 879.49) 27.63, p.001, p
2.308), and
Cue Trial (F(13.12, 813.44) 57.73, p.001, p
2.48)
0
20
40
60
80
100
12345678
tnemegduJ evitciderP naeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
0
20
40
60
80
100
1 4 7 10131619222528313437404346495255
tnemegduJ evitciderP naeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
0
20
40
60
80
100
12345678910
tnemegduJ evitciderP
naeM
Trials
Intermixed Group - Cue A
Intermixed Group - Cue B
Continuous Group - Cue A
Continuous Group - Cue B
Figure 5. Mean predictive judgments of each of the three phases in Experiment 3. Acquisition, Extinction, and
Reacquisition are presented in the top, middle, and bottom panels, respectively. The error bars are the SEM.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
interactions were all statistically significant. Furthermore, the double
Group Cue Trial interaction was also significant (F(13.12,
813.44) 24.47, p.001, p
2.283).
An additional analysis of Cue A showed significant effects of
Trials (F(13.35, 827.81) 73.78, p.001, p
20.543) and the
Trials Group interaction (F(13.35, 827.81) 28.65, p
.001, p
20.316). As can be seen, this interaction is because of
the different course of extinction of Cue A in each group.
However, a series of ttests confirmed that there were no
significant differences across groups in responding to Cue A in
the last eight trials (all ps.157, with all the differences
between means of the groups under 3.38 points), indicating that
the same level of extinction had been reached in both groups at
the end of the extinction phase, at variance with the results from
previous experiments.
Reacquisition phase. The bottom panel of Figure 5 depicts
the results of the Reacquisition phase. As in Experiments 1 and 2,
the main effect of Group (F(1, 62) 2.219, p.141, p
2.035),
and the Group Cue interaction (F(1, 62) 1.81, p.183, p
2
.028) failed to reach statistical significance. However, we found
significant effects of Cue (F(1, 62) 2399.17, p.001, p
2
.975), Trial (F(3.11, 192.68) 263.44, p.001, p
2.809),
Group Trial (F(3.11, 192.68) 5.60, p.001, p
2.083), and
Cue Trial interactions (F(3.87, 239.91) 198.59, p.001,
p
2.762). Most important, we also found a significant Group
Cue Trial interaction (F(3.87, 239.91) 5.32, p.001, p
2
.079).
To further explore the critical Group Cue Trial interaction,
we ran separate analysis for each of the cues. Judgments for Cue
A showed a significant effect of Trial (F(2.77, 171.55) 278.97,
p.001, p
2.818) and a significant Group Trial interaction
(F(2.77, 171.55) 6.46, p.001, p
2.094). This interaction
confirms that the rate of reacquisition of Cue A differed between
groups. The main effect of Group did not reach statistical signif-
icance (F(1, 62) 2.07, p.155, p
2.032). In the case of
judgments for Cue B, none of the main effects or the interaction
was statistically significant (all ps0.394).
A series of post hoc ttests were run for differences between
responses to Cue A from both groups across the different trials,
showing that there were significant differences in Trial 2
(t(59.05) 3.12, p.003, d
s
0.783) and Trial 3 (t(61.73)
2.06, p.044, d
s
0.517). Judgments for A did not differ
significantly in the rest of the trials (largest t1.13, smallest p
.265).
Curve fitting. We computed the best-fitting curves to the
reacquisition of Cue A following the procedure described in pre-
vious experiments. Figure 6 depicts the best-fitting curves. The
parameters aand bthat produce best-fitting results for the Inter-
mixed group were a⫽⫺0.020 (95% CI [0.079, 0.039]) and b
1.268 (95% CI [1.167, 1.369]), respectively, R
2
.72 for the
model. In the case of the Continuous group their values were a
0.040 (95% CI [0.037 to 0.117]) and b1.718 (95% CI [1.508
to 1.929]), being the variance that the model explained R
2
.66.
As in Experiments 1 and 2, the best-fitting parameter bwas
significantly higher in the Continuous than in the Intermixed
Group, confirming that reacquisition was faster in the former
group.
General Discussion
The results reported provide evidence of a slower reacquisition
after an intermixed or partial extinction procedure in human con-
tingency learning. Specifically, the rate or reacquisition of a pre-
viously extinguished response was lower after a partial than after
a standard extinction procedure. This effect appeared when control
conditions consisted of deterministic cue-outcome relationships
(Experiment 1) as well as probabilistic cue-outcome relationships
(Experiment 2). And importantly, Experiment 3 showed that the
differential reacquisition was still obtained even when expectancy
judgments had reached asymptotic levels of extinction. Thus,
slower reacquisition cannot be regarded as a by-product of a
previous incomplete extinction but the genuine effect of intermix-
ing reinforced trials during the extinction treatment.
In Experiments 1 and 2, responses to Cue A differed in both
groups of participants at the end of the extinction phase. This
happened even when the probability of reinforcement was rela-
tively low (12.5% in Experiment 2). This fact introduces a poten-
tial confound in the interpretation of the results, as it makes it
harder to compare the reacquisition rate of both groups. However,
we would argue that in fact this supports the robustness of the
effect. Although participants in the Continuous group had lower
levels of responding at the end of the extinction phase than those
in the Intermixed Group, participants‘ responses from the former
group increased during the initial trials of the Reacquisition phase.
Therefore, should both groups had reached the same level of
responding at the end of the extinction phase, even sharper differ-
ences between groups would be expected during the first reacqui-
sition trials. This conclusion is supported by our curve fitting
results. Parameter b, which represented the reacquisition rate, was
lower in the Intermixed groups even when controlling for the
Figure 6. Data points of all the participants from each group (Intermixed
group, blue dots and blue line; Continuous group, red dots and red line) in
Experiment 3, representing predictive judgments for each trial. The line
depicted is the nonlinear function that was fitted to these data. Each data
point has been displaced a small random distance in the horizontal axis to
facilitate its visual display. See the online article for the color version of
this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
initial level of responding during the Reacquisition phase (Param-
eter a). On the other hand, the results from Experiment 3 proved
that the slower reacquisition effect was not because of differences
in the response level at the end of the extinction phase in both
groups, but to the partial extinction programmed.
The high level of responding obtained despite the low number of
reinforced trials in Experiments 1 and 2 may have been because of
changes in the level of unconditioned responding to the outcome.
It has been suggested that the level of unconditioned responding
produced by an aversive stimulus can vary because of several
factors. Repeated presentation of the outcome can lead to its
habituation (e.g., Rescorla, 1973), and whenever an outcome is
correctly anticipated, the level of unconditioned responding is
reduced (Dunsmoor, Bandettini, & Knight, 2008; Grings, 1973;
Lykken & Tellegen, 1974), and made less aversive (Schell &
Grings, 1971). In our experiments, a reduction in the predictability
of the outcome could have led to an increase in its aversiveness,
causing higher levels of conditioned responding. Consistently with
this idea, response levels in Experiment 3 extinguished completely
when the reinforcement schedule diminished progressively during
extinction, rendering the absence of the outcome more and more
predictable.
Overall, the pattern of results obtained mimicked the target
effect found in the animal conditioning literature (Bouton et al.,
2004; Gershman, Jones, Norman, Monfils, & Niv, 2013; Woods &
Bouton, 2007). To the best of our knowledge, there is only another
demonstration of a similar partial extinction manipulation in hu-
man learning. van den Akker, Havermans, and Jansen (2015) have
showed a slowed reacquisition of verbal expectancy, using food as
an outcome in an appetitive task. Put together, these findings point
to the possibility of a general effect across types of reinforcers,
although more evidence is required.
Although the level of uncertainty (i.e., nondeterministic cue-
outcome relationships) in Experiments 2 and 3 was more similar
across groups than in Experiment 1, they were not completely
equated. Previous studies have shown that increased uncertainty
can lead to higher learning rates (e.g., Behrens, Woolrich, Walton,
& Rushworth, 2007; Courville, Daw, & Touretzky, 2006; although
see Le Pelley, 2004 and Le Pelley & McLaren, 2003). However,
our results show the opposite pattern of results, as the Intermixed
group, in which the level of uncertainty during extinction was
higher than in the Continuous group, showed a slower learning rate
during the Reacquisition phase. Therefore, it is hard to explain the
pattern of data obtained solely on this difference of uncertainty
between the conditions.
On the other hand, our results are compatible with the theoret-
ical proposal by Bouton et al. (2004). According to this account,
when extinction takes place, reinforced trials become a cue that
indicates whether the current context is the extinction or the
acquisition context. This potential role of the reinforced trials as a
contextual cue is degraded by introducing these reinforced trials
during the extinction phase. Thus, reinforced trials would not
univocally signal the acquisition context, which in turn, would
finally lead to a slower reacquisition. However, this is not the only
explanation available, as other theoretical models for this effect
have been put forward. For example, according to Gershman, Blei,
and Niv (2010) the mechanism involved would be quite different.
They propose that persisting large prediction error provide a signal
that is used to segment learning. Future trials will be coded as part
of a new context, with new associations. In the case of Experiment
3 according to this theory, persisting large prediction errors would
occur during the change from the acquisition to the extinction
phase in the case of our control groups. This would be because of
the sudden shift from a reinforced to a nonreinforced scenario.
However, this should take place to a lesser extent in the case of the
experimental group, which changes from a reinforced to a partially
reinforced scenario. In the former, acquisition and extinction
would be encoded in separate associations, as they would be
segmented in different contexts, while in the latter a single asso-
ciation would be updated across the trials of both phases (see
Gershman et al., 2013 for a detailed discussion).
The main objective of this study was to provide evidence of the
effect of partial reinforcement during extinction in human contin-
gency learning, and more experiments are required to better un-
derstand the mechanisms underlying this effect. For example, in
our curve fitting analysis, we have implicitly assumed that reac-
quisition can be modeled as a negatively accelerated curve with
different learning rates in each group. However, this might not be
necessarily the case. Authors like Charles R. Gallistel (Gallistel
2012a, 2012b; Gallistel, Fairhurst, & Balsam, 2004) have proposed
that learning is better described as an all-or-nothing process, in
which organisms switch between multiple states that have different
levels of response. From this point of view, the learning curve
described in most studies is only the effect of averaging the
responses of multiple participants who make this transition at
different moments. This could have happened in this series of
experiments. Specific data analyses, beyond the scope of this
work, have to be used to properly test this possibility and differ-
entiate between these two proposals (see Blanco & Morís, 2016,
for an extended discussion of this problem and possible data
analysis solutions based on Bayesian modeling). Furthermore, it
must be noted that theories like Bouton’s model can easily accom-
modate both types of mechanism. Additionally, the implications of
both potential mechanisms for relapse reduction are not substan-
tially different either. No matter whether the partial extinction
treatment has altered the learning rate of reacquisition or has
delayed the moment when participants detect the new contingency,
relapse may be understood as having been reduced. Future exper-
iments are needed to specifically test what mechanisms underlie
this phenomenon in humans, their boundary conditions and how to
optimize the reduction of relapse.
The relevance of the results from Experiment 3 is not only
theoretical. The pattern of results obtained in Experiment 3 showed
that partial extinction may potentiate extinction effects even if
maximal extinction has already been achieved. Of course, reaching
the lowest possible levels of responding once extinction is com-
plete constitutes one of the most important therapeutic outcomes in
anxiety disorder treatments. The results of Experiment 3 showed
that a partial extinction procedure does not prevent complete
extinction of expectancy ratings. Although these results are still far
from being directly applicable to clinical scenarios, they suggest
that partial extinction is not necessarily an obstacle to reach low
levels of conditioned responding.
To explore more thoroughly the possible clinical implications of
the effect found, it would be necessary to know whether similar
effects may be obtained using a more aversive paradigm (note the
mild aversive sound used in our task) and registering alternative
responses to expectancy ratings, such as subjective fear ratings or
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
behavioral responses for example, the potentiated startle response
or avoidance behavior, that might also reflect more emotional
aspects of fear. These different measures may too be tapping
different processes engaged in human learning that may interact in
different ways and, importantly, that can be dissociated (e.g.,
Morís, Cobos, Luque, & López, 2014; Sevenster, Beckers, &
Kindt, 2012; Soeter & Kindt, 2011). However, this interest should
not obscure the relevance of expectancy ratings themselves, as
danger expectancies play an important role in cognitive theories of
fear and anxiety and these expectancy ratings may be regarded as
a highly valid measure of fear conditioning (see Boddez, Baeyens,
Luyten, Vansteenwegen, Hermans, & Beckers, 2013).
Slower reacquisition can be of great applied relevance in those
situations in which the patient might have to experience again the
aversive situation leading to the emotional disorder, either because
of observational learning (Askew & Field, 2008; Todd & Pi-
etrowski, 2007), or a direct re-exposure, like in the case of social
phobias or posttraumatic stress disorder (Craske et al., 2014). For
example, patients suffering from social anxiety disorder may re-
experience, after exposure therapy, a truly aversive event related to
social failure (e.g., a deeply embarrassing or traumatic situation in
a social context), that may be viewed as a reacquisition experience.
Our result suggests the idea that a partial version of exposure
therapy may induce some form of resilience to manage future
reencounters with truly anxiogenic social situations. Nonetheless,
this is not the most widespread relapse effect, nor the most com-
mon experimental model of relapse. Finding an equivalent allevi-
ating effect of these other forms of relapse may, in principle,
increase the clinical interest of the results found. Only future
research will have a say in this.
Because of its easy implementation, and its potential use to
reduce relapse, intermixed extinction might constitute in the future
a way to improve extinction-based therapies. It is also a very
interesting phenomenon to further develop the theories of associa-
tive learning and extinction, and it can serve as an experimental
tool to further test their predictions. Moreover, there are many
open questions regarding this effect, its boundary conditions, and
how to maximize it. All in all, it may be regarded as a promising
venue of research.
References
Askew, C., & Field, A. P. (2008). The vicarious learning pathway to fear
40 years on. Clinical Psychology Review, 28, 1249–1265. http://dx.doi
.org/10.1016/j.cpr.2008.05.003
Beckers, T., Krypotos, A.-M., Boddez, Y., Effting, M., & Kindt, M.
(2013). What’s wrong with fear conditioning? Biological Psychology,
92, 90–96. http://dx.doi.org/10.1016/j.biopsycho.2011.12.015
Behrens, T. E., Woolrich, M. W., Walton, M. E., & Rushworth, M. F.
(2007). Learning the value of information in an uncertain world. Nature
Neuroscience, 10, 1214–1221. http://dx.doi.org/10.1038/nn1954
Blanco, F., & Morís, J. (2016). Bayesian methods for solving long-standing
problems in associative learning: The case of PREE. Manuscript sub-
mitted for publication.
Boddez, Y., Baeyens, F., Luyten, L., Vansteenwegen, D., Hermans, D., &
Beckers, T. (2013). Rating data are underrated: Validity of US expec-
tancy in human fear conditioning. Journal of Behavior Therapy and
Experimental Psychiatry, 44, 201–206. http://dx.doi.org/10.1016/j.jbtep
.2012.08.003
Bouton, M. E. (1993). Context, time, and memory retrieval in the inter-
ference paradigms of Pavlovian learning. Psychological Bulletin, 114,
80–99. http://dx.doi.org/10.1037/0033-2909.114.1.80
Bouton, M. E. (2002). Context, ambiguity, and unlearning: Sources of
relapse after behavioral extinction. Biological Psychiatry, 52, 976–986.
http://dx.doi.org/10.1016/S0006-3223(02)01546-9
Bouton, M. E. (2004). Context and behavioral processes in extinction.
Learning & Memory, 11, 485–494. http://dx.doi.org/10.1101/lm.78804
Bouton, M. E., Woods, A. M., & Pineño, O. (2004). Occasional reinforced
trials during extinction can slow the rate of rapid reacquisition. Learning
and Motivation, 35, 371–390. http://dx.doi.org/10.1016/j.lmot.2004.05
.001
Branch, M. A., Coleman, T. F., & Li, Y. (1999). A subspace, interior, and
conjugate gradient method for large-scale bound-constrained minimiza-
tion problems. SIAM Journal on Scientific Computing, 21, 1–23. http://
dx.doi.org/10.1137/S1064827595289108
Capaldi, E. J. (1967). A sequential hypothesis of instrumental learning. In
K. W. Spence & J. T. Spence (Eds.), The psychology of learning and
motivation (pp. 67–156). New York, NY: Academic Press.
Capaldi, E. J. (1994). The sequential view: From rapidly fading stimulus
traces to the organization of memory and the abstract concept of number.
Psychonomic Bulletin & Review, 1, 156–181. http://dx.doi.org/10.3758/
BF03200771
Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the
future of cognitive science. Behavioral and Brain Sciences, 36, 181–
204. http://dx.doi.org/10.1017/S0140525X12000477
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Courville, A. C., Daw, N. D., & Touretzky, D. S. (2006). Bayesian theories
of conditioning in a changing world. Trends in Cognitive Sciences, 10,
294–300. http://dx.doi.org/10.1016/j.tics.2006.05.004
Craske, M. G., Liao, B., Brown, L., & Vervliet, B. (2012). Role of
inhibition in exposure therapy. Journal of Experimental Psychopathol-
ogy, 3, 322–345. http://dx.doi.org/10.5127/jep.026511
Craske, M. G., & Mystkowski, J. L. (2006). Exposure therapy and extinc-
tion: Clinical studies. In M. G. Craske, D. Hermans, & D. Vansteenwe-
gen (Eds.), Fear and learning: Basic science to clinical application (pp.
217–233). Washington, DC: APA Books.
Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., & Vervliet, B.
(2014). Maximizing exposure therapy: An inhibitory learning approach.
Behaviour Research and Therapy, 58, 10–23. http://dx.doi.org/10.1016/
j.brat.2014.04.006
Dunsmoor, J. E., Bandettini, P. A., & Knight, D. C. (2008). Neural correlates of
unconditioned response diminution during Pavlovian conditioning. NeuroIm-
age, 40, 811–817. http://dx.doi.org/10.1016/j.neuroimage.2007.11.042
Friston, K. (2003). Learning and inference in the brain. Neural Networks,
16, 1325–1352. http://dx.doi.org/10.1016/j.neunet.2003.06.005
Gallistel, C. R. (2012a). Extinction from a rationalist perspective. Behav-
ioural Processes, 90, 6680. http://dx.doi.org/10.1016/j.beproc.2012.02
.008
Gallistel, C. R. (2012b). On the evils of group averaging: Commentary on
Nevin’s “Resistance to extinction and behavioral momentum.” Behav-
ioural Processes, 90, 98–99. http://dx.doi.org/10.1016/j.beproc.2012.02
.013
Gallistel, C. R., Fairhurst, S., & Balsam, P. (2004). The learning curve:
Implications of a quantitative analysis. Proceedings of the National
Academy of Sciences of the United States of America, 101, 13124
13131. http://dx.doi.org/10.1073/pnas.0404965101
Gershman, S. J., Blei, D. M., & Niv, Y. (2010). Context, learning, and
extinction. Psychological Review, 117, 197–209. http://dx.doi.org/10
.1037/a0017808
Gershman, S. J., Jones, C. E., Norman, K. A., Monfils, M. H., & Niv, Y.
(2013). Gradual extinction prevents the return of fear: Implications for
the discovery of state. [Advance online publication]. Frontiers in Be-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 MORÍS, BARBERIA, VADILLO, ANDRADES, AND LÓPEZ
havioral Neuroscience, 7, 164. http://dx.doi.org/10.3389/fnbeh.2013
.00164
Grings, W. W. (1973). Cognitive factors in electrodermal conditioning. Psycho-
logical Bulletin, 79, 200–210. http://dx.doi.org/10.1037/h0033883
IBM Corp. (2012). IBM SPSS statistics for Windows, Version 21.0. Ar-
monk, NY: IBM Corp.
Le Pelley, M. E. (2004). The role of associative history in models of
associative learning: A selective review and a hybrid model. The Quar-
terly Journal of Experimental Psychology Section B, 57, 193–243.
http://dx.doi.org/10.1080/02724990344000141
Le Pelley, M. E., & McLaren, I. P. L. (2003). Learned associability and
associative change in human causal learning. The Quarterly Journal of
Experimental Psychology: Section B, 56, 68–79. http://dx.doi.org/10
.1080/02724990244000179
Longmore, R. J., & Worrell, M. (2007). Do we need to challenge thoughts
in cognitive behavior therapy? Clinical Psychology Review, 27, 173–
187. http://dx.doi.org/10.1016/j.cpr.2006.08.001
Lykken, D. T., & Tellegen, A. (1974). On the validity of the preception
hypothesis. Psychophysiology, 11, 125–132. http://dx.doi.org/10.1111/j
.1469-8986.1974.tb00833.x
Milad, M. R., & Quirk, G. J. (2012). Fear extinction as a model for translational
neuroscience: Ten years of progress. Annual Review of Psychology, 63, 129
151. http://dx.doi.org/10.1146/annurev.psych.121208.131631
Mineka, S., & Zinbarg, R. (2006). A contemporary learning theory per-
spective on the etiology of anxiety disorders: It’s not what you thought
it was. American Psychologist, 61, 10–26. http://dx.doi.org/10.1037/
0003-066X.61.1.10
Morís, J., Cobos, P. L., Luque, D., & López, F. J. (2014). Associative
repetition priming as a measure of human contingency learning: Evi-
dence of forward and backward blocking. Journal of Experimental
Psychology: General, 143, 77–93. http://dx.doi.org/10.1037/a0030919
Norton, P. J., & Price, E. C. (2007). A meta-analytic review of adult
cognitive-behavioral treatment outcome across the anxiety disorders.
Journal of Nervous and Mental Disease, 195, 521–531. http://dx.doi
.org/10.1097/01.nmd.0000253843.70149.9a
Pavlov, I. P. (1927). Conditioned reflexes. London: Oxford University
Press.
Rachman, S. (1966). Studies in desensitization. 3. Speed of generalization.
Behaviour Research and Therapy, 4, 7–15. http://dx.doi.org/10.1016/
0005-7967(66)90038-6
Rescorla, R. A. (1973). Effect of US habituation following conditioning.
Journal of Comparative and Physiological Psychology, 82, 137–143.
http://dx.doi.org/10.1037/h0033815
Ricker, S. T., & Bouton, M. E. (1996). Reacquisition following extinction
in appetitive conditioning. Animal Learning & Behavior, 24, 423–436.
http://dx.doi.org/10.3758/BF03199014
Schell, A. M., & Grings, W. W. (1971). Judgments of UCS intensity and
diminution of the unconditioned GSR. Psychophysiology, 8, 427–432.
http://dx.doi.org/10.1111/j.1469-8986.1971.tb00475.x
Sevenster, D., Beckers, T., & Kindt, M. (2012). Retrieval per se is not
sufficient to trigger reconsolidation of human fear memory. Neurobiol-
ogy of Learning and Memory, 97, 338–345. http://dx.doi.org/10.1016/j
.nlm.2012.01.009
Shanks, D. R. (2010). Learning: From association to cognition. Annual
Review of Psychology, 61, 273–301. http://dx.doi.org/10.1146/annurev
.psych.093008.100519
Soeter, M., & Kindt, M. (2011). Disrupting reconsolidation: Pharmacolog-
ical and behavioral manipulations. Learning & Memory, 18, 357–366.
http://dx.doi.org/10.1101/lm.2148511
The MathWorks, Inc. (2012). MATLAB and statistics toolbox release
2012b. The MathWorks, Inc.: Natick, MA.
Todd, J. T., & Pietrowski, J. L. (2007). Animal models of exposure
therapy: A selective review. In D. C. Richard & D. Lauterbach (Eds.),
Handbook of exposure therapies (pp. 29–59). New York, NY: Aca-
demic Press. http://dx.doi.org/10.1016/B978-012587421-2/50003-X
van den Akker, K., Havermans, R. C., & Jansen, A. (2015). Effects of
occasional reinforced trials during extinction on the reacquisition of
conditioned responses to food cues. Journal of Behavior Therapy and
Experimental Psychiatry, 48, 50–58. http://dx.doi.org/10.1016/j.jbtep
.2015.02.001
Vervliet, B., Craske, M. G., & Hermans, D. (2013). Fear extinction and
relapse: State of the art. Annual Review of Clinical Psychology, 9,
215–248. http://dx.doi.org/10.1146/annurev-clinpsy-050212-185542
Woods, A. M., & Bouton, M. E. (2007). Occasional reinforced responses
during extinction can slow the rate of reacquisition of an operant
response. Learning and Motivation, 38, 56–74. http://dx.doi.org/10
.1016/j.lmot.2006.07.003
Received March 11, 2015
Revision received March 8, 2016
Accepted March 10, 2016
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
SLOWER REACQUISITION AFTER PARTIAL EXTINCTION
... Fig. 2 shows the mean responses per minute in the last extinction or for both groups. This finding is similar to other studies that have reported lower reacquisition after partial extinction (Bouton et al., 2004;Morís et al., 2017), and probably similar processes are involved in both, partial extinction and delay of reinforcement. The present study assessed the effects of delay of reinforcement on ABA renewal. ...
Article
ABA renewal was assessed by decreasing response rates by means of extinction or a 12-s delay of reinforcement in one experiment with rats. In a first phase, rats were trained to lever press for water in Context A. Rats were exposed to a second phase in Context B, where lever-pressing decreased for all rats; for one group of rats, lever-pressing decreased by means of extinction, while lever-pressing decreased by means of a 12-s delay for the other group. During a third phase, both groups were once again exposed to Context A with no reinforcer deliveries for lever-pressing. Once rats were exposed to the latter phase, renewal was observed for both groups. Results suggest that delay of reinforcement can be used as an alternative procedure to help reduce behavior and mitigate relapse.
... In CET, one way to implement this technique is by encouraging individuals to occasionally consume small amounts of food during cue exposure exercises, in order to learn that eating a small amount of food is no longer a cue for overeating [13]. In support, animal and human appetitive conditioning [51,76], as well as fear conditioning studies [77,78], has reported evidence for occasional reinforcements during extinction to reduce relapse of responding (i.e., rapid reacquisition). In studies on CET, occasional reinforced extinction has not yet been studied in isolation, but some studies have incorporated this technique in exposure sessions by letting participants take small bites of high-calorie food previously associated with overeating [74•, 75]. ...
Article
Full-text available
Purpose of review: This review provides an overview of recent findings relating to the role of Pavlovian conditioning in food cue reactivity, including its application to overeating and weight loss interventions. Recent findings: Both in the laboratory and in real life, cue-elicited appetitive reactivity (e.g., eating desires) can be easily learned, but (long-term) extinction is more difficult. New findings suggest impaired appetitive learning in obesity, which might be causally related to overeating. The clinical analogue of extinction-cue exposure therapy-effectively reduces cue-elicited cravings and overeating. While its working mechanisms are still unclear, some studies suggest that reducing overeating expectancies is important. Summary: Pavlovian learning theory provides a still undervalued theoretical framework of how cravings and overeating can be learned and how they might be effectively tackled. Future studies should aim to elucidate inter-individual differences in Pavlovian conditioning, study ways to strengthen (long-term) extinction, and investigate the working mechanisms of cue exposure therapy.
Article
Full-text available
Most associative models typically assume that learning can be understood as a gradual change in associative strength that captures the situation into one single parameter, or representational state. We will call this view single-state learning. However, there is ample evidence showing that under many circumstances different relationships that share features can be learned independently, and animals can quickly switch between expressing one or another. We will call this multiple-state learning. Theoretically, it is understudied because it needs a different data analysis approach from those usually employed. In this paper, we present a Bayesian model of the Partial Reinforcement Extinction Effect (PREE) that can test the predictions of the multiple-state view. This implies estimating the moment of change in the responses (from the acquisition to the extinction performance), both at the individual and at the group levels. We used this model to analyze data from a PREE experiment with three levels of reinforcement during acquisition (100%, 75% and 50%). We found differences in the estimated moment of switch between states during extinction, so that it was delayed after leaner partial reinforcement schedules. The finding is compatible with the multiple-state view. It is the first time, to our knowledge, that the predictions from the multiple-state view are tested directly. The paper also aims to show the benefits that Bayesian methods can bring to the associative learning field.
Article
Full-text available
While many researchers have largely focused on principles of systematic desensitization and habituation in explaining fear extinction, these processes have mixed evidence at best. In particular, these models do not account for spontaneous recovery or reinstatement of fear, nor do they explain the context dependency of extinction or rapid reacquisition. This may in part account for the significant number of patients who fail to respond to our available treatments which rely on these principles in designing exposure sessions. However, recent research is converging to suggest that an inhibitory model of fear reduction, in which the original feared association (CS-US) remains but is inhibited by a newly formed association (CS-noUS) representing safety, holds promise in explaining the long-term attenuation of fear and anxiety. This paper reviews research in a number of areas, including neuroimaging, psychophysiology, and psychopharmacology that all provide support for the inhibition model of anxiety. Limitations to this body of research are discussed, along with recommendations for future research and suggestions for improving exposure therapy for fear and anxiety disorders. Clinical implications discussed in this paper include incorporating random and variable practice in exposure sessions, multiple contexts, and pharmacological aides, among others.
Article
Full-text available
Fear memories are notoriously difficult to erase, often recovering over time. The longstanding explanation for this finding is that, in extinction training, a new memory is formed that competes with the old one for expression but does not otherwise modify it. This explanation is at odds with traditional models of learning such as Rescorla-Wagner and reinforcement learning. A possible reconciliation that was recently suggested is that extinction training leads to the inference of a new state that is different from the state that was in effect in the original training. This solution, however, raises a new question: under what conditions are new states, or new memories formed? Theoretical accounts implicate persistent large prediction errors in this process. As a test of this idea, we reasoned that careful design of the reinforcement schedule during extinction training could reduce these prediction errors enough to prevent the formation of a new memory, while still decreasing reinforcement sufficiently to drive modification of the old fear memory. In two Pavlovian fear-conditioning experiments, we show that gradually reducing the frequency of aversive stimuli, rather than eliminating them abruptly, prevents the recovery of fear. This finding has important implications for theories of state discovery in reinforcement learning.
Article
In animal models of exposure therapy, there remains a high degree of correspondence between observations and research findings on the use of various forms of exposure therapy in humans with the results of animal analog experiments. The effects of all major parameters of exposure therapy procedures including duration of treatment, the use of distraction and contingent reinforcement, and the action of a range of drugs are highly similar in animals and humans. Close correspondences have been found for the effects of drugs in combination with exposure therapy. Anxiety in animal models responds to exposure-like treatments in a way that is similar in humans. One important implication of the close correspondence between animal and human behavior in anxiety and exposure treatment situation has to do with conceptual and pragmatic status of some newer modes of theorizing. Great success has been obtained with animal models that remain closely tied to descriptive principles of behavior.
Article
In four experiments utilizing an appetitive conditioning preparation, reacquisition of conditioned responding was found to occur both rapidly and slowly following extinction. In Experiment 1, acquisition of responding to a tone that had been conditioned and extinguished occurred more rapidly than acquisition in either a group that received equivalent exposure to the food unconditioned stimulus or a 'rest' control group that received only exposure to the apparatus in the first two phases. However, reacquisition was impaired relative to acquisition in a 'learning-experienced' group that had previously received conditioning and extinction with a different stimulus. Experiments 2 and 3 produced similar results, but also found that high responding during reacquisition was confined to trials that followed reinforced, rather than nonreinforced, trials. Experiment 4, in which very few initial conditioning trials were used, produced reacquisition that was slow compared with both learning-experienced and rest controls. The results are consistent with a role for sequential learning: Reacquisition is rapid when animals have learned that reinforced trials signal other reinforced trials.
Article
Background and objectives Successful long-term dieting appears to be difficult, and part of its difficulty might be explained by processes related to classical appetitive conditioning. Increasing the speed of extinction of appetitive responses to food cues and decreasing the magnitude of returns of these responses could help increase the long-term effectiveness of weight loss attempts. Two extinction techniques hypothesized to slow down rapid reacquisition of conditioned appetitive responses were investigated: the provision of 1) occasional reinforced extinction trials (OR) and 2) unpaired unconditioned stimuli (USs) during extinction (UNP). Methods After acquisition, participants (N = 90) received one of three extinction trainings: OR, UNP, or normal extinction (control), followed by a reacquisition phase. Their desire to eat, US expectancy, and salivation were measured. Effects of impulsivity on different phases of appetitive conditioning were also assessed. Results It was found that both extinction techniques were successful in reducing the rate of reacquisition of US expectancies. Participants in the OR condition also demonstrated a slower extinction of US expectancies and desires to eat. However, the reacquisition of conditioned desires was not affected by either extinction technique. Impulsivity did not moderate responses during acquisition or extinction, but appeared to slow down the reacquisition of conditioned desires. Limitations US expectancies and eating desires were not completely extinguished, and a few differences in baseline responses caused difficulty in interpreting some of the findings. Conclusions It is concluded that the provision of occasional reinforced extinction trials and unpaired USs seem promising techniques to slow down reacquisition, but that additional studies are needed.