Content uploaded by Julie Y.L. Chow
Author content
All content in this area was uploaded by Julie Y.L. Chow on Mar 27, 2025
Content may be subject to copyright.
45
Occasion Setting in Humans: Norm or Exception?
In this commentary, we propose that occasion setting in humans may be even more widespread than
Leising and colleagues assume. In contrast to animal work, our research using the feature negative
procedure (A+ AX−) reveals substantial individual dierences in what people learn about the feature
(X). We discuss ndings showing that the majority of participants in our experiments learn something
akin to occasion setting, and we present reasons for why this may be the case. We conclude that
occasion setting may be the norm in humans because it allows existing learning to be preserved and
allows for the possibility that the eect of a cue is unique to its accompanying target.
Keywords: occasion setting, feature negative, modulation, conditioned inhibition, prevention learning
Leising, Nerz, Solorzano-Restrepo, and Bond (2025)
present a comprehensive review on the theoretical and
methodological issues surrounding occasion setting in
a number of species. We agree that occasion setting, or
modulation, is a potentially widespread phenomenon in
human learning and that greater consideration should be
paid to the possibility that humans learn hierarchical or
higher-order associations in addition to direct or rst-or-
der associations. In fact, our ndings in causal learning
procedures lead us to believe that in humans, modulation
may be the norm rather than the exception.
Causal learning often requires individuals to make
inferences about cues and their outcomes based on in-
complete or ambiguous information. For example, in a
feature negative discrimination1 commonly used in the
occasion-setting literature (A+ AX−), it is unclear whether
X directly prevents the outcome (i.e., conditioned inhibi-
tion or prevention) or whether X prevents A from causing
the outcome (i.e., negative occasion setting, or what we
have termed modulation). Situations of high ambiguity
1. We focus on negative occasion setting in this commentary, as
we have investigated negative occasion setting in our research.
are ripe for the observation of individual dierences in
what people learn. Indeed, we observed this exact result in
multiple experiments using self-report (Chow et al., 2022,
2024; Lee & Lovibond, 2021; Lovibond & Lee, 2021;
Lovibond et al., 2022). To our surprise, we found that very
few participants spontaneously described their learning by
using words such as “inhibition” or “prevention” despite
prevailing assumptions that learning a direct negative
association is dominant under these conditions. Instead,
we found that a large proportion of participants reported a
modulatory causal structure (i.e., X stopped A from causing
an allergic reaction) over a preventative structure (i.e., X
prevented allergic reactions) or a congural structure (i.e.,
A and X together produced no allergic reaction). These
ndings were surprising because conditioned inhibition
(i.e., prevention learning) has been seen as the default
mode of learning in feature negative discriminations, with
serial presentation of the AX compound assumed to be
necessary in order to observe occasion setting (Holland,
1984, 1989, 1991; Holland & Lamarre, 1984; Lamarre &
Holland, 1985; see Bonardi et al., 2017, and Fraser & Hol-
land, 2019, for reviews). In contrast, a direct comparison
ISSN: 1911-4745
DOI:10.3819/CCBR.2025.200003 Volume 20, 2025
Jessica C. Lee
University of Sydney
University of New South Wales
Julie Y.-L. Chow
University of New South Wales
Peter F. Lovibond
University of New South Wales
COMMENTARY
46
COMPARATIVE COGNITION & BEHAVIOR REVIEWS
Author Note: Correspondence concerning this article should be
addressed to Jessica Lee at jessica.c.lee@sydney.edu.au
Jessica Lee, School of Psychology, University of Sydney, Camp-
erdown, 2006, New South Wales, Australia.
Lee, Chow, and Lovibond
of simultaneous and serial presentation of stimuli revealed
no signicant dierence in the kinds of causal structure
endorsed by participants, with the majority of participants
endorsing a modulatory structure in both conditions
(Lovibond & Lee, 2021).
In retrospect, the dominance of modulation learning
makes sense. In the presence of ambiguity due to limited
information, participants may lean toward modulation,
as it is the more conservative inference to make if par-
ticipants are unable to observe the eects of the feature
X with other cues. Modulation also helps to preserve
existing learning about the target A, which may be less
resource intensive when A is being learned alongside
AX. Consistent with this idea, Fraser and Holland (2019)
pointed out that one of the hypotheses guiding early ani-
mal research was that occasion setting is more likely when
the target is learned about faster than the feature. Under
this hypothesis, our failure to obtain a dierence between
serial and simultaneous presentation in feature negative
learning in humans (Lovibond & Lee, 2021) may be ex-
plained by assuming that learning of the target was rapid
in both conditions. Rapid learning may occur because both
serial and simultaneous procedures include trials in which
the target is presented without any other cues, making its
direct relationship with the outcome unambiguous and
easy to identify. We generally nd rapid learning under
these conditions, making it more likely that participants
will try to preserve this learning through modulation (a
form of theory protection, discussed next).
The majority of participants not only report learning
a modulatory causal structure but also seem to behave in
ways that are consistent with occasion setting. The rst
piece of evidence comes from studies examining the de-
gree of transfer observed by modulation and prevention
subgroups in a summation test (Leising et al., 2025, Test
2; see Rescorla, 1969). A summation test involves training
another cue as a predictor of the outcome (e.g., B+), and
then combining the feature X with that predictor (BX),
to test its ability to modulate responding to the predictor.
Summation is observed if the addition of the negative
feature to the separately trained predictor (BX) reduces
responding relative to the predictor alone (B). We con-
sistently nd strong summation for participants reporting
a modulatory causal structure, albeit weaker summation
than for those reporting a prevention structure (Chow et
al., 2022; Lee & Lovibond, 2021; Lovibond & Lee, 2021).
Notably, transfer was determined by the degree of similar-
ity between the training and test targets. Further, we have
repeatedly failed to show complete transfer of inhibition
in summation tests, even among individuals who report
learning a preventive causal structure. We have argued
that these results are consistent with “preventors” having
also learned a modulatory causal structure but being
more willing to generalize the modulatory properties of
the feature to new stimuli (Chow et al., 2022). Thus, the
summation test may be viewed as more like a general-
ization test, meaning it is unclear whether the dierence
between modulation and prevention is actually qualitative
or whether there is simply a dierence in degree (Chow et
al., 2022). The implication of these ndings is that Test 2
proposed by Leising et al. may not be very diagnostic in
humans. Researchers who wish to use this test should con-
sider the possibility that individual dierences may exist
in what people learn even among those who have received
identical experimental treatment and that the dierence
between modulators and preventors might be subtle.
The second piece of evidence comes from experi-
ments showing that extinction of a feature is most eective
using procedures that nullify the modulatory properties
of the feature. After feature negative training (A+ AX−),
we found reversal of the inhibitory properties of X after
“no-modulation” training (A+ AX+). Reversal also oc-
curred when the no-modulation contingencies included
a novel cue (M+ MX+). In contrast to the predictions of
the Rescorla-Wagner model (Rescorla & Wagner, 1972),
extinction of the feature (X−) did not have the same eect.
These results suggest that what is learned in feature nega-
tive discriminations is a modulatory causal structure, akin
to occasion setting.
A third piece of evidence is that a preventative cue
(i.e., a conditioned inhibitor) fails to show greater retarda-
tion of subsequent excitatory conditioning compared with
a latently inhibited cue (Lovibond et al., 2023). In a series
of experiments, we compared the properties of a negative
feature X trained in a feature negative discrimination (A+
AX−) with an equivalent cue E presented in compound
with a cue (D), which was separately presented with no
feedback about the outcome (D DE−; see Lee et al., 2022,
for experiments validating this no-feedback procedure).
According to the Rescorla-Wagner model (Rescorla &
Wagner, 1972), X should become a conditioned inhibitor,
whereas E should undergo latent inhibition. Because latent
inhibition is assumed to be dierent to “true” inhibition, X
should show stronger inhibition than E in both a summation
47
Volume 20, 2025
OCCASION SETTING IN HUMANS
and retardation test (Rescorla, 1969). In a retardation test,
the degree to which a cue shows impairment in learning
when it is subsequently paired with the outcome is used
as an index of inhibitory learning. We found that X did
indeed show more transfer than E in a summation test, but
there was no dierence between the speed of acquisition
(X+ vs. E+) in a retardation test (Lovibond et al., 2023).
The similar levels of retardation observed for the negative
feature X and the latently inhibited cue E contradict the
intuition that the inhibitor should be slower given that
it starts at negative associative strength, but can be ex-
plained by assuming that X has become an occasion setter.
Occasion setters should show strong transfer, because as
mentioned the summation test involves generalization
(Chow et al., 2022), and any retardation observed can be
attributed to latent inhibition, since the level of retardation
observed was equivalent to that of the trained latent in-
hibitor. We emphasize that this is not the only explanation
of our results (see Lovibond et al., 2023); nevertheless,
it is worthwhile considering that other processes may be
occurring alongside occasion setting.
The nal piece of evidence comes from experiments
investigating the phenomenon of “protection from extinc-
tion.” Extinction occurs when a cue that was previously
paired with an outcome is no longer paired with that
outcome (A+ / A−), and protection from extinction occurs
when a stimulus is presented with the extinguished cue
and thereby prevents extinction of A from occurring (A+
/ AX−). This procedure is similar to the feature negative
discrimination, except that the trial types are presented
in dierent phases (A+ and then A−). Traditional models
of learning predict that protection of extinction should
occur only if the added cue X is inhibitory, because the
predictions from X (inhibitory) and A (excitatory) will
cancel each other out, resulting in no prediction error and
therefore no associative change to A. In contrast to this
prediction, protection from extinction seems to occur when
the added cue X is neutral (Chow et al., 2024, Experiment
3), excitatory (Lovibond et al., 2000), and even inferred
but not physically present (Chow et al., 2024), suggesting
that there is more than prediction error at play.
In Chow et al. (2024), we found that participants
showed protection from extinction when the cover story
suggested the presence of a hidden cause that could plau-
sibly explain the absent outcome on the extinction trials.
This nding is signicant because it demonstrates that an
additional cue does not need to be present at the onset of
the extinction trials for protection from extinction to occur.
The insight also aligns with well-known ndings demon-
strating the context-specicity of extinction (see Bouton,
1993, 2004). Here, the context can be seen as the hidden
cue that serves to modulate (and therefore protect) the
preexisting belief that the target cue causes the outcome.
If we assume the hypothesis mentioned earlier—occasion
setting is more likely to eventuate when learning of the tar-
get is faster than the feature—then it makes perfect sense
that modulation should occur in extinction, as by denition
the target association must be acquired rst in order to be
extinguished (A+ then AX−, where X is the context).
We concluded from our ndings that people have a
natural tendency to “protect” existing beliefs, leading them
to explain away contradictory information using whatever
stimuli or information is available (Chow et al., 2024). If
this tendency for theory protection is robust (see also Chan
et al., 2024; Spicer et al., 2020, 2022), then occasion set-
ting or modulation will be the dominant form of learning.
Although theory protection may seem suboptimal when an
enduring change has indeed occurred, this tendency helps
to maintain stability of learning in a constantly changing
environment. In other words, rather than starting from
scratch (unlearning or remapping associations), which
would lead to erratic behavior, it may be more ecient for
a learning system to retain existing knowledge and learn
in what specic contexts that knowledge applies. Inter-
estingly, this exact idea is implemented in the Category
Abstraction Learning Model (Schlegelmilch et al., 2022).
To summarize, we agree with Leising et al. (2025)
that modulation, or occasion setting, is more widespread
than assumed. However, we depart in proposing that oc-
casion setting may be the dominant form of learning with
procedures such as the feature negative discrimination. We
have claimed that discriminations that are ambiguous will
produce individual dierences in the content of learning,
and we have advocated for self-report as an additional
test that researchers might want to consider. We reviewed
evidence from our lab demonstrating that the majority
of participants show behavior consistent with occasion
setting, and we presented two explanations for this pref-
erence. First, occasion setting, or modulatory learning,
may be preferable because it is more conservative when
there is limited information regarding the degree to which
the properties of the feature generalize. Second, occasion
setting allows preservation of existing learning, which
may result in more stable and optimal learning over time
in comparison to constantly relearning or remapping as-
sociations. If we are correct, a challenge for the eld will
be to identify the enabling conditions that allow occasion
setting to occur which will allow occasion setting to be in-
tegrated alongside existing mechanisms in formal theories
of learning.
48
COMPARATIVE COGNITION & BEHAVIOR REVIEWS
Lee, Chow, and Lovibond
References
Bonardi, C., Robinson, J., & Jennings, D. (2017). Can
existing associative principles explain occasion set-
ting? Some old ideas and some new data. Behavioural
Processes, 137, 5–18. https://doi.org/10.1016/j.
beproc.2016.07.007
Bouton, M. E. (1993). Context, time, and memory retriev-
al in the interference paradigms of Pavlovian learning.
Psychological Bulletin, 114(1), 80–99. https://doi.
org/10.1037/0033-2909.114.1.80
Bouton, M. E. (2004). Context and behavioral processes
in extinction. Learning & Memory, 11, 485–494. https://
doi.org/10.1101/lm.78804
Chan, Y. Y., Lee, J. C., Fam, J. P., Westbrook, R. F., &
Holmes, N. M. (2024). The role of uncertainty in reg-
ulating associative change. Journal of Experimental
Psychology: Animal Learning and Cognition, 50(2),
77–98. https://doi.org/10.1037/xan0000375
Chow, J. Y. L., Lee, J. C., & Lovibond, P. F. (2022).
Inhibitory summation as a form of generalization.
Journal of Experimental Psychology: Animal Learning
and Cognition, 48(2), 86–104. https://doi.org/10.1037/
xan0000320
Chow, J. Y. L., Lee, J. C., & Lovibond, P. F. (2024). Using
unobserved causes to explain unexpected outcomes:
The eect of existing causal knowledge on protection
from extinction by a hidden cause. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition,
50(7), 1167–1185. https://doi.org/10.1037/xlm0001306
Fraser, K. M., & Holland, P. C. (2019). Occasion setting.
Behavioral Neuroscience, 133(2), 145–175. https://doi.
org/10.1037/bne0000306
Holland, P. C. (1984). Dierential eects of reinforcement
of an inhibitory feature after serial and simultaneous fea-
ture negative discrimination training. Journal of Exper-
imental Psychology: Animal Behavior Processes, 10(4),
461–475. https://doi.org/10.1037/0097-7403.10.4.461
Holland, P. C. (1989). Transfer of negative occasion set-
ting and conditioned inhibition across conditioned and
unconditioned stimuli. Journal of Experimental Psy-
chology: Animal Behavior Processes, 15(4), 311–328.
https://doi.org/10.1037/0097-7403.15.4.311
Holland, P. C. (1991). Acquisition and transfer of occasion
setting in operant feature positive and feature negative
discriminations. Learning and Motivation, 22, 366–387.
https://doi.org/10.1016/0023-9690(91)90002-P
Holland, P. C., & Lamarre, J. (1984). Transfer of inhibition
after serial and simultaneous feature negative discrimi-
nation training. Learning and Motivation, 15, 219–243.
https://doi.org/10.1016/0023-9690(84)90020-1
Lamarre, J., & Holland, P. C. (1985). Acquisition and
transfer of feature-negative discriminations. Bulletin
of the Psychonomic Society, 23(1), 71–74. https://doi.
org/10.3758/BF03329783
Lee, J. C., Le Pelley, M. E., & Lovibond, P. F. (2022).
Nonreactive testing: Evaluating the eect of withhold-
ing feedback in predictive learning. Journal of Exper-
imental Psychology: Animal Learning and Cognition,
48(1), 17–28. https://doi.org/10.1037/xan0000311
Lee, J. C., & Lovibond, P. F. (2021). Individual
dierences in causal structures inferred during
feature negative learning. Quarterly Journal of Ex-
perimental Psychology, 74(1), 150–165. https://doi.
org/10.1177/1747021820959286
Leising, K., Nerz, J., Solorzano-Restrepo, J., & Bond S.
R. (2025). Are you studying occasion setting? A re-
view for inquiring minds. Comparative Cognition and
Behavior Reviews, 20, 5-44. https://doi.org/10.3819/
CCBR.2025.200002
Lovibond, P. F., Chow, J. Y. L., & Lee, J. C. (2023).
Retardation of acquisition after conditioned inhibition
and latent inhibition training in human causal learning.
Journal of Experimental Psychology: Animal Learning
and Cognition, 49(2), 75–86. https://doi.org/10.1037/
xan0000351
Lovibond, P. F., Chow, J. Y. L., Tobler, C., & Lee, J. C.
(2022). Reversal of inhibition by no-modulation train-
ing but not extinction in human causal learning. Journal
of Experimental Psychology: Animal Learning and
Cognition, 48(4), 336–348. https://doi.org/10.1037/
xan0000328
Lovibond, P. F., Davis, N. R., & O’Flaherty, A. S. (2000).
Protection from extinction in human fear conditioning.
Behaviour Research and Therapy, 38, 967–983. https://
doi.org/10.1016/S0005-7967(99)00121-7
49
Volume 20, 2025
OCCASION SETTING IN HUMANS
Lovibond, P. F., & Lee, J. C. (2021). Inhibitory causal struc-
tures inferred through feature negative learning. Quar-
terly Journal of Experimental Psychology, 74(12), 2165–
2181. https://doi.org/10.1177/17470218211022252
Rescorla, R. A. (1969). Pavlovian conditioned inhibition.
Psychological Bulletin, 72(2), 77–84. https://doi.
org/10.1037/h0027760
Rescorla, R. A., & Wagner, A. R. (1972). A theory of
Pavlovian conditioning: Variations in the eectiveness
of reinforcement and non-reinforcement. In A. H. Black
& W. F. Prokasy (Eds.), Classical conditioning II (pp.
64–99). Appleton-Century-Crofts.
Schlegelmilch, R., Wills, A. J., & von Helversen, B.
(2022). A cognitive category-learning model of rule
abstraction, attention learning, and contextual modula-
tion. Psychological Review, 129(6), 1211–1248. https://
doi.org/10.1037/rev0000321
Spicer, S. G., Mitchell, C. J., Wills, A. J., Blake, K. L., &
Jones, P. M. (2022). Theory protection: Do humans pro-
tect existing associative links? Journal of Experimental
Psychology: Animal Learning and Cognition, 48(1),
1–16. https://doi.org/10.1037/xan0000314
Spicer, S. G., Mitchell, C. J., Wills, A. J., & Jones, P. M.
(2020). Theory protection in associative learning: Hu-
mans maintain certain beliefs in a manner that violates
prediction error. Journal of Experimental Psychology:
Animal Learning and Cognition, 46, 151–161. https://
doi.org/10.1037/xan0000225
