Analogical inference: the role of awareness in abstract learning

The acquisition and flexible expression of complex relations is often attributed to declarative memory processes. The extent to which such tasks may be done implicitly has not been sufficiently explored. We report that analogical or transfer processes may be accomplished implicitly. Our analogy task requires acquisition of a transverse patterning set, and then tests for transfer on an unrelated set. Participants learn the relations A>B (given a choice between A and B choose A) and B>C and the unrelated set X>Y and Y>Z. Only the experimental group was trained on the transverse pair C>A. At test all trials are unreinforced: A?B, B?C, A?C, X?Y, Y?Z, X?Z. Analogy was observed when the experimental group chose Z>X at greater frequency than controls who uniformly chose X>Z. Analogy occurred with or without awareness of the transfer. The capacity to transfer relations to an analogous set demonstrates a level of flexibility and abstraction not generally thought to be possible for implicit processes.     

Transforming valences through transitive inference: How are faces emotionally dissonant?

Information that is emotionally incongruous with self-concepts can produce feelings of unease. This implies that embedding incongruous information in newly formed relational structures would have little effect on their previous emotive properties. Alternatively, Relational Frame Theory highlights the importance of contextualized stimulus-stimulus relations, where the structure of a relational series is key in determining the function of its elements. To see whether series membership can mitigate ‘dissonance’ when a salient element is employed, the present investigation trained and tested a seven-term relational series (X>A>B>C>D>E>Y) using blurred faces as stimuli. Specifically, Stimuli X, A, B, D, E and Y were blurred unfamiliar faces and Stimulus C constituted of the participant’s own blurred face, assumed to be more salient than the former. To assess how the valences of the related stimuli were transformed by relational series membership, self-report ratings and electroencephalographic (EEG) recordings were collected before and after training the X>A>B>C>D>E>Y series. These pre vs. post contrasts revealed that, for unfamiliar faces, stimulus valence transformed as a function of relational structure. Conversely, the lack of difference in pre vs. post contrasts of Stimulus C, which maintained a high valence, suggest that relational series membership may not suffice to mitigate emotionally dissonant information.     

Derived manding in children with autism: synthesizing Skinner's verbal behavior with relational frame theory

Mand functions for two stimuli (A1 and A2) were trained for 3 children with autism and were then incorporated into two related conditional discriminations (A1-B1/A2 -B2 and B1-C1/B2-C2). Tests were conducted to probe for a derived transfer of mand response functions from A1 and A2 to C1 and C2, respectively. When 1 participant failed to demonstrate derived transfer of mand response functions, transfer training using exemplars was conducted. When participants had demonstrated derived transfer of mand functions, the X1 and X2 tokens that were employed as reinforcers for mand responses were incorporated into two conditional discriminations (X1-Y1/X2-Y2 and Y1-Z1/Y2-Z2). Tests were conducted for derived transfer of reinforcing functions. Finally, tests were conducted to determine if the participants would demonstrate derived manding for the derived reinforcers (present C1 and C2 to mand for Z1 and Z2, respectively). Derived transfer of functions was observed when the sequence of training and testing was reversed (i.e., training and testing reinforcing functions before mand response functions) and when only minimal instructions were provided.     

Transfer of inhibition after serial feature negative discrimination training

Three experiments examined the transfer of inhibition in Pavlovian serial feature negative (A+, X→ A−) discriminations in a conditioned suppression situation with rat subjects. Consistent with our previous report (P. C. Holland & J. Lamarre, Learning and Motivation, 15, 219–243), the feature (X) showed little or no ability to inhibit suppression to another conditioned excitor (B) that had been consistently reinforced. Nor was substantial transfer observed to excitors that had been partially reinforced, or conditioned, extinguished, and then reconditioned. However, X readily inhibited suppression to an excitor that had been trained within another serial feature negative discrimination (B+, Y → B−). These latter data are difficult to reconcile with our previous proposal that inhibitors established with serial procedures act on particular conditioned stimulus-unconditioned stimulus associations. We suggest that serial feature negative discriminations endow both the feature (X) and the excitor (A) with special properties.     

The role of similarity in human associative learning

In 2 experiments, humans received sequences of patterns that were similar (AX-->BX, AY-->BY, AZ-->BZ) or dissimilar (CX-->DY, CY-->DZ, CZ-->DX). The patterns were portrayed as bugs that could be eliminated with 2 insecticide sprays (red or blue). Either spray eliminated bugs with Features A and C, and participants learned by trial and error to use one spray (e.g., red) to eliminate bugs with Feature B and the other spray (e.g., blue) to eliminate those with Feature D. In Experiment 1, participants' spray choice for bugs with Feature A came to match that used to eliminate bugs with Feature B, but there was no such associative transfer between Features C and D. That is, similarity promoted associative transfer of responding between paired patterns when the features used to manipulate similarity (i.e., X, Y, and Z) were irrelevant. In Experiment 2, in which X, Y, and Z were relevant to the solution of configural discrimination, similarity hindered such associative transfer. These results complement those found in pigeons (R. A. Rescorla & D. J. Gillan, 1980) and indicate that similarity should not be accorded independent status as a principle of associative learning.     

Retrospective revaluation effects following serial compound training and target extinction

Using a conditioned suppression task, two experiments examined retrospective revaluation effects after serial compound training in a release from overshadowing design. In Experiment 1, serial X → A+ training produced suppression to target A, which was enhanced when preceded by feature X, whereas X by itself elicited no suppression. Subsequent A− presentations extinguished responding to A, but had no effect on either responding to X → A or X alone. However, the addition of A− trials did enhance the ability of feature X to elicit suppression to a novel target, B, suggesting retrospective revaluation of X’s properties. Experiment 2 showed that the enhanced transfer effect, observed in Experiment 1, was independent of the training history of the target (B− or Y → B+/B−). Together, these results suggest that feature X did not retrospectively acquire excitatory strength or occasion setting power, but rather a generalized ability to increase responding to any other cue.     

Acquired equivalence between stimuli trained in the same context

The role of context was examined in human acquired equivalence. Participants were trained on two conditional discriminations. In the first conditional discrimination, if sample A1 was presented, choice of comparison B1, but not B2, was correct, and if sample A2 was presented, choice of comparison B2, but not B1, was correct. In the second conditional discrimination, if sample X1 was presented, choice of comparison Y1, but not Y2, was correct, and if sample X2 was presented, choice of comparison Y2, but not Y1, was correct. In each conditional discrimination, one of the conditional relations was trained in context 1 (e.g., A1 → B1 and X1 → Y1) and the other was trained in context 2 (i.e., A2 → B2 and X2 → Y2). On test trials, when conditional stimuli from the two conditional discriminations were interchanged (e.g., sample A1 was presented with comparisons Y1 and Y2) and were presented in a neutral context, positive transfer resulted. That is, in the absence of the training context, stimuli that shared a common context on different trials in training came to be treated as equivalent.     

An analysis of generalized contextual control of conditional discriminations

This research asked whether performance engendered by contextual control procedures would generalize to novel matching-to-sample stimulus arrangements. Two studies were conducted with young adult participants. In Study 1, participants first were trained to perform the contextually controlled conditional discrimination, X-AB, where the sample-comparison relations A1B1 and A2B2 were reinforced in the presence of contextual stimulus X1, but the relations A1B2 and A2B1 were reinforced in the presence of X2. Then, a new conditional discrimination, CD, was established via an unreinforced-conditional-selection procedure. Next, participants were tested for X-CD contextual control performance. Participants selected the originally established CD relations in the presence of X1, but the opposite relations in the presence of X2. Next, an additional conditional relation, EF, was established. Then, participants received trials consisting of entirely novel contextual stimuli, Z1 and Z2, and EF samples and comparisons. Selections were consistent with contextual control; that is, participants selected the originally established EF relations in the presence of one of the novel contextual Z stimuli, but selected the opposite EF relations in the presence of the other contextual Z stimulus. Study 2 systematically replicated these results with naive participants and demonstrated the necessity of first establishing a conditional discrimination prior to tests for generalized contextual control. The findings are discussed in terms of unreinforced conditional selection, stimulus classes, and new ways in which contextual control performances can emerge.     

Simultaneous and sequential Feature Negative discriminations: Elemental learning and occasion setting in human Pavlovian conditioning

Using a conditioned suppression task, we investigated simultaneous (XA−/A+) vs. sequential (X → A−/A+) Feature Negative (FN) discrimination learning in humans. We expected the simultaneous discrimination to result in X (or alternatively the XA configuration) becoming an inhibitor acting directly on the US, and the sequential discrimination in X becoming a negative occasion setter acting indirectly on the A-US link. After simultaneous FN training, X+ training completely abolished discriminative XA/A responding (Experiment 1), and X transferred inhibition to new targets B regardless of their training history (B+ or YB−/B+) (Experiment 2), suggesting X became a simple inhibitor. After sequential FN training, X showed the predicted selective transfer to a target B that also had been modulated (Y → B−/B+), not to a simple excitor (B+) (Experiment 4), but turning X into an excitor (X+) likewise disrupted discriminative X → A/A responding (Experiment 3). This suggests that X acquired a combination of modulatory and direct inhibitory properties, and that the joint contribution of both components is necessary for the suppression of the target-induced US activation.     

Relations among equivalence, naming, and conflicting baseline control

Three studies were conducted with different groups of 6 students each to explore the effects of training class-inconsistent relations and naming on demonstrations of emergent arbitrary stimulus relations. In all studies, two three-member equivalence classes of Greek symbols (A1B1C1 and A2B2C2) emerged as a result of training in conditional discriminations. Two new symbols were introduced (X and Y), and additional conditional discriminations were trained, whereby X was designated as the positive discriminative stimulus (S+) and Y was designated as the negative discriminative stimulus (S-) for A1 and B2. Conversely, Y was designated as the S+ and X as the S- for B1 and A2. This introduced conflicting sources of control within and between classes. In Study 1, subjects were not provided with names for the stimuli. In Study 2, the experimenter provided common names for the stimuli within each class. In Study 3, the subjects were required to use the common names during conditional discrimination training and test-trial blocks. In all experiments, equivalence responding with respect to the original classes was disrupted for some subjects subsequent to learning the new relations. Furthermore, in Studies 2 and 3, there were frequent examples of noncorrespondence between observed (listener or speaker) naming patterns and derived relations. These results support the view that demonstrations of equivalence are subject to control from a variety of sources rather than being fundamentally dependent on naming.     

Transfer between contextual conditional discriminations: an examination of how stimulus conjuctions are represented

In 2 experiments, when rats were placed in 1 pair of contexts, A and B, 2 relationships were in force (X --> food and Y --> no food), and when they were placed in another pair of contexts, C and D, the complementary relationships were operative (Y --> food and X --> no food). In Experiment 1, rats then received a 2nd discrimination that was either contextually congruent (in A and B, Y --> food and X --> no food; in C and D, X --> food and Y --> no food) or contextually incongruent (in A and D, Y --> food and X --> no food; in C and B, X --> food and Y --> no food) with the 1st discrimination. In Experiment 2, the 1st discrimination, involving X and Y, was interleaved with a 2nd discrimination, involving V and W, that was again either congruent (in A and B, V --> food and W --> no food) or incongruent (in A and D, V --> food and W --> no food) with the 1st discrimination. The congruent discriminations were acquired more readily than were the incongruent discriminations.     

Dissociations among ABA, ABC, and AAB recovery effects

In a human predictive learning experiment, the strengths of ABA, ABC, and AAB recovery effects after discrimination reversal learning were compared. Initially, a discrimination between two stimuli (X+, Y−) was trained in Context A. During Phase 2, participants received discrimination reversal training (X−, Y+) either in Context A (Group AAB) or in Context B (Group ABA, Group ABC). Testing occurred in the reversal context and in a neutral context (Group AAB, Group ABC) or in the reversal context and in the acquisition context (Group ABA). Recovery effects were similar in ABA and ABC procedures, whereas no AAB recovery effect occurred. Furthermore, the rate of discrimination reversal learning during Phase 2 was unaffected by whether training was conducted in the same or in a different context as initial learning. Results indicate that contextual stimuli are processed before interfering information is presented. These representations, however, do not acquire the ability to control performance.     

Transfer of specific contextual functions to novel conditional discriminations

Three adolescents and 4 children participated in studies designed to examine contextually controlled conditional discrimination performance. In Study 1, participants selected Comparison B1 in the presence one stimulus (A1) and Comparison B2 in the presence of another stimulus (A2) using a matching-to-sample procedure. Next, contextual stimuli X1 or X2 were presented, such that in the presence of X1, selection of B1 given A1 and selection of B2 given A2 were reinforced; and in the presence of X2, selection of B2 given A1 and selection of B1 given A2 were reinforced. Then, new conditional discriminations were taught with Stimuli E and F. When the contextual Stimuli X1 and X2 were presented, participants selected the same comparisons as previously established in the EF relations in the presence of X1, but the opposite comparison as in the EF relations in the presence of X2. The results then were replicated with new Stimuli G and H. In Study 2, a new conditional discrimination, CD, was taught. Then, four combinations of two-element samples--C1 and D1, C2 and D2, C1 and D2, or C2 and D1--were presented with X1 and X2 as comparisons. Five of 6 participants selected X1 in the presence of C1 and D1 or C2 and D2, and selected X2 in the presence of C1 and D2 or C2 and D1. Finally, in Study 3, two new discriminations IJ and JK were taught. Then, the transitive IK relations were tested with X1 and X2 as contextual stimuli. The 4 participants selected K1 in the presence of I1 and K2 in the presence of I2 when the contextual stimulus was X1--demonstrating class formation--and selected the other comparisons when the contextual stimulus was X2. These results suggest that the contextual control functions of X1 and X2 transferred even to relations that had not been directly taught. These results extend those demonstrating generalized contextual control by showing transfer of functions of the contextual stimuli in transitivity tests and when the former contextual stimuli were presented as comparisons.     

Constructing Relational Sentences: Establishing Arbitrarily AppliCable Comparative Relations with the Relational Completion Procedure

The present study examined the effectiveness of the Relational Completion Procedure (RCP) in establishing arbitrarily applicable comparative relations after training with either linear (Experiment 1) or non-linear (Experiment 2) arbitrary relations. Linear relations were those presented in a sequential order (i.e., A-B, B-C, C-D, etc.), while nonlinear relations were presented in a nonsequential order (i.e., C-D, B-C, A-B, etc.). After nonarbitrary relational training and testing to establish the contextual functions of MORE THAN and LESS THAN, adults underwent constructed-response nonarbitrary relational training and testing with physically dissimilar stimuli in which they were required to “complete” relational sentences in the presence of the cues. Next, during arbitrary relational training with abstract stimuli, participants were randomly assigned to 1 of 3 training groups: All-More (E > D, D > C., C > B, and B > A), All-Less (A < B, B < C., C < D, and D < E), and Less-More (A < B, B < C., D > C., and E > D). Finally, during arbitrary relational testing, probes for mutually entailed relations were followed by probes for 1-and 2-node combinatorially entailed relations. The yield, or percentage of participants passing, was 83% in Experiment 1 and 91.6% in Experiment 2. Our findings demonstrate that the RCP was successful in establishing arbitrarily applicable comparative relations with adults regardless of the order in which the arbitrary relational training trials were presented.

     

A derived transfer of simple discrimination and self-reported arousal functions in spider fearful and non-spider-fearful participants

Two experiments investigated the derived transfer of functions through equivalence relations established using a stimulus pairing observation procedure. In Experiment 1, participants were trained on a simple discrimination (A1+/A2-) and then a stimulus pairing observation procedure was used to establish 4 stimulus pairings (A1-B1, A2-B2, B1-C1, B2-C2). Subsequently, a transfer of the simple discrimination functions through equivalence relations was observed (e.g., C1+/C2-). These procedures were modified in Experiment 2, which demonstrated that spider-fearful and non-spider-fearful participants show differing levels of a transfer of self-reported arousal functions for stimuli used in equivalence relations with video-based material depicting scenes with spiders. The results demonstrate that the stimulus pairing observation procedure provides a viable alternative to matching-to-sample, and also offer tentative support for a derived-relations model of the acquisition of anxiety responses in at least one sub-clinical population.     

Acquired equivalence and distinctiveness of cues using a sensory-preconditioning procedure

In Experiment 1 rats were presented with two flavours, A and B, that were accompanied either by a third common flavour, X, or by two different flavours, X and Y, respectively. An aversion was then established to A, and the extent to which this aversion generalized to B was measured. Stimulus generalization was found to be more marked when A and B had both been presented with the same flavour, X, during pre-exposure than when A and B had been presented with X and Y. In Experiment 2 three target flavours, A, B, and C, were initially presented. In one pre-exposure condition presentations of A and B were accompanied by X, and C was presented alone; in a second condition A and B were presented in isolation, and C was accompanied by X. After an aversion had been established to A, half of the animals in each of the pre-exposure conditions were tested with B, and the remainder were tested with C. In both of the pre-exposure conditions the generalized aversion was more substantial to B than to C. An analysis of these results in terms of the mechanisms that have been supposed to underlie sensory preconditioning is presented.     

Reduced discriminability following perceptual learning with odours

In this experiment, the effects of prior experience on odour perception and discrimination were explored. Participants repeatedly sniffed a mixture composed of two odours, AX, as well as smelling two further odours alone, B and Y. After this training phase, participants were asked to rate the similarity of the odours A and X, B and Y, and a non-exposed pair C and Z. A and X were judged as significantly more alike than the other pairs. Exactly the same pattern emerged on a second test, in which participants were asked to select the odd odour out of sets of three. It was consistently harder for participants to pick the odd odour when the stimuli were drawn from the AX pair (eg A versus AX versus AX). Not only do these findings demonstrate that prior experience can affect odour perception, a finding not predicted by theories of odour perception based solely upon the physiochemical properties of odours, they also suggest that experience can act to selectively decrease discriminability.     

Mechanisms of contextual control when contexts are informative to solve the task

An experiment was conducted using a human instrumental learning task with the goal of evaluating the mechanisms underlying the deleterious effect of context-switching on responding to an unambiguous stimulus when contexts are informative to solve the task. Participants were trained in a context-based reversal discrimination in which two discriminative stimuli (X and Y) interchange their meaning across contexts A and B. In context A, discriminative stimulus Z consistently announced that the relationship between a specific instrumental response (RI) and a specific outcome (O1) was in effect. Performance in the presence of stimulus Z was equally deteriorated when the test was conducted outside the training context, regardless of whether the test context was familiar (context B) or new (context C). This result is consistent with the idea that participants code all the information presented in an informative context as context-specific with the context playing a role akin to an occasion setter.     

Acquired equivalence and distinctiveness of cues: I. Exploring a neural network approach

Rats were placed in 4 contexts (A, B, C, D) where they received 2 auditory stimuli (X, Y); in A and B, presentations of X were paired with food and those of Y were not, and in C and D, Y was paired with food and X was not. Rats then received combinations of contexts that had provided congruent (AB, CD) or incongruent (AD, CB) information about X and Y's relationship to food. Responding was more variable during congruent than incongruent trials (Experiment 1) and was systematically increased and decreased during congruent (relative to incongruent) trials by the presentation of food or no food, respectively (Experiment 2). These results support a connectionist approach to acquired changes in stimulus distinctiveness.     

Reflexivity without identity matching training: A first demonstration

Until now, the equivalence property of reflexivity—matching physically identical stimuli to themselves after training on a set of arbitrary matching relations—has not been demonstrated in any animal, human or nonhuman. Previous reports of reflexivity have either implicitly or explicitly involved reinforced training on other identity matching relations. Here we demonstrate reflexivity without prior identity matching training. Pigeons received concurrent successive matching training on three arbitrary matching tasks: AB (hue–form), BC (form–hue), and AC (hue–hue with different hues in the A and C sets). Afterwards, pigeons were tested for BB (form–form) reflexivity. Consistent with the predictions of Urcuioli's ( 2008 ) theory, pigeons preferentially responded to B comparison stimuli that matched the preceding B sample stimuli in testing (i.e., BB reflexivity). A separate experiment showed that a slightly different set of arbitrary matching baseline relations yielded a theoretically predicted “anti‐reflexivity” (or emergent oddity) effect in two of five pigeons. Finally, training on just two arbitrary successive matching tasks (AB and BC) did not yield any differential BB responding in testing for five of eight pigeons, with two others showing reflexivity and one showing antireflexivity. These data complement previous findings of symmetry and transitivity (the two other properties of equivalence) in pigeons.