A role for stimulus generalization in conditional discrimination learning

In each of three experiments on discrimination learning by rats, whether or not a 10-sec target stimulus was followed by food was determined by the nature of a 2-min background stimulus that accompanied it. A conditional discrimination was employed in Experiment 1 such that background A indicated food would follow one target but not the other, whereas this relationship between the targets and food was reversed in the presence of background B. Experiment 2 employed two feature-positive discriminations. Subsequent test trials revealed that the background for one discrimination was able to enhance responding during the target for the other discrimination. Experiment 3 employed a feature-positive and a feature-negative discrimination prior to test trials in which each target was presented separately during a compound of both background stimuli. The compound enhanced responding to the target from the feature-positive discrimination and reduced it to the target from the feature-negative discrimination. We suggest that to accommodate all these findings, the best explanation is provided by a configural model of Pavlovian conditioning.     

The generalization function is determined by one subject's probability learning data

In his recent review of probability learning, Estes (1972, p. 96) notes that “the chief limitation on effective application” of the models for learning with a continuum of responses to have emerged from the linear and pattern learning models is the failure to specify the smearing or generalization function. He notes that “a useful technique has been developed by Levine for estimating this function from individual data.” This paper describes, gives the rationale, and proves the validity of the previously unpublished technique.     

Asymmetry of generalization decrement in causal learning

Two experiments required volunteers to learn which of various “planes” caused high levels of pollution. Novel test items were then rated as causes of pollution. Items created by adding novel features were rated at the same level as that of the original training items but items created by removing features received reduced ratings. This asymmetry of generalization decrement was not predicted by a well-known configural model of stimulus representation (Pearce, 1987, 1994) but was predicted by a recently proposed model of stimulus representation, the replaced-elements model (Brandon, Vogel, & Wagner, 2000).     

A factorial analysis of Thurstone's sixteen primary mental abilities tests

This article deals with a factorial analysis of Thurstone'sPrimary Mental Abilities Tests. The analysis was made in order to determine whether the tests designated to measure a particular ability would be found upon completion of the analysis to be clustered together with significant loadings and be isolated from the tests of the other abilities. The results show that most of the tests specifically designed to measure six of the abilities were isolated with loadings varying in size. The tests for the remaining ability were not isolated. It is also shown that the tests are complex in nature, measuring more than any one single ability.     

Stimulus generalization in two associative learning processes

Recent studies involving nonlinear discrimination problems suggest that stimuli in human associative learning are represented configurally with narrow 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.     

Infants' learning,memory, and generalization of learning for bimodal events

Study 1 investigated whether infants 3 and 7 months of age show differential learning of and memory for sight-sound pairs depending on whether or not temporal synchrony was present; memory was assessed after a 10-min and 1-week interval. Study 2 examined whether 7-month-olds show generalization of learning when they encounter novel bimodal events that are similar (changes in size, orientation, or color, and spectral sound properties) to the sight-sound pairs learned 1 week earlier based on temporal synchrony. For Study 1, infants received a familiarization phase followed by a paired-comparison preference procedure to assess for learning of the sight-sound pairs. One week later a memory test was given. Results confirmed that 7-month-olds had no difficulty learning auditory-visual pairings regardless of whether or not events were temporally synchronous, and they remembered these 10 min and 1 week later. In contrast, 3-month-olds showed poorer learning of sight-sound associations in the no-synchrony than synchrony conditions, and memory for sight-sound pairs 1 week later was shown only for the synchrony conditions. Results for Study 2 revealed generalization of learning of bimodal pairings under all stimulus conditions after a 1-week interval at 7 months of age. Implications of these findings for development of intersensory knowledge are discussed.     

Thurstone's discriminal processes fifty years later

Four issues are discussed concerning Thurstone's discriminal processes: the distributions governing the representation, the nature of the response decision rules, the relation of the mean representation to physical characteristics of the stimulus, and factors affecting the variance of the representation. A neural schema underlying the representation is proposed which involves samples in time of pulse trains on individual neural fibers, estimators of parameters of the several pulse trains, samples of neural fibers, and an aggregation of the estimates over the sample. The resulting aggregated estimate is the Thurstonian representation. Two estimators of pulse rate, which is monotonic with signal intensity, are timing and counting ratios and two methods of aggregation are averaging and maximizing. These lead to very different predictions in a speed-accuracy experiment; data indicate that both estimators are available and the aggregation is by averaging. Magnitude estimation data are then used both to illustrate an unusual response rule and to study the psychophysical law. In addition, the pattern of variability and correlation of magnitude estimates on successive trials is interpreted in terms of the sample size over which the aggregation takes place. Neural sample size is equated with selective attention, and is an important factor affecting the variability of the representation. It accounts for the magical number seven phenomenon in absolute identification and predicts the impact of nonuniform distributions of intensities on the absolute identification of two frequencies. 1977 Psychometric Society Presidential Address. This work was supported in part by a grant of the National Science Foundation to Harvard University. I wish to express my appreciation to S. Burbeck, D. M. Green, M. Shaw, and B. Wandell for their useful comments on an earlier draft of this paper.     

Factorial analysis of Thurstone's seven primary abilities

Intercorrelation coefficients among Thurstone's seven primary mental abilities scores were obtained from scores of 170 freshmen engineering students on the experimental edition of thePrimary Mental Abilities Tests. These correlation coefficients were factored to four factors, interpreted as a general factor, a reasoning factor, a verbal factor, and, tentatively, a specific memory factor. The finding of a general factor for a college population corroborates the Thurstones' finding of a general factor for eighth-grade children.     

Object labeling influences infant phonetic learning and generalization

Different kinds of speech sounds are used to signify possible word forms in every language. For example, lexical stress is used in Spanish (/‘be.be/, ‘he/she drinks’ versus /be.’be/, ‘baby’), but not in French (/‘be.be/ and /be.’be/ both mean ‘baby’). Infants learn many such native language phonetic contrasts in their first year of life, likely using a number of cues from parental speech input. One such cue could be parents’ object labeling, which can explicitly highlight relevant contrasts. Here we ask whether phonetic learning from object labeling is abstract—that is, if learning can generalize to new phonetic contexts. We investigate this issue in the prosodic domain, as the abstraction of prosodic cues (like lexical stress) has been shown to be particularly difficult. One group of 10-month-old French-learners was given consistent word labels that contrasted on lexical stress (e.g., Object A was labeled /‘ma.bu/, and Object B was labeled /ma.’bu/). Another group of 10-month-olds was given inconsistent word labels (i.e., mixed pairings), and stress discrimination in both groups was measured in a test phase with words made up of new syllables. Infants trained with consistently contrastive labels showed an earlier effect of discrimination compared to infants trained with inconsistent labels. Results indicate that phonetic learning from object labeling can indeed generalize, and suggest one way infants may learn the sound properties of their native language(s).     

Mediated generalization in discrimination learning by Rhesus monkeys

Monkeys received discrimination training in which the choice of either a simultaneous or a serial compound of two visual images was rewarded before we assessed the monkeys' conditioned preference for one of the images from the compound. This preference was reduced or blocked if the other image had been associated with reward rather than nonreward prior to compound training. By contrast, the preference was enhanced if the other image was associated with reward rather than nonreward after compound training. The magnitudes of the blocking and enhancement were unaffected by the temporal structure, simultaneous or serial, of the stimulus compound. These results are discussed in terms of the representation of stimulus compounds, the role of within-compound associations, and the similarities between serial and simultaneous visual compounds as processed by monkeys.     

Enhanced visuomotor learning and generalization in expert surgeons

Although human motor learning has been intensively studied for many decades, it remains unknown whether group differences are present in expert cohorts that must routinely cope with and learn new visuomotor mappings such as expert minimally invasive surgeons. We found that expert surgeons compensate for a visuomotor perturbation more rapidly than naive controls. Modelling indicates that these differences in expert behavioural performance reflects greater trial-to-trial retention, as opposed to greater trial-to-trial learning rate. We also found that surgeons generalize to novel reach directions more broadly than controls, a result which was subsequently confirmed by our modelling. In general, our findings show that minimally invasive surgeons exhibit enhanced visuomotor learning and spatial generalization.     

Mediated generalization in discrimination learning by Rhesus monkeys

Monkeys received discrimination training in which the choice of either a simultaneous or a serial compound of two visual images was rewarded before we assessed the monkeys' conditioned preference for one of the images from the compound. This preference was reduced or blocked if the other image had been associated with reward rather than nonreward prior to compound training. By contrast, the preference was enhanced if the other image was associated with reward rather than nonreward after compound training. The magnitudes of the blocking and enhancement were unaffected by the temporal structure, simultaneous or serial, of the stimulus compound. These results are discussed in terms of the representation of stimulus compounds, the role of within-compound associations, and the similarities between serial and simultaneous visual compounds as processed by monkeys.     

Conceptual similarity promotes generalization of higher order fear learning

We tested the hypothesis that conceptual similarity promotes generalization of conditioned fear. Using a sensory preconditioning procedure, three groups of subjects learned an association between two cues that were conceptually similar, unrelated, or mismatched. Next, one of the cues was paired with a shock. The other cue was then reintroduced to test for fear generalization, as measured by the skin conductance response. Results showed enhanced fear generalization that correlated with trait anxiety levels in the group that learned an association between conceptually similar stimuli. These findings suggest that conceptual representations of conditional stimuli influence human fear learning processes.     

A parallel rule activation and rule synthesis model for generalization in category learning

This paper proposes a distinction between primary generalization (transfer from stored exemplars to perceived targets) and secondary generalization (transfer from inferred abstractions to perceived targets). This distinction is embodied in the parallel rule activation and rule synthesis (PRAS) model, a production model capable of exemplar-based and abstraction-based categorization. As an exemplar model, the PRAS model is related to the generalized context model (Nosofsky, 1984). Exemplars are stored in memory encoded as condition-action rules. Working as an exemplar-based model, rules are activated on the basis of their strength and their similarity to the current to-be-categorized instance. Similarity between a target and a stored exemplar is weighted for attention to the dimensions of the psychological space. Depending on the value of a special parameter, the PRAS model is also able to operate as an abstraction model. In the latter case, it attempts to construct generalizing productions, which are activated according to the same rules as the exemplar-specific rules. The model is described in detail. It is applied to a number of important observations described in the research literature, and an experiment is reported that tested the usefulness of the proposed secondary-generalization mechanism. Finally, the discussion elaborates on the implications of the present study for further research.