Figure 2 - uploaded by Jessica C Lee
Content may be subject to copyright.
Mean outcome prediction ratings (+/-1 SE of the mean) for each causal structure subgroup in the training phase in Experiment 1.

Mean outcome prediction ratings (+/-1 SE of the mean) for each causal structure subgroup in the training phase in Experiment 1.

Source publication
Article
Full-text available
Traditional associative learning theories predict that training with feature negative (A+/AB-) contingencies leads to the feature B acquiring negative associative strength and becoming a conditioned inhibitor (i.e., prevention learning). However, feature negative training can sometimes result in negative occasion setting, where B modulates the effe...

Contexts in source publication

Context 1
... Figure 2 shows mean outcome predictions from the training phase for each subgroup. It is clear from Fig- ure 2 that participants showed differential responding to stimuli predicting the outcome versus those that predict no outcome. ...
Context 2
... 2 shows mean outcome predictions from the training phase for each subgroup. It is clear from Fig- ure 2 that participants showed differential responding to stimuli predicting the outcome versus those that predict no outcome. From the second presentation, there are high overall predictive ratings of the outcome for the A+, C+, and GH+ trials, and low overall predictive ratings of the outcome for the AB-, DE-, and F-trials. ...
Context 3
... Figure 2 shows mean outcome predictions from the training phase for each subgroup. It is clear from Fig- ure 2 that participants showed differential responding to stimuli predicting the outcome versus those that predict no outcome. ...
Context 4
... 2 shows mean outcome predictions from the training phase for each subgroup. It is clear from Fig- ure 2 that participants showed differential responding to stimuli predicting the outcome versus those that predict no outcome. From the second presentation, there are high overall predictive ratings of the outcome for the A+, C+, and GH+ trials, and low overall predictive ratings of the outcome for the AB-, DE-, and F-trials. ...

Citations

... Importantly, our findings can help reconcile apparent discrepancies between rodent and human associative learning strategies. Whereas rodents typically approach feature positive (A:X+ / X-) or negative (A:X-/ X+) discrimination problems using simple, non-hierarchical associative strategies (resorting to hierarchical strategies only when simpler strategies fail to establish accurate predictions), humans spontaneously approach these problems using hierarchical associative strategies 46,47 . Our findings suggest that differences in pre-existing schema, rather than fundamental species differences in learning processes might account for this discrepancy (task-naïve rodent lacking relevant schema, unlike human participants who are well-aware of the possibility of hierarchical relationships). ...
Preprint
Full-text available
Reward cues are often ambiguous; what is good in one context is not necessarily good in another context. To solve this ambiguity, animals form hierarchical associations in which the context acts as a gatekeeper in the retrieval of the appropriate cue-evoked memory, ensuring context-appropriate behavior. These hierarchical associative structures also influence future learning by promoting the formation of new context-dependent associations (leading to the inference of context-dependency for new associations). The orbitofrontal cortex (OFC) and the dorsal hippocampus (DH) are both proposed to encode a "cognitive map" that includes the representation of hierarchical, context-dependent, associations. However the causal role of the OFC and DH in the different functional properties of hierarchical associations remains controversial. Here we used chemogenetic inactivations, in rats, to examine the role of OFC and DH in 1) the contextual regulation of performance, and 2) the contextual learning bias conferred by hierarchical associations. We show that OFC is required for both manifestations of hierarchical associations. In contrast, DH contribution appears limited to the contextual learning bias. This study provides novel insight into the different functional properties of context-dependent hierarchical associations, and establishes the OFC as a critical orchestrator of these different contextual effects.
... As outlined above, models such as Pearce-Hall use prediction error differently, where it is used to update CS-US and CS-no US associations in different ways (Table 2), in addition to being multiplied by a flexible associability parameter, and an additional salience parameter than can be set differently for excitatory and inhibitory trials. This can allow for situations where CS-US is learned quicker than CS-no US, and account for inter-individual differences in the clarity of inhibitory learning (Baetu & Baker, 2012;Chow et al., 2023; J. C. Lee & Lovibond, 2020). For instance, some people may have a greater inhibitory learning rates than others, which may result in a more robust safety memory, which in turn can be used as a predictor of individual variation in trait anxiety and negative affect (Browning et al., 2015;Laing et al., 2019). ...
Preprint
Full-text available
Safety learning involves associating stimuli with the absence of threats, enabling the inhibition of fear and anxiety. Despite growing interest in psychology, psychiatry, and neuroscience research, safety learning lacks a formal consensus definition, leading to inconsistent methodologies and varied results. Conceptualized as a form of inhibitory learning (conditioned inhibition), safety learning can be understood through formal learning theories, such as the Rescorla-Wagner and Pearce-Hall models. This review aims to establish a principled conceptualization of ‘Pavlovian safety learning’, identifying cognitive mechanisms that generate it safety, and boundary conditions that constrain it. Based on these observations, we define Pavlovian safety learning as an active associative process, where surprising threat- omission (safety prediction error) acts as a salient reinforcing event. Instead of producing neutral or non-aversive states, the safety learning process endows stimuli with positive association to ‘safety’. The resulting stimulus-safety memories counteract the influence of fear memories, promoting fear regulation, positive affect, and relief. We critically analyze traditional criteria of conditioned inhibition for their relevance to safety and propose areas for future innovation. A principled concept of Pavlovian safety learning may reduce methodological inconsistencies, stimulate translational research, and facilitate a comprehensive understanding of an indispensable psychological construct.
... In addition, we were not able to include a separate extinguished control cue in the experiment since it would also be subject to the same protective cause as A. Instead, we focused our analysis on comparing subgroups of PROTECTION BY A HIDDEN CAUSE 11 participants who did and did not infer a hidden cause during extinction of cue A, as defined by their responses to the open-ended question at the end of the experiment. These responses were classified by raters who were blind to the test data, as we have done successfully in several previous studies on generalization and inhibition (Lee, Hayes & Lovibond, 2018;Lee & Lovibond, 2021). We then used the difference in predictive and causal ratings to cues C and A as a measure of the extent of extinction in order to compare the subgroups. ...
... The contrast comparing causal ratings of cue A to cue C showed a main effect of cue, using the allergist task (e.g., Lee & Lovibond, 2021;Chow et al., 2022), reflecting potential uncertainty about the causal status of non-reinforced presentations (see also probe test results from Spicer et al., 2022 using a chemical scenario). ...
... In the allergist task used in the present study, foods are often the only type of cue presented throughout the experiment. Previous research from our lab using a feature-negative design (intermixed A+/AB-trials) have shown that people are able to learn inhibitory relationships between food and allergic reactions, and provide highly negative (preventive) ratings to B when asked to judge whether the food by itself causes or prevents an allergic reaction from occurring (e.g., Lee & Lovibond, 2021). ...
Article
Full-text available
People often rely on the covariation between events to infer causality. However, covariation between cues and outcomes may change over time. In the associative learning literature, extinction provides a model to study updating of causal beliefs when a previously established relationship no longer holds. Prediction error theories can explain both extinction and protection from extinction when an inhibitory (preventive) cue is present during extinction. In three experiments using the allergist causal learning task, we found that protection could also be achieved by a hidden cause that was inferred but not physically present, so long as that cause was a plausible preventer of the outcome. We additionally showed complete protection by a physically presented cue that was neutral rather than inhibitory at the outset of extinction. Both findings are difficult to reconcile with dominant prediction error theories. However, they are compatible with the idea of theory protection, where the learner attributes the absence of the outcome to the added cue (when present) or to a hidden cause, and therefore does not need to revise their causal beliefs. Our results suggest that prediction error encourages changes in causal beliefs, but the nature of the change is determined by reasoning processes that incorporate existing knowledge of causal mechanisms and may be biased toward preservation of existing beliefs.
... We also included a Retardation of acquisition 8 brief instruction check which participants were required to pass before proceeding. For further details, please see Lee and Lovibond (2021) and Lee et al. (2022). ...
Article
Full-text available
Inhibitory stimuli are slow to acquire excitatory properties when paired with the outcome in a retardation test. However, this pattern is also seen after simple non-reinforced exposure: latent inhibition. It is commonly assumed that retardation would be stronger for a conditioned inhibitor than for a latent inhibitor, but there is surprisingly little empirical evidence comparing the two in either animals or humans. Thus, retardation after inhibitory training could in principle be attributable entirely to latent inhibition. We directly compared the speed of excitatory acquisition after conditioned inhibition and matched latent inhibition training in human causal learning. Conditioned inhibition training produced stronger transfer in a summation test, but the two conditions did not differ substantially in a retardation test. We offer two explanations for this dissociation. One is that learned predictiveness attenuated the latent inhibition that otherwise would have occurred during conditioned inhibition training, so that retardation in that condition was primarily due to inhibition. The second explanation is that inhibitory learning in these experiments was hierarchical in nature, similar to negative occasion-setting. By this account, the conditioned inhibitor was able to negatively modulate the test excitor in a summation test, but was no more retarded than a latent inhibitor in its ability to form a direct association with the outcome.
... Recent research on humans' causal judgments has called into question the sharp boundary between conditioned inhibition and negative occasion setting, at least on procedural grounds. Lee and Lovibond (2021; see also Glautier & Brudan, 2019) found that the self-reported properties of the modulatory function exerted by a target stimulus that underwent feature-negative discrimination training (A+/AX−) varied considerably between participants. This was the case even though the degree of discrimination at the end of the training protocol was roughly equivalent for participants regardless of their type of self-report. ...
... These authors evaluated the patterns of connections developed across different simulations for the same task. Interestingly, similar to the results reported by Lee and Lovibond (2021; see also , the neural networks varied in their effective solutions to the discrimination task. In short, the participants in the study of Lee and Lovibond (2021) solved the feature-negative discrimination task via different inferential structures, while the artificial networks in Castiello et al. (2021) study solved the negative patterning via different connectivity patterns. ...
... Interestingly, similar to the results reported by Lee and Lovibond (2021; see also , the neural networks varied in their effective solutions to the discrimination task. In short, the participants in the study of Lee and Lovibond (2021) solved the feature-negative discrimination task via different inferential structures, while the artificial networks in Castiello et al. (2021) study solved the negative patterning via different connectivity patterns. ...
Article
Full-text available
Inhibition can be defined as a phenomenon in which an agent prevents or suppresses a behavioral state that would otherwise occur. Associative learning studies have extensively examined how experiences shape the acquisition of inhibitory behavioral tendencies across many species and situations. Associative inhibitory phenomena can be studied at various levels of analysis. One could focus on the trajectory of behavioral change involved in learning from negative statistical associations between discrete events (inhibitory learning). Alternatively, one could be interested in the effects of accumulated experience with those negative associations (conditioned inhibition). One could rather be interested in how organisms implement what they learn through experiences involving negative associations (response inhibition). Yet, one could inquire into how the capacity of learning negative associations and performing accordingly varies between individuals and along time for the same individual (inhibitory control). This article presents a tentative taxonomy addressing different levels of analysis of associative inhibitory phenomena by using different terms for each. In addition, recent evidence and certain unresolved issues at each level are thoroughly scrutinized and contrasted with prior findings. The empirical and theoretical advances made by modeling inhibition as an associative learning phenomenon have provided scaffolds for the current knowledge and emerging accounts of the topic. Some of those emerging accounts have the potential to bridge different levels of analysis and foster “cross-pollination” of ideas among broad fields beyond associative learning.
... As in Lee and Lovibond (2021), the experimental stimuli (A-L) were selected randomly from a pool of 16 food pictures that included a verbal label (e.g., "chicken"). The allergic reaction outcome consisted of the text "Allergic Reaction!" accompanied by a sad face emoticon. ...
... The primary exclusion criteria were those used in Lee and Lovibond (2021) and . Specifically, participants were excluded if they reported having written anything down during the experiment (n=26) or if they failed the acquisition criterion which was an average rating > 75 for the stimuli that predicted the outcome (A+, C+, GH+) and an average rating < 25 for the stimuli that predicted no outcome (AB-, DE-, F-) in the last block of Phase 1 (n=26). ...
Article
Full-text available
One of the many strengths of the Rescorla and Wagner (1972) model is that it accounts for both excitatory and inhibitory learning using a single error-correction mechanism. However, it makes the counterintuitive prediction that nonreinforced presentations of an inhibitory stimulus will lead to extinction of its inhibitory properties. Zimmer-Hart and Rescorla (1974) provided the first of several animal conditioning studies that contradicted this prediction. However, the human data are more mixed. Accordingly, we set out to test whether extinction of an inhibitor occurs in human causal learning after simultaneous feature negative training with a conventional unidirectional outcome. In 2 experiments with substantial sample sizes, we found no evidence of extinction after presentations of the inhibitory stimulus alone in either a summation test or causal ratings. By contrast, 2 "no-modulation" procedures that contradicted the original training contingencies successfully reversed inhibition. These results did not differ substantially as a function of participants' self-reported causal structures (configural/modulation/prevention). We hypothesize that inhibitory learning may be intrinsically modulatory, analogous to negative occasion-setting, even with simultaneous training. This hypothesis would explain why inhibition is reversed by manipulations that contradict modulation but not by simple extinction, as well as other properties of inhibitory learning such as imperfect transfer to another excitor. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... Participants provided online consent for their participation. The procedure used in the experiments reported here closely followed the procedure used in Lee and Lovibond (2021) and the simultaneous procedure in Lovibond and Lee (2021). ...
... In the final phase of the study, participants were asked to assess the role of Cue B in a three-alternative forced-choice (3AFC) question. An image of Cue B was displayed at the top of the screen, and participants were asked to select the option that best described what they thought about the role of Cue B. As in Lee and Lovibond (2021) and Lovibond and Lee (2021), the configural option was, "It is hard to know the exact role of individual objects such as this one. I concentrated on remembering which combinations of objects caused an allergic reaction and worked from there"; the modulation option was, "It prevented allergic reactions caused by specific objects"; and the prevention option was, "It prevented allergic reactions in general." ...
... Participants' data were excluded from data analysis based on the same two exclusion criteria used in Lee and Lovibond (2021) and Lovibond and Lee (2021). Data were excluded if participants reported writing down information during the task (n = 9), or if they failed to meet the training criterion (n = 9), which was average training ratings greater than 75 for predictive cues and less than 25 for nonpredictive cues. ...
Article
Full-text available
Inhibitory learning after feature negative training (A+/AB-) is typically measured by combining the Feature B with a separately trained excitor (e.g., C) in a summation test. Reduced responding to C is taken as evidence that B has properties directly opposite to those of C. However, in human causal learning, transfer of B's inhibitory properties to another excitor is modest and depends on individual differences in inferred causal structure. Here we ask whether instead of opposing processes, a summation test might instead be thought of in terms of generalization. Using an allergist task, we tested whether inhibitory transfer would be influenced by similarity. We found that transfer was greater when the test stimuli were from the same semantic category as the training stimuli (Experiments 1 and 2) and when the test excitor had previously been associated with the same outcome (Experiment 3). We also found that the similarity effect applied across all self-reported causal structures. We conclude it may be more helpful to consider transfer of inhibition as a form of conceptual generalization rather than the arithmetic summation of opposing processes. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... This no-feedback testing procedure is ubiquitous in predictive learning experiments. In the simplest case, participants are simply informed that they will not receive feedback for the upcoming test phase (e.g., Lee & Lovibond, 2021;Lee et al., 2019;Uengoer et al., 2020), or are forewarned that this will occur for the final series of trials (e.g., Shanks & Darby, 1998;Williams et al., 1994). In fear-conditioning studies, the experimenter can unplug the shock electrodes to show participants that shocks are not possible prior to the test phase; participants are then asked to make hypothetical expectancy ratings of shock (Lee et al., 2018;2019;Wong & Lovibond, 2017). ...
... For example, participants may be told that outcome data for particular trials have been "lost" (e.g., Le Pelley & McLaren, 2001;2004), that "no information is available" (e.g., Collins & Shanks, 2006;George & Kruschke, 2012), or they may be presented with question marks "???" (e.g., Luque et al., 2017;Wills et al., 2007;2014). In contrast, other studies do not present any alternative feedback and participants simply progress to the next trial after making their prediction response (e.g., Lee & Lovibond, 2021;Lee et al., 2019;Livesey et al., 2019;Lee & Livesey, 2012). Studies also vary in the degree to which the missing-outcome feedback is integrated into the cover story. ...
... Alternatively, participants may be told that the missing outcome feedback is part of the experiment (e.g., Shanks & Darby, 1998). Most commonly, participants are given no additional information and simply told to expect feedback to be withheld in upcoming trials (e.g., Lee & Lovibond, 2021;Williams et al., 1994). ...
Article
Full-text available
Learning of cue-outcome relationships in associative learning experiments is often assessed by presenting cues without feedback about the outcome and informing participants to expect no outcomes to occur. The rationale is that this "no-feedback" testing procedure prevents new learning during testing that might contaminate the later test trials. We tested this assumption in 4 predictive learning experiments where participants were tasked with learning which foods (cues) were causing allergic reactions (the outcome) in a fictitious patient. We found that withholding feedback in a block of trials had no effect on causal ratings (Experiments 1 and 2), but it led to regression toward intermediate ratings when the missing feedback was embedded in the causal scenario and information about the outcome replaced by a "?" (Experiment 3). A factorial experiment manipulating cover story and feedback revealed that the regression-to-baseline effect was primarily driven by presentation of the "?" feedback (Experiment 4). We conclude that the procedure of testing without feedback, used widely in studies of human cognition, is an appropriate way of assessing learning, as long as the missing data are attributed to the experimenter and the absence of feedback is not highlighted in a way that induces uncertainty. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
... Although developed to account for animal conditioning data, the Rescorla-Wagner model has also been suggested to provide a good account of causal learning in humans, where excitatory associative strength corresponds to the belief that the stimulus causes the outcome, and inhibitory associative strength corresponds to the belief that the stimulus prevents the outcome (Dickinson et al., 1984). We have recently been interested in exploring individual differences in the beliefs acquired by human participants in a causal learning task involving a feature negative (A+/ AB−) discrimination (Lee & Lovibond, 2021; see also Glautier & Brudan, 2019). We have found that, contrary to the Rescorla-Wagner model, relatively few participants express what they have learned in terms of a causal structure in which feature B directly prevents the outcome. ...
... Our previous findings (Lee & Lovibond, 2021) were obtained with a simultaneous AB compound (A and B were presented at the same time). However, the modulatory participants show, in both their causal structure beliefs and transfer, a type of learning that is more commonly observed in animal conditioning with a serial procedure in which the feature B precedes A in time, referred to as negative occasion-setting. ...
... It is clear that we do not yet have a consistent empirical picture of inhibitory learning in humans. Our own work suggests that probing the causal beliefs acquired by participants might provide insight into how they are learning, including the possibility that group-level results are an amalgam of distinct patterns displayed by subgroups of participants (Lee et al., 2018;Lee & Lovibond, 2021). In the present project, we carried out three experiments to explore the impact of serial feature negative training on self-reported causal structure as well as transfer in a summation test. ...
Article
Full-text available
We have previously reported that human participants trained with a simultaneous feature negative discrimination (intermixed A+ / AB- trials) show only modest transfer of inhibitory properties of the feature B to a separately trained excitor in a summation test (Lee & Lovibond, 2021). Self-reported causal structure suggested that many participants learned that the effect of the feature B was somewhat specific to the excitor it had been trained with (modulation), rather than learning that the feature prevented the outcome (prevention). This pattern is reminiscent of the distinction between negative occasion-setting and conditioned inhibition in the animal conditioning literature. However, in animals, occasion-setting is more commonly seen with a serial procedure in which the feature (B) precedes the training excitor (A). Accordingly, we ran three experiments to compare serial with simultaneous training in an allergist causal judgment task. Transfer in a summation test was stronger to a previously modulated test excitor compared to a simple excitor after both simultaneous and serial training. There was a numerical trend towards a larger effect in the serial group, but it failed to reach significance and the Bayes Factor indicated support for the null. Serial training had no differential effect on self-reported causal structure, and did not significantly reduce overall transfer. After both simultaneous and serial training, transfer was strongest in participants who reported a prevention structure, replicating and extending our previous results to a previously modulated excitor. These results suggest that serial feature negative training does not promote a qualitatively different inhibitory causal structure compared to simultaneous training in humans.
... As a result, insights on safety learning in contemporary research may be limited by factors that are less theory-driven, and more the product of pragmatic methodological concerns. By contrast, 'Pavlovian conditioned inhibition' (also referred to as 'feature-negative discrimination', (Holland and Lamarre, 1984;Lee and Lovibond, 2020) is a canonical inhibitory learning paradigm, representing a predetermined effort to model associations between neutral cues and the omission of salient outcomes (Choraźyna, 1962;Konorski, 1967;Konorski and Miller, 1937;Miller and Konorski, 1969;Pavlov, 1927). The following sections outline the procedure's experimental parameters. ...
Article
Safety learning occurs when an otherwise neutral stimulus comes to signal the absence of threat, allowing organisms to use safety information to inhibit fear and anxiety in nonthreatening environments. Although it continues to emerge as a topic of relevance in biological and clinical psychology, safety learning remains inconsistently defined and under-researched. Here, we analyse the Pavlovian conditioned inhibition paradigm and its application to the study of safety learning in humans. We discuss existing studies; address outstanding theoretical considerations; and identify prospects for its further application. Though Pavlovian conditioned inhibition presents a theoretically sound model of safety learning, it has been investigated infrequently, with decade-long interims between some studies, and notable methodological variability. Consequently, we argue that the full potential of conditioned inhibition as a model for human safety learning remains untapped, and propose that it could be revisited as a framework for addressing timely questions in the behavioural and clinical sciences.