The formation of linked perceptual classes

Multiple-exemplar training with stimuli in four domains induced two new fill-based (A1' and A2') and satellite-image-based (B1' and B2') perceptual classes. Conditional discriminations were established between the endpoints of the A1' and B1' classes as well as the A2' and B2' classes. The emergence of linked perceptual classes was evaluated by the performances occasioned by nine cross-class probes that contained fill variants as samples and satellite variants as comparisons, along with nine other cross-class probes that consisted of satellite variants as samples and fill variants as comparisons. The 18 probes were first presented serially and then concurrently. Class-consistent responding indicated the emergence of linked perceptual classes. Of the linked perceptual classes, 70% emerged during the initial serial test. An additional 20% of the linked perceptual classes emerged during the subsequently presented concurrent test block. Thus, linked perceptual classes emerged on an immediate or delayed basis. Linked perceptual classes, then, share structural and fuctional similarities with equivalence classes, generalized equivalence classes, cross-modal classes, and complex maturally occurring categories, and may clarify processes such as intersensory perception.     

The role of common motor responses in stimulus categorization by preschool children

The purpose of this study was to assess the role of common motor responses as the "speaker" behavior on stimulus class formation, and the emergence of functional classes. Experiment 1 examined whether training one motor response to a set of three stimuli and a second motor response to another set of three stimuli would result in correct category-sort responses for 5 typically developing preschool children. Three of the children passed the categorization tests. Experiment 2 examined whether the classes formed in Experiment 1 were functional classes, and whether participants who did not pass categorization tests in Experiment 1 would do so following common vocal tact training. The 2 participants who failed categorization tests in Experiment 1 passed these tests in Experiment 2, although none of the participants passed the tests for functional classes. The results of the current study did not unequivocally support the naming hypothesis. Future research should therefore evaluate other possible sources of control that aid in stimulus categorization.     

Naming and categorization in young children: vocal tact training

In three experiments, 2- to 4-year-old children, following pretraining with everyday objects, were presented with arbitrary stimuli of differing shapes. In Experiment 1A, 9 subjects were trained one common tact response, "zag," to three of these and a second tact, "vek," to another three. In category match-to-sample Test 1, 4 subjects sorted accurately when required only to look at the sample before selecting from five comparisons. The remaining 5 subjects succeeded in Test 2, in which they were required to tact the sample before selecting comparisons. Experiment 1B showed, for 2 of these subjects, that tact training with 12 arbitrary stimuli established two six-member classes that were still intact 6 weeks later. In Experiment 2, 3 new subjects participated in a common tact training procedure that ensured that none of the exemplars from the same class were presented together prior to the test for three-member classes. Two subjects passed category Test 1 and the third passed Test 2. Tests showed subjects' listener behavior in response to hearing /zog/ and /vek/ to be in place. These experiments indicate that common naming is effective in establishing arbitrary stimulus classes and that category match-to-sample testing provides a robust measure of categorization.     

Typicality effects in contingency-shaped generalized equivalence classes

Two experiments were conducted using match-to-sample methodologies in an effort to model lexical classes, which include both arbitrary and perceptual relations between class members. Training in both experiments used a one-to-many mapping procedure with nonsense syllables as samples and eight sets of abstract stimuli as comparisons. These abstract stimuli differed along a number of dimensions, four of which were critical to the experimenter-defined class membership. Stimuli in some comparison sets included only one of the class-defining features, but stimuli in other sets included two, three, or all four of the critical features. After mastery of the baseline training, three types of probe tests were conducted: symmetry, transitivity/equivalence, and novel probe tests in which the training nonsense syllables served as samples, and comparisons were novel abstract stimuli that included one or more of the class-defining features. Symmetry and transitivity/equivalence probe tests showed that the stimuli used in training became members of equivalence classes. The novel stimuli also became class members on the basis of inclusion of any of the critical features. Thus these probe tests revealed the formation of open-ended generalized equivalence classes. In addition, typicality effects were observed such that comparison sets with more critical features were learned with fewer errors, responded to more rapidly, and judged to be better exemplars of the class. Contingency-shaped stimulus classes established through a match-to-sample procedure thus show several important behavioral similarities to natural lexical categories.     

Contextual control of stimulus generalization and stimulus equivalence in hierarchical categorization

The purpose of this study was to determine whether hierarchical categorization would result from a combination of contextually controlled conditional discrimination training, stimulus generalization, and stimulus equivalence. First, differential selection responses to a specific stimulus feature were brought under contextual control. This contextual control was hierarchical in that stimuli at the top of the hierarchy all evoked one response, whereas those at the bottom each evoked different responses. The evocative functions of these stimuli generalized in predictable ways along a dimension of physical similarity. Then, these functions were indirectly acquired by a set of nonsense syllables that were related via transitivity relations to the originally trained stimuli. These nonsense syllables effectively served as names for the different stimulus classes within each level of the hierarchy.     

Naming and categorization in young children: II. Listener behavior training

Following pretraining with everyday objects, 1- to 4-year-old children received listener training with three pairs of arbitrary stimuli of differing shapes. For each pair, 9 children were trained to select one stimulus in response to the spoken word /zog/ and the other to the spoken word /vek/. Next, in the look-at-sample category match-to-sample test, none categorized the six stimuli correctly when asked to look at the sample before selecting from five comparisons. Seven of these children failed a subsequent test of corresponding speaker behavior (tact test); following tact training, 5 of them passed either a repeat of the look-at-sample category test (2 subjects) or an alternative category test (3 subjects) in which they were required to tact the sample before selecting comparisons. The remaining 2 failed both category tests. Of the 2 who passed the tact test, 1 passed the tact-sample category test; the other failed to complete category testing. Two children were next given a second stimulus set. One passed the look-at-sample category test and the tact test; the other failed both tests but passed the tact-sample category test after tact training. The results show that 1- to 4-year-old children may learn listener behavior without corresponding speaker behavior. The results also show that common listener behavior is not sufficient to establish arbitrary stimulus classes, and they are consistent with the proposition that naming may be necessary for categorization of such stimuli.     

Discrimination of polymorphous stimulus sets by pigeons

Pigeons were trained to perform a visual discrimination between stimulus sets in which the presence of any two of three positive features made a stimulus positive, while any two of three negative features made it negative (there were thus three different positive and three different negative stimuli). After training, the birds were exposed to test stimuli that contained either all three positive or all three negative features. In Experiment I three pigeons were successfully trained by a successive method, and subsequently responded to the test stimuli as though they were positive or negative respectively. In Experiment II four pigeons were trained by a simultaneous method. Three learned the discrimination and generalized appropriately to the test stimuli, but they showed no preference between positive test and positive training stimuli, nor any consistent difference in speed of response to them; and similar results were found for negative stimuli. It is argued from this that the pigeons learned to respond to the stimuli as patterns (configurations of features) rather than to the constituent features, but that they generalized to the test stimuli by using the common features. The experiments show that pigeons could in principle learn to discriminate natural polymorphous classes (such as “pigeon” or “person”) without using any single feature, but neither the present experiments nor earlier ones demonstrating discriminations of such natural classes establish that pigeons make use of polymorphous concepts in the same way as people.     

The relation between stimulus function and equivalence class formation

Fifty participants were exposed to a simple discrimination-training procedure during which six S+ functions were established for six arbitrary stimuli, and S- functions were established for a further six stimuli. Following this training, each participant was exposed to one of five conditions. In the S+ condition, participants were exposed to a stimulus equivalence training and testing procedure using only the six S+ stimuli as samples and comparisons. In the S+/S- condition, participants were exposed to the same training and testing sequence as in the S+ condition, the difference being that three S+ and three S- stimuli were used as sample and comparison stimuli, with each set of three corresponding to the trained equivalence relations. In the S+/S- mixed condition, the S+ and S- stimuli were assigned to their roles as samples and comparisons in a quasi-random order. In the S- condition, all six S- stimuli were used. The no-function condition served as a control condition and employed stimuli for which no stimulus-control functions had been established. The results showed that, on average, participants required more testing trials to form equivalence relations when the stimuli involved were functionally similar rather than functionally different. Moreover, participants required more test trials to form equivalence relations when novel arbitrary stimuli, rather than functionally distinct stimuli, were used as samples and comparisons. The speed of acquisition of stimulus equivalence was also related to the number of functionally similar stimuli established before training. These findings indicate a variety of ways in which the emergence of equivalence relations is affected by the functional classes in which the relevant stimuli participate.     

Establishing stimulus equivalences among retarded adolescents.

Two separate sets of stimuli were established as stimulus classes in a match-to-sample task by using each stimulus as both a sample and choice stimulus for every other member within that set. After classes were established, the subjects were taught to select one member of each class when an associated auditory stimulus was presented. After this training, three of six subjects were able to select the remaining members of each stimulus class in response to the appropriate auditory stimuli without direct training. The other three subjects also demonstrated transfer after additional members of each stimulus class were brought under the control of the auditory stimuli through direct training.     

Transfer to intermediate forms following concept discrimination by pigeons: chimeras and morphs

Two experiments examined pigeons' generalization to intermediate forms following training of concept discriminations. In Experiment 1, the training stimuli were sets of images of dogs and cats, and the transfer stimuli were head/body chimeras, which humans tend to categorize more readily in terms of the head part rather than the body part. In Experiment 2, the training stimuli were sets of images of heads of dogs and cats, and the intermediate stimuli were computer-generated morphs. In both experiments, pigeons learned the concept discrimination quickly and generalized with some decrement to novel instances of the categories. In both experiments, transfer tests were carried out with intermediate forms generated from both familiar and novel exemplars of the training sets. In Experiment 1, the pigeons' transfer performance, unlike that of human infants exposed to similar stimuli, was best predicted by the body part of the stimulus when the chimeras were formed from familiar exemplars. Spatial frequency analysis of the stimuli showed that the body parts were richer in high spatial frequencies than the head parts, so these data are consistent with the hypothesis that categorization is more dependent on local stimulus features in pigeons than in humans. There was no corresponding trend when the chimeras were formed from novel exemplars. In Experiment 2, when morphs of training stimuli were used, response rates declined smoothly as the proportion of the morph contributed by the positive stimulus fell, although results with morphs of novel stimuli were again less orderly.     

Extending sequence-class membership with matching to sample

Three normal adults were first trained to point sequentially to each member of several pairs of visual stimuli. This baseline training established one class of stimuli to which subjects responded first, and another class of stimuli to which they responded second. Then, in a matching-to-sample procedure, baseline-sequence stimuli served as samples and new visual stimuli served as comparisons. Subjects were trained to choose one group of new comparisons when the sample was a "first" stimulus from the sequence baseline, and to choose the other new comparison stimuli when the sample was a "second" from the sequence baseline. When the new stimuli were then presented as pairs in the posttest, two subjects pointed to them in sequences predictable on the basis of the stimulus-class membership established during matching to sample. The failure of one subject to demonstrate sequential transfer was shown to be a consequence of the failure of the matching-to-sample procedure to establish stimulus classes. The production of sequences that were not directly trained suggested an empirical approach to the analysis of simple grammatical behavior.     

The effects of differential training procedures on linked perceptual class formation

When the stimuli in one perceptual class (A') become related to the stimuli in another perceptual class (B'), the two are functioning as a single linked perceptual class. A common linked perceptual class would be the sounds of a person's voice (class A') and the pictures of that person (class B'). Such classes are ubiquitous in real world settings. We describe the effects of a variety of training procedures on the formation of these classes. The results could account for the development of naturally occurring linked perceptual classes. Two perceptual classes (A' and B') were formed in Experiment 1. The endpoints of the A' class were called anchor (Aa) and boundary (Ab) stimuli. Likewise, the anchor and boundary stimuli in the B' class were represented as Ba and Bb. In Experiment 2, the A' and B' classes were linked by the establishment of one of four cross-class conditional discriminations: Aa-->Ba, Aa-->Bb, Ab-->Ba, or Ab-->Bb. Results were greatest after Aa-->Bb training, intermediate after Aa-->Ba and Ab-->Ba training, and lowest after Ab-->Bb training. Class formation was influenced by the interaction of the anchor/boundary values and the sample/comparison functions of the stimuli used in training. Experiment 3 determined whether class formation was influenced by different sets of two cross-class conditional discriminations: Aa-->Ba and Ab-->Bb, or Aa-->Bb and Ab-->Ba. Both conditions produced equivalent results. Similarities were attributable to the use of anchor stimuli as samples and boundary stimuli as comparisons in each training condition. Finally, the results afterjoint Aa-->Ba and Ab-->Bb training were much greater than those produced by summing the results of Aa-->Ba training alone and Ab-->Bb training alone. This same synergy was not observed after joint Aa-->Bb and Ab-->Ba training or either alone.     

On the role of covarying functions in stimulus class formation and transfer of function

This experiment investigated whether directly trained covarying functions are necessary for stimulus class formation and transfer of function in humans. Initial class training was designed to establish two respondent-based stimulus classes by pairing two visual stimuli with shock and two other visual stimuli with no shock. Next, two operant discrimination functions were trained to one stimulus of each putative class. The no-shock group received the same training and testing in all phases, except no stimuli were ever paired with shock. The data indicated that skin conductance response conditioning did not occur for the shock groups or for the no-shock group. Tests showed transfer of the established discriminative functions, however, only for the shock groups, indicating the formation of two stimulus classes only for those participants who received respondent class training. The results suggest that transfer of function does not depend on first covarying the stimulus class functions.     

Secondary generalization and categorization in pigeons

In Experiment 1, one group of pigeons learned to classify a set of stimuli into the human language classes cat, flower, car, and chair (categorization); another group learned to classify the same set into arbitrary classes (pseudocategorization). Then, both groups were trained on a new categorization task and their performance compared to that of a control group that had no initial classification training. Hull's (1943) notion of secondary generalization (generalization that is not based on physical similarity but on mediating associations) predicts that categorization experience will facilitate the learning of a new categorization task, whereas pseudocategorization experience will impair it. However, in Experiment 1, performance on the new categorization task was not differently affected by prior experience. In Experiment 2, pigeons initially trained to classify a set of 48 stimuli (original training) were later trained to classify a subset of four of these stimuli using new responses (reassignment training). Then, they were tested on the 44 remaining stimuli. Performance better accorded with original than with reassignment training, indicating that categorization training did not lead to the formation of equivalence classes of stimuli, in which the equivalence relationship is mediated by secondary generalization. The lack of evidence of secondary generalization implies that our pigeons failed to meet Lea's (1984) criterion for conceptual behavior.     

Development and crossmodal transfer of contextual control of emergent stimulus relations

Six normally capable adults first learned three conditional relations in each of two prospective equivalence classes via match-to-sample training with figures as conditional (sample) and discriminative (comparison) stimuli. Then one trained conditional relation in each prospective class was brought under the control of contextual stimuli, two dictated nonsense syllables. Test performances indicated the emergence of untrained conditional relations, and therefore two equivalence classes, that were conditional on the contextual stimuli. These tests involved untrained combinations of contextual stimuli and stimuli in conditional relations, suggesting that the contextual stimuli functioned independently to control conditional relations rather than forming compound stimuli with samples and comparisons in training. Next, two novel figures were made equivalent to each of the original dictated contextual stimuli by match-to-sample training and testing. On subsequent tests, all subjects demonstrated transfer of conditional control of untrained conditional relations from the original auditory contextual stimuli to equivalent visual stimuli. These outcomes further supported the conclusion that the contextual stimuli exerted true conditional control over conditional relations in the equivalence classes and were not merely elements of compound stimuli.     

Stimulus dependence in probabilistic category learning

The current study investigates whether probabilistic categorization on the Weather Prediction task involves a single, modality/domain general learning mechanism or there are modality/domain differences. The same probabilistic categorization task was used in three modalities/domains and two modes of presentation. Cues consisted of visual, auditory-verbal or auditory-nonverbal stimuli, and were presented either sequentially or simultaneously. Results show that while there was no general difference in performance across modalities/domains, the mode of presentation affected them differently. In the visual modality, simultaneous performance had a general advantage over sequential presentation, while in the auditory conditions, there was an initial advantage of simultaneous presentation, which disappeared, and in the non-verbal condition, gave over to a sequential advantage in the later stages of learning. Data suggest that there are strong peripheral modality effects; however, there are no signs of modality/domain of stimuli centrally affecting categorization.     

Independent contexts for category ratings: A range-frequency analysis

Subjects were instructed to rate with a single set of categories an intermingled series of smaller squares and larger circles, establishing a separate scale of size for each domain. The major finding was that the scale established for one domain was independent of the skewing of sizes in the other domain. However, subjects were also able to combine the two domains into a single context when instructed to do so, particularly when the smallest circle was only slightly larger than the largest square. A range-frequency analysis of the rating suggests that under certain conditions the endpoints defining the subjective range depend upon the stimulus frequencies: the subjective endpoint is more extreme when the closest stimuli are presented with greater relative frequency.     

The effect of logarithmic transformation on estimating the parameters of the generalized matching law

This study examined stimulus class membership established via stimulus-reinforcer relations. Mentally retarded subjects learned conditional discriminations with four two-member sets of visual stimuli (A, B, C, and D). On arbitrary-matching trials, they selected comparison stimuli B1 and B2 conditionally upon samples A1 and A2, respectively, and C1 and C2 conditionally upon B1 and B2, respectively. On identity-matching trials, they selected all stimuli as comparisons conditionally upon identical stimuli as samples. Throughout training, correct selections of A1, B1, C1, and D1 were followed by one reinforcer, R1, and those of A2, B2, C2, and D2 were followed by another, R2. Subsequent tests documented the formation of two four-member stimulus classes, A1-B1-C1-D1 and A2-B2-C2-D2. The class membership of the A, B, and C stimuli could have been based on equivalence relations that resulted from the arbitrary-matching training. D1 and D2 had never appeared on arbitrary-matching trials, however. Their class membership must have been based on relations with R1 and R2, respectively. Results thus confirm a previous finding that stimulus classes can be expanded via stimulus-reinforcer relations. They also define more precisely the potential nature of those classes and the conditions under which class membership can be established.     

An evaluation of intraverbal training and listener training for teaching categorization skills

The effects of vocal intraverbal training and listener training on the emergence of untrained categorization skills were evaluated. Five typically developing preschool children initially learned to name a number of previously unfamiliar visual stimuli. Each child then received one of two types of training. Intraverbal training involved reinforcing vocally emitted category names by the child in the presence of a spoken exemplar name. Listener training involved reinforcing the selection of visual stimuli by the child in the presence of a spoken category name. A multiple baseline design was used to evaluate the effects of training on each child's intraverbal category naming and stimulus selection. Pre- and posttests were used to assess additional emergent relations. Four of 5 participants did not demonstrate the emergence of any untrained relations. The current findings suggest that additional research is needed to determine the optimal sequencing of conditions in programs designed to teach categorization skills.     

CONDITIONAL DISCRIMINATION VS. MATCHING TO SAMPLE: AN EXPANSION OF THE TESTING PARADIGM

A subject's performance under a conditional-discrimination procedure defines conditional relations between stimuli: “If A1, then B1; if A2, then B2.” The procedure may also generate matching to sample. If so, the stimuli will be related not only by conditionality, but by equivalence: A1 and B1 will become equivalent members of one stimulus class, A2 and B2 of another. One paradigm for testing whether a conditional-discrimination procedure has generated equivalence relations uses three sets of stimuli, A, B, and C, three stimuli per set. Subjects learn to select Set-B and Set-C comparisons conditionally upon Set-A samples. Having been explicitly taught six sample-comparison relations, A1B1, A1C1, A2B2, A2C2, A3B3, and A3C3, subjects prove immediately capable of matching the B- and C-stimuli; six new relations emerge (B1C1, B2C2, B3C3, C1B1, C2B2, C3B3). The 12 stimulus relations, six taught and six emergent, define the existence of three three-member stimulus classes, A1B1C1, A2B2C2, and A3B3C3. This paradigm was expanded by introducing three more stimuli (Set D), and teaching eight children not only the AB and AC relations but DC relations also—selecting Set-C comparisons conditionally upon Set-D samples. Six of the children proved immediately capable of matching the B- and D-stimuli to each other. By selecting appropriate Set-B comparisons conditionally upon Set-D samples, and Set-D comparisons conditionally upon Set-B samples, they demonstrated the existence of three four-member stimulus classes, A1B1C1D1, A2B2C2D2, and A3B3C3D3. These larger classes were confirmed by the subjects' success with the prerequisite lower-level conditional relations; they were also able to select Set-D comparisons conditionally upon samples from Sets A and C, and to do the BC and CB matching that defined the original three-member classes. Adding the three DC relations therefore generated 12 more, three each in BD, DB, AD, and CD. Enlarging each class by one member brought about a disproportionate increase in the number of emergent relations. Ancillary oral naming tests suggested that the subject's application of the same name to each stimulus was neither necessary nor sufficient to establish classes of equivalent stimuli.