Predictive and diagnostic learning within causal models: Asymmetries in cue competition.

Several researchers have recently claimed that higher order types of learning, such as categorization and causal induction, can be reduced to lower order associative learning. These claims are based in part on reports of cue competition in higher order learning, apparently analogous to blocking in classical conditioning. Three experiments are reported in which Ss had to learn to respond on the basis of cues that were defined either as possible causes of a common effect (predictive learning) or as possible effects of a common cause (diagnostic learning). The results indicate that diagnostic and predictive reasoning, far from being identical as predicted by associationistic models, are not even symmetrical. Although cue competition occurs among multiple possible causes during predictive learning, multiple possible effects need not compete during diagnostic learning. The results favor a causal-model theory. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The special status of actions in causal reasoning in rats.

A. P. Blaisdell, K. Sawa, K. J. Leising, and M. R. Waldmann (2006) reported evidence for causal reasoning in rats. After learning through Pavlovian observation that Event A (a light) was a common cause of Events X (an auditory stimulus) and F (food), rats predicted F in the test phase when they observed Event X as a cue but not when they generated X by a lever press. Whereas associative accounts predict associations between X and F regardless of whether X is observed or generated by an action, causal-model theory predicts that the intervention at test should lead to discounting of A, the regular cause of X. The authors report further tests of causal-model theory. One key prediction is that full discounting should be observed only when the alternative cause is viewed as deterministic and independent of other events, 2 hallmark features of actions but not necessarily of arbitrary events. Consequently, the authors observed discounting with only interventions but not other observable events (Experiments 1 and 2). Moreover, rats were capable of flexibly switching between observational and interventional predictions (Experiment 3). Finally, discounting occurred on the very first test trial (Meta-Analysis). These results confirm causal-model theory but refute associative accounts. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A Causal-Model Theory of Conceptual Representation and Categorization.

This article presents a theory of categorization that accounts for the effects of causal knowledge that relates the features of categories. According to causal-model theory, people explicitly represent the probabilistic causal mechanisms that link category features and classify objects by evaluating whether they were likely to have been generated by those mechanisms. In 3 experiments, participants were taught causal knowledge that related the features of a novel category. Causal-model theory provided a good quantitative account of the effect of this knowledge on the importance of both individual features and interfeature correlations to classification. By enabling precise model fits and interpretable parameter estimates, causal-model theory helps place the theory-based approach to conceptual representation on equal footing with the well-known similarity-based approaches. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Comparing elemental and configural associative theories in human causal learning: A case for attention.

In two causal learning experiments with human participants, the authors compared various associative theories that assumed either elemental (unique cue, modified unique cue, replaced elements model, and Harris' model) or configural processing of stimuli (Pearce's theory and a modification of it). The authors used modified patterning problems initially suggested by Redhead and Pearce (1995). Predictions for all theories were generated by computer simulations. Both configural theories and the unique cue approach failed to account for the observations. The replaced elements model was able to account for part of the data, but only if the replacement parameters could vary across discrimination problems. The Harris model and the modified unique cue approach, assuming that the salience of stimuli decreases with an increasing number of stimuli in a compound, successfully accounted for all of our data. This success implies that attentional factors should be explicitly taken into account in associative learning theory. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Retrospective revaluation of causal judgments under positive and negative contingencies.

Retrospective revaluation of causal judgments was investigated in a 2-stage procedure. In the 1st stage, compounds of 2 cues were associated with the outcome, whereas in the 2nd stage, a cue from each compound was trained by itself. Associating this cue with the outcome in the 2nd stage had no detectable effect on the causal rating of the other cue from the compound, whereas presenting it without the outcome enhanced the causal rating of the other cue. The retrospective revaluation of the causal rating of these productive cues and also of preventative cues depended on consistent pairing of the cues during compound training, suggesting a role for within-compound associations. These results favor associative accounts of retrospective revaluation that use separate excitatory and inhibitory learning processes rather than a general error-correcting learning algorithm. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A model of probabilistic category learning.

A new connectionist model (named RASHNL) accounts for many "irrational" phenomena found in nonmetric multiple-cue probability learning, wherein people learn to utilize a number of discrete-valued cues that are partially valid indicators of categorical outcomes. Phenomena accounted for include cue competition, effects of cue salience, utilization of configural information, decreased learning when information is introduced after a delay, and effects of base rates. Exps 1 and 2 replicate previous experiments on cue competition and cue salience, and fits of the model provide parameter values for making qualitatively correct predictions for many other situations. The model also makes 2 new predictions, confirmed in Exps 3 and 4. The model formalizes 3 explanatory principles: rapidly shifting attention with learned shifts, decreasing learning rates, and graded similarity in exemplar representation. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Associative History Affects the Associative Change Undergone by Both Presented and Absent Cues in Human Causal Learning.

R. A. Rescorla (2000) noted that a number of influential theories of associative learning do not take the associative history of cues (i.e., the prior training that they have received) into account when calculating the associative change undergone by those cues. The authors tested this assumption in a human causal learning paradigm and found associative history to be an important determinant of the learning undergone by cues that are presented on a trial. Moreover, associative history was also found to influence the amount of retrospective revaluation undergone by absent cues. These findings conflict with models of causal learning in which the associative change undergone by an element of a cue compound is governed by a summed error term (e.g., R. A. Rescorla & A. R. Wagner, 1972). (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Learning in a changing environment.

Multiple cue probability learning studies have typically focused on stationary environments. We present 3 experiments investigating learning in changing environments. A fine-grained analysis of the learning dynamics shows that participants were responsive to both abrupt and gradual changes in cue–outcome relations. We found no evidence that participants adapted to these types of change in qualitatively different ways. Also, in contrast to earlier claims that these tasks are learned implicitly, participants showed good insight into what they learned. By fitting formal learning models, we investigated whether participants learned global functional relationships or made localized predictions from similar experienced exemplars. Both a local (the associative learning model) and a global learning model (the Bayesian linear filter) fitted the data of the first 2 experiments. However, the results of Experiment 3, which was specifically designed to discriminate between local and global learning models, provided more support for global learning models. Finally, we present a novel model to account for the cue competition effects found in previous research and displayed by some of our participants. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The experimental task influences cue competition in human causal learning.

Participants were shown A+ and C- trials followed by AB+ and CD+ trials. These trials were embedded in a causal learning task in which participants had to learn either the relationship between different foods and allergic reactions or the relationship between different stocks and an increase in the stock market index. The authors orthogonally varied the manner in which the different cues were presented to participants during training. Cue competition was related to the causal learning scenario but not to the manner in which the different cues were presented. These results question claims of a human bias toward configural processing that were based on difficulties in finding cue competition in some previous causal learning experiments. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The hide-and-seek of retrospective revaluation: Recovery from blocking is context dependent in human causal learning.

In a typical blocking procedure, pairings of a compound consisting of 2 stimuli, A and X, with the outcome are preceded by pairings of only A with the outcome (i.e., A+ then AX+). This procedure is known to diminish responding to the target cue (X) relative to a control group that does not receive the preceding training with blocking cue A. We report 2 experiments that investigated the effect of extinguishing a blocking cue on responding to the target cue in a human causal learning paradigm (i.e., A+ and AX+ training followed by A– training). The results indicate that extinguishing a blocking cue increases conditioned responding to the target cue. Moreover, this increase appears to be context dependent, such that increased responding to the target is limited to the context in which extinction of the blocking cue took place. We discuss these findings in the light of associative and propositional learning theories. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Cues trained apart compete for behavioral control in rats: Convergence with the associative interference literature.

Contemporary theories of associative learning require cues be trained in compound for cue competition (interference) to occur. That is, Cues A and X should compete for behavioral control only if training consists of AX-outcome (O) trials and not if each cue is separately paired with O (i.e., X–O and A–O). Research with humans challenges this view by showing that A–O trials interpolated between training and testing of a X–O association impair responding to X (i.e., retroactive interference). In six conditioned suppression studies with rats, the authors demonstrate that two cues trained apart can each interfere with the potential of the other to predict the outcome. The authors conclude that this type of interference (a) reflects a failure to retrieve the target association due to priming at test of the interfering association and (b) is attenuated if the outcome is of high biological significance. These findings parallel previous reports in verbal learning research and suggest that a similar associative structure underlies some types of associations in nonverbal subjects. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Effects of trial order on contingency judgments: A comparison of associative and probabilistic contrast accounts.

Four experiments examined trial sequencing effects in human contingency judgment. In Experiments 1–3, ratings of contingency between a target cue and outcome were affected by the presentation order of a series of trials distributed in 2 distinct blocks and showed a recency bias. Experiment 4 replicated this effect when the trials were partly intermixed. These recency effects are predicted by an associative learning model that computes associative strengths trial by trial and incorporates configural coding of cues but are problematic for probabilistic contrast accounts, which currently have no provision in the contingency computation for the differential weighting of trials as a function of their order of presentation. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Associative and rule-based accounts of cue interaction in contingency judgment.

This study examined how people detect and assess the strength of contingent relationships between pairs of events. Some researchers have suggested that contingency learning is analogous to classical conditioning and that contingency judgment is based on psychological associations formed during learning. Others have rejected the associative account in favor of a rule-based account involving higher level statistical and causal reasoning. The results of 4 experiments in which college-student participants performed a simulated medical-judgment task, showed that the rule-based account does not provide a sufficient explanation of cue-interaction effects in contingency learning and judgment. Elements of the associative account are needed to explain the entire range of contingency judgment phenomena. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The Outcome Specificity of Learned Predictiveness Effects: Parallels Between Human Causal Learning and Animal Conditioning.

Two experiments examined the outcome specificity of a learned predictiveness effect in human causal learning. Experiment 1 indicated that prior experience of a cue-outcome relation modulates learning about that cue with respect to a different outcome from the same affective class but not with respect to an outcome from a different affective class. Experiment 2 ruled out an interpretation of this effect in terms of context specificity. These results indicate that learned predictiveness effects in human causal learning index an associability that is specific to a particular class of outcomes. Moreover, they mirror demonstrations of the reinforcer specificity of analogous effects in animal conditioning, supporting the suggestion that, under some circumstances, human causal learning and animal conditioning reflect the operation of common associative mechanisms. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Stimulus Generalization in Two Associative Learning Processes.

Recent studies involving nonlinear discrimination problems suggest that stimuli in human associative learning are represented configurally with nairow generalization, such that presentation of stimuli that are even slightly dissimilar to stored configurations weakly activate these configurations. The authors note that another well-known set of findings in human associative learning, cue-interaction phenomena, suggest relatively broad generalization. Three experiments show that current models of human associative learning, which try to model both nonlinear discrimination and cue interaction as the result of 1 process, fail because they cannot simultaneously account for narrow and broad generalization. Results suggest that human associative learning involves (a) an exemplar-based process with configural stimulus representation and narrow generalization and (b) an adaptive learning process characterized by broad generalization and cue interaction. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Recovery from blocking between outcomes.

Contemporary associative learning research largely focuses on cue competition phenomena that occur when 2 cues are paired with a common outcome. Little research has been conducted to investigate similar phenomena occurring when a single cue is trained with 2 outcomes. Three conditioned lick suppression experiments with rats assessed whether treatments known to alleviate blocking between cues would also attenuate blocking between outcomes. In Experiment 1, conditioned responding recovered from blocking between outcomes when a long retention interval was interposed between training and testing. Experiment 2 obtained recovery from blocking between outcomes when the blocking outcome was extinguished after the blocking treatment. In Experiment 3, a recovery from blocking between outcomes occurred when a reminder stimulus was presented in a novel context prior to testing. Collectively, these studies demonstrate that blocking of outcomes, like blocking of cues, appears to be caused by a deficit in the expression of an acquired association. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Simultaneous activation of the representations of absent cues results in the formation of an excitatory association between them.

It is a common assumption of associative theories of learning that no change in the strength of an associative connection between 2 cues is possible in the absence of those cues. However, recently suggested modifications to associative theory (A. Dickinson & J., Burke 1996) have questioned this assumption by arguing that if the representations of 2 cues are simultaneously retrieved from memory, an association will be formed between them even though the cues themselves are not present. A flavor preference procedure was used to find evidence for such associations. In 3 experiments a novel excitatory connection was formed between the representations of peppermint and sucrose in their absence. This suggests that the assumption that cues cannot undergo a change of associative strength in their absence should be abandoned. The tension between the current results and accounts of mediated conditioning is discussed, and some suggestions regarding the difference between the 2 procedures are proposed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Spatial association learning by rufous hummingbirds (Selasphorus rufus): Effects of relative spacing among stimuli.

Gestalt theory predicts that when cues are spatially separated from response locations, associative learning is faster when distance between cue–response location pairs is increased. This prediction was tested with 20 rufous hummingbirds that learned to select rewarding feeders signaled by a spatially separated light cue in 4 treatments in which distance between cues and feeders and between cue–feeder pairs was varied. As has been shown for other animals, the hummingbirds learned more slowly when the distance between cues and feeder was increased, and as predicted by Gestalt theory, they learned faster at a given distance when distance between cue–feeder pairs was increased. This result suggests that spatial association is influenced by the proximity of other stimuli in the visual field. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Dissociations between expectancy and performance in simple and two-choice reaction-time tasks: A test of associative and nonassociative explanations.

Perruchet, Cleeremans, and Destrebecqz (2006) reported a striking dissociation between trends in the conscious expectancy of an event and the speed of a response that is cued by that event. They argued that this indicates the operation of independent processes in human associative learning. However, there remains a strong possibility that this dissociation is not a consequence of associative learning and is instead caused by changes in vigilance or sensitivity based on the recency of events on previous trials. Three experiments tested this possibility with versions of a cued reaction time task in which trends in performance could not be explained by these nonassociative factors. Experiment 1 introduced a dual-response version of the task, in which response-related vigilance should be held relatively constant, and Experiments 2 and 3 used a differential conditioning procedure to separate the influence of recent response cue presentation from the recent associative history of the trial events. In all experiments, similar trends in reaction time were evident, suggesting a genuine influence of associative learning on response performance. Experiment 3 demonstrated that the associative contribution to these trends was not caused by commensurate changes in expectancy of the response cue. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Reasoning rats: Forward blocking in Pavlovian animal conditioning is sensitive to constraints of causal inference.

Forward blocking is one of the best-documented phenomena in Pavlovian animal conditioning. According to contemporary associative learning theories, forward blocking arises directly from the hardwired basic learning rules that govern the acquisition or expression of associations. Contrary to this view, here the authors demonstrate that blocking in rats is flexible and sensitive to constraints of causal inference, such as violation of additivity and ceiling considerations. This suggests that complex cognitive processes akin to causal inferential reasoning are involved in a well-established Pavlovian animal conditioning phenomenon commonly attributed to the operation of basic associative processes. (PsycINFO Database Record (c) 2010 APA, all rights reserved)