The role of comparison in perceptual learning: effects of concurrent exposure to similar stimuli on the perceptual effectiveness of their unique features

In three experiments, rats were given concurrent exposure to a compound flavor (AX) and to one of the elements of the compound (X). The perceptual effectiveness of A was then assessed by a test involving generalization of a conditioned aversion. Comparison was made with a preexposure procedure in which the compound and the common element were presented on separate trials, either in alternation or in separate blocks of trials. The effectiveness of the unique cue was less after blocked preexposure than after either of the other procedures; concurrent preexposure did not produce a greater effect than did alternating preexposure. These results challenge the suggestion that concurrent preexposure engages a special comparison process that will facilitate this form of perceptual learning.     

The role of inhibitory associations in perceptual learning

Preexposure to two compound flavors (AX and BX) typically enhances their discriminability: An aversion conditioned to AX will generalize less to BX, especially if the preexposure regime has involved alternated presentations of AX and BX rather than presenting all AX trials before BX trials (or vice versa). One possible explanation of this finding is that alternating preexposure establishes inhibitory associations between the two unique features A and B, thus counteracting the generalization produced by excitatory associations between X and A and between X and B, which might result in either the retrieval of B on a conditioning trial to AX, or the retrieval of A on a test trial to BX. Three experiments on flavor aversion conditioning in rats tested these predictions. Experiment 1 suggested that the more important of these excitatory associations was that which allowed X to retrieve A on the test trial to BX. Experiment 2 suggested that the more important inhibitory association was that which allowed B to inhibit the representation of A on this test trial. Experiment 3 provided direct evidence of the role of this inhibitory B⊣A association.     

The role of test stimuli in matching to compound samples by pigeons

Two pigeons matched to sample in a three-key operant conditioning chamber. In Experiment I, two different kinds of samples were presented on the center key.Element samples were members of one of two sample sets — colors (a red or blue disk) or lines (a vertical or horizontal orientation of a set of white lines). These samples were followed by their respective sample sets on the side keys as comparison stimuli.Compound samples consisted of a set of lines superimposed on a colored disk. Following these samples, either sample set could appear as comparison stimuli. Matching to compound samples was less accurate than matching to element samples. One interpretation is that sharing of attention among elements of a compound sample weakened stimulus control by each element. A different interpretation is that an element sample controlled matching better because it was physically identical to a comparison stimulus whereas a compound sample was not. Experiments II–IV evaluated this “generalization decrement” alternative by testing element- vs. compound sample control with both element and compound comparison stimuli. Irrelevant elements were added to form compound comparison stimuli, some of which were physically identical to a preceding compound sample, but never identical to an element sample. In all experiments, the addition of irrelevant elements of comparison stimuli reduced sample control. However, the generalization decrement hypothesis failed to predict how differences in performance maintained by element and compound samples were affected by different tests of sample control. Matching accuracy appeared to be independently determined by the number of elements in a sample and whether irrelevant elements were present during tests of sample control.     

Concept learning by pigeons: Matching-to-sample with trial-unique video picture stimuli

Pigeons were trained to match-to-sample with several new methodologies: a large number of stimuli, computer-drawn color picture stimuli, responses monitored by a computer touch screen, stimuli presented horizontally from the floor, and grain reinforcement delivered onto the picture stimuli. Following acquisition, matching-to-sample concept learning was assessed by transfer to novel stimuli on the first exposure to pairs of novel stimuli. One group (trial-unique), trained with 152 different pictures presented once daily, showed excellent transfer (80% correct). Transfer and baseline performances were equivalent, indicating that the matching-to-sample concept had been learned. A second group (2-stimulus), trained with only two different pictures, showed no evidence of transfer. These results are discussed in terms of the effect of numbers of exemplars on previous failures to find concept learning in pigeons, and the implications of the positive finding from this experiment on abstract concept learning and evolutionary cognitive development.     

The neuroscience of perceptual categorization in pigeons: A mechanistic hypothesis

We are surrounded by an endless variation of objects. The ability to categorize these objects represents a core cognitive competence of humans and possibly all vertebrates. Research on category learning in nonhuman animals started with the seminal studies of Richard Herrnstein on the category “human” in pigeons. Since then, we have learned that pigeons are able to categorize a large number of stimulus sets, ranging from Cubist paintings to English orthography. Strangely, this prolific field has largely neglected to also study the avian neurobiology of categorization. Here, we present a hypothesis that combines experimental results and theories from categorization research in pigeons with neurobiological insights on visual processing and dopamine-mediated learning in primates. We conclude that in both fields, similar conclusions on the mechanisms of perceptual categorization have been drawn, despite very little cross-reference or communication between these two areas to date. We hypothesize that perceptual categorization is a two-component process in which stimulus features are first rapidly extracted in a feed-forward process, thereby enabling a fast subdivision along multiple category borders. In primates this seems to happen in the inferotemporal cortex, while pigeons may primarily use a cluster of associative visual forebrain areas. The second process rests on dopaminergic error-prediction learning that enables prefrontal areas to connect top down the relevant visual category dimension to the appropriate action dimension.     

Reinstatement of postresponse stimuli prior to reward in delayed-reward discrimination learning by pigeons

Delayed-reward learning in pigeons was examined using a simultaneous red-green visual discrimination task in which the conditions during the delay interval were varied between groups. The nondifferential group received training in which the stimulus present during the 1-min delay was the same following a peck on the correct and incorrect colors. The other three groups received 1-min delay training in which different stimuli occurred in the delay interval following correct and incorrect choices. The differential group received continuous, differential stimuli during the delay. The reinstatement group received the differential stimuli in the 10 sec immediately following the choice and during the last 10 sec of the delay. The reversedcue group was treated in the same way, except that the 10-sec delay stimulus immediately following an incorrect response was also presented for 10 sec prior to reward on correct choices, and the stimulus following a correct response also occurred 10 sec before nonreward on incorrect choices. Nondifferential birds failed to learn the discrimination, while differential and reinstatement birds learned it readily. The reversed-cue birds learned to choose the incorrect stimulus. Differential and reinstatement birds showed no decrement in performance when the delay was increased to 2 min. These findings suggest that similarity of prereward and postresponse delay stimuli controls choice responding in long-delay learning, a finding compatible with both memorial and conditioned reinforcement interpretations.     

The nature of discrimination learning in pigeons

The results from five experiments are considered in relation to two of Spence's (1937, 1938) proposals concerning discrimination learning. In Experiments 1 and 2, we investigated whether his ideas about the interaction between excitatory and inhibitory generalization gradients can be used to understand how animals solve a complex patterning discrimination. The results supported a development of his proposals as put forward by Pearce (1994), provided a modification was made to Pearce's rulefor determining the shape ofthe generalization gradient. In Experiments 3, 4, and 5, we examined whether animals would pay more attention to stimuli that are relevant, rather than irrelevant, to the solution of a discrimination. The results supported this proposal for stimuli comprising visual patterns, but not for those comprising plain colors. The results also indicated that change of attention was a consequence of preliminary receptor-exposure acts, as envisaged by Spence, and not of more central changes in attention.     

Factors affecting conditional discrimination learning by pigeons: II. Physical and temporal characteristics of stimuli

Four experiments were performed to determine the stimulus characteristics that favor the development of conditional stimulus control in the single reversal paradigm with pigeon subjects. In Experiment 1, pigeons were trained on a successive discrimination between tone frequencies ranging from 350 to 3500 Hz in a particular houselight context condition (houselight-on or -off). The subjects then were trained on the reversal of the tone discrimination in the alternative context. Subsequent tone-frequency generalization testing in the two contexts indicated that they had failed to gain conditional control over the pigeons’ discriminative performance. Such control was obtained in Experiment 2, in which the two problems were alternated daily for 32 sessions of training. The gradients then peaked at the appropriate S+ value in each context. In Experiment 3, the key colors (blue vs. red) served as contexts while pigeons learned a successive discrimination in which the discriminative cues were houselight-on versus houselight-off conditions. This was followed by a reversal of the discrimination in the alternative key-color context condition. The key colors were effective conditional cues in this situation. In a previous experiment (Thomas, McKelvie, & Mah, 1985), key color had been ineffective as a conditional cue when the discriminative cues were lines superimposed on the colored background. In Experiment 4, key color was effective when the color and lines were presented on a single key as in the earlier experiment, but were sequenced such that the onset of the key color preceded and then overlapped the presentation of the lines. We concluded that conditional discriminations are easiest for pigeons when visual cues are used, but the conditional and discriminative cues must be presented in such a way that they do not combine to form a psychological compound.     

Backward associations: Differential learning about stimuli that follow the presence versus the absence of food in pigeons

During training intended to establish positive and negative backward associations, pigeons received periodic 3-sec illuminations of a grain magazine, with food simultaneously available on 50% of the illuminations. For backward-positive subjects, a red keylight followed food trials only; for backward-negative subjects, it followed no-food trials only. During this training, the birds hardly ever pecked the red stimulus (no evidence of backward conditioning). However, when the red stimulus subsequently preceded and signaled grain for all birds, the backward-positive subjects pecked sooner and more often than did the backward-negative subjects, and control subjects performed at an intermediate level. Increases in the amount of original training did not significantly change the facilitatory effects of the backward-positive stimulus, but seemed to enhance the retarding effects of the backward-negative stimulus. These group differences cannot be attributed to what happened after the backward stimuli during original training. The features of the present technique that could possibly have produced strong and long-lasting positive and negative backward associations are discussed in comparison to prior research that has indicated weak or transitory evidence of true backward conditioning.     

Contextual stimuli and memory retrieval in pigeons

Pigeons were given free-operant successive discrimination training in which, on alternate days, two different interdimensional problems were employed (color positive and line angle negative; a different color negative and a different line angle positive). Between days, these problems could be construed as intradimensional ones. For one group, training was conducted on each problem in the presence of the same ambient (contextual) stimuli, while for a second group each problem was trained in a different context. For a third group, these two contexts were randomly related to the problems. Postdiscrimination stimulus generalization gradients showed that peak shifts were obtained in both the same context and the random context groups, but no peak shifts were found for the group which learned each problem in a different context. Results were consistent with Spear’s (1973) treatment of animal memory which attributes retention test performance to the operation of a “context-elicited” retrieval process. An alternative uniprocess conditioning account of these and similar data was also discussed.     

Artificial grammar learning in pigeons

In two experiments, we developed a new methodology for studying complex stimulus control by spatial sequences of letters generated by artificial grammars. An artificial grammar is a system of rules that defines which letter sequences or strings are “grammatical.” In Experiment 1, pigeons learned to respond differently to strings conforming to a grammar versus strings that were nongrammatical distortions. Several different criteria all suggested that performance was controlled both by some short chunks of strings shared between reinforced training strings and novel transfer strings and by more complex sequential regularities. In Experiment 2, pigeons quickly and accurately learned to respond differently to strings conforming to one or the other of two different artificial grammars. As in Experiment 1, performance was controlled both by some short chunks and by more complex sequential regularities. The results are interpreted in terms of family resemblance and pose new goals for theories of complex stimulus control.     

Simultaneous processing of visual and spatial stimuli in pigeons

Pigeons were trained to symbolically match comparison stimuli to either visual sample stimuli presented on a center key or to spatial sample stimuli presented on side keys. Tests were carried out in which visual and spatial cues were simultaneously presented in compound and short-term memory was probed for either visual or spatial information. Symmetrical interference with the matching of visual and spatial components of compounds was found when the visual and spatial cues were presented on separate keys. However, when visual and spatial cues were superimposed on the same side key, no interference was observed relative to element control tests. Discussion of these findings focuses on accounts in terms of limited processing capacity, coding decrement, and receptor orientation mechanisms.     

Varieties of perceptual learning

Although most studies of perceptual learning in human participants have concentrated on the changes in perception assumed to be occurring, studies of nonhuman animals necessarily measure discrimination learning and generalization and remain agnostic on the question of whether changes in behavior reflect changes in perception. On the other hand, animal studies do make it easier to draw a distinction between supervised and unsupervised learning. Differential reinforcement will surely teach animals to attend to some features of a stimulus array rather than to others. But it is an open question as to whether such changes in attention underlie the enhanced discrimination seen after unreinforced exposure to such an array. I argue that most instances of unsupervised perceptual learning observed in animals (and at least some in human animals) are better explained by appeal to well-established principles and phenomena of associative learning theory: excitatory and inhibitory associations between stimulus elements, latent inhibition, and habituation.     

Sustained attention,intelligence, and the crucial role of perceptual processes

The contribution of perceptual processes to the correlation between measures of sustained attention and intelligence was investigated in considering computerized and paper–pencil tests of sustained attention. Perceptual processes were represented by tasks demanding signal detection and stimulus discrimination. A total of 117 participants completed two attention tests, as well as two tests that required signal detection and stimulus discrimination. Intelligence was measured by means of Advanced Progressive Matrices (APM) and Berliner Intelligenzstruktur-Test (BIS). Structural equation modeling showed that intelligence was best predicted by one independent latent variable, which included loadings of sustained attention, speed of signal detection, and of stimulus discrimination. The investigation of the structure of prediction revealed speed of stimulus discrimination and genuine sustained attention as major predictors of intelligence.     

Postevent cues bias recognition performance in pigeons

In the present study, we examined whether the presentation of postevent cues would bias recognition in a visual delayed matching-to-sample task with pigeons. Postevent cues were either consistent with the original target stimulus (i.e., they were the same as the correct choice option at recognition), inconsistent (i.e., they were the same as the incorrect recognition option), or neutral (i.e., they were different from both the correct and the incorrect recognition options). In Experiment 1, a single colored light served as the target stimulus. In Experiment 2, the target stimulus was one of three lights presented in a sequence. Both experiments demonstrated that recognition choices were biased toward the option corresponding to the postevent cue, but only if the cue occurred at the end of the delay interval. The present results mirror those found using the misinformation paradigm with human subjects.     

An investigation of ambiguous-cue learning in pigeons

Two experiments demonstrated that pigeons can solve a simultaneous discrimination in which on half the trials the positive and an ambiguous cue (A) are presented and on half the trials choice is between A and the negative stimulus. In Experiment 1, where a relatively nondistinctive A cue was used, performance on the former type of trial was superior to that shown on the latter. In Experiment 2, where a distinctive A cue was provided, this pattern of results was reversed. These findings are interpreted in terms of an approach-avoidance explanation first proposed by Leary (1958). Experiment 3 tested and confirmed a central prediction of this explanation by showing that in an orthodox simultaneous discrimination, occasional reinforcements of the negative stimulus produce less accurate performance than do nonreinforcements of the positive.     

Association of conditioned stimuli during serial conditioning by pigeons

CS2-CS1-US autoshaping was given to hungry pigeons and evidence of CS2-CS1 association was sought. Pigeons pecked that key of a two-key compound CS2 located where the upcoming CS1 key light was to occur. Such CS2-CS1 association was: (1) quickly acquired, (2) quite durable, and (3) stronger with simultaneous than with successive discrimination tasks.