Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback—an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Unsupervised category learning with integral-dimension stimuli

Despite the recent surge in research on unsupervised category learning, the majority of studies have focused on unconstrained tasks in which no instructions are provided about the underlying category structure. Relatively little research has focused on constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. The few studies that have addressed this issue have focused almost exclusively on stimuli for which it is relatively easy to attend selectively to the component dimensions (i.e., separable dimensions). In the present study, we investigated the ability of participants to learn categories constructed from stimuli for which it is difficult, if not impossible, to attend selectively to the component dimensions (i.e., integral dimensions). The experiments demonstrate that individuals are capable of learning categories constructed from the integral dimensions of brightness and saturation, but this ability is generally limited to category structures requiring selective attention to brightness. As might be expected with integral dimensions, participants were often able to integrate brightness and saturation information in the absence of feedback--an ability not observed in previous studies with separable dimensions. Even so, there was a bias to weight brightness more heavily than saturation in the categorization process, suggesting a weak form of selective attention to brightness. These data present an important challenge for the development of models of unsupervised category learning.     

Unsupervised categorization with a child sample: category cohesion development

Studies into categorization have demonstrated that the ability to form concepts is an essential ability in cognitive development. For example, before a decision about anything can be made, firstly category concepts need to be acquired in order to make efficient decisions about that situation. The present study explored a particular type of category learning, not previously explored in this particular context – unsupervised categorization with 16 items and two dimensions, and comparing specifically children vs. adults. Previous studies have typically focused on simpler designs such as three items of two dimensions in the triad tasks, or a greater number of dimensions but with much fewer items per category in other unsupervised settings. This study investigated unsupervised categorization with two levels of task difficulty, and compared two different populations, children and adults. The findings revealed that adults performed better for the easy condition but there was no difference between these groups for the more difficult category task. The findings also revealed that unsupervised categorization in more complex settings result in more one dimensional sorting, for both children and adults. The results are discussed in the context of unsupervised categorization development abilities in children.     

Thirty-something categorization results explained: selective attention, eyetracking, and models of category learning

An eyetracking study testing D. L. Medin and M. M. Schaffer's (1978) 5-4 category structure was conducted. Over 30 studies have shown that the exemplar-based generalized context model (GCM) usually provides a better quantitative account of 5-4 learning data as compared with the prototype model. However, J. D. Smith and J. P. Minda (2000) argued that the GCM is a psychologically implausible account of 5-4 learning because it implies suboptimal attention weights. To test this claim, the authors recorded undergraduates' eye movements while the students learned the 5-4 category structure. Eye fixations matched the attention weights estimated by the GCM but not those of the prototype model. This result confirms that the GCM is a realistic model of the processes involved in learning the 5-4 structure and that learners do not always optimize attention, as commonly supposed. The conditions under which learners are likely to optimize attention during category learning are discussed.     

Measuring category intuitiveness in unconstrained categorization tasks

What makes a category seem natural or intuitive? In this paper, an unsupervised categorization task was employed to examine observer agreement concerning the categorization of nine different stimulus sets. The stimulus sets were designed to capture different intuitions about classification structure. The main empirical index of category intuitiveness was the frequency of the preferred classification, for different stimulus sets. With 169 participants, and a within participants design, with some stimulus sets the most frequent classification was produced over 50 times and with others not more than two or three times. The main empirical finding was that cluster tightness was more important in determining category intuitiveness, than cluster separation. The results were considered in relation to the following models of unsupervised categorization: DIVA, the rational model, the simplicity model, SUSTAIN, an Unsupervised version of the Generalized Context Model (UGCM), and a simple geometric model based on similarity. DIVA, the geometric approach, SUSTAIN, and the UGCM provided good, though not perfect, fits. Overall, the present work highlights several theoretical and practical issues regarding unsupervised categorization and reveals weaknesses in some of the corresponding formal models.     

Contrasting rule-based and similarity-based category learning: The effects of mood and prior knowledge on ambiguous categorization

This study adopts a dual-system view of category learning. The findings suggest that consumers who learn a dominant feature as a verbal rule for a product category will classify a new ambiguous product according to that feature even if it more closely resembles a different product category. The findings also demonstrate that dominant features can bias categorization toward a less prototypical category in the event that the new product breaks the rule. We refer to this phenomenon as criterial inferencing. Lastly, we offer unique empirical evidence to suggest that mood influences category learning and thus attenuates the criterial inferencing bias.     

The effect of category variability in perceptual categorization

Exemplar and distributional accounts of categorization make differing predictions for the classification of a critical exemplar precisely halfway between the nearest exemplars of 2 categories differing in variability. Under standard conditions of sequential presentation, the critical exemplar was classified into the most similar, least variable category, consistent with an exemplar account. However, if the difference in variability is made more salient, then the same exemplar is classified into the more variable, most likely category, consistent with a distributional account. This suggests that participants may be strategic in their use of either strategy. However, when the relative variability of 2 categories was manipulated, participants showed changes in the classification of intermediate exemplars that neither approach could account for.     

The impact of category separation on unsupervised categorization

Most previous research on unsupervised categorization has used unconstrained tasks in which no instructions are provided about the underlying category structure or in which the stimuli are not clustered into categories. Few studies have investigated constrained tasks in which the goal is to learn predefined stimulus clusters in the absence of feedback. These studies have generally reported good performance when the stimulus clusters could be separated by a one-dimensional rule. In the present study, we investigated the limits of this ability. Results suggest that even when two stimulus clusters are as widely separated, as in previous studies, performance is poor if within-category variance on the relevant dimension is nonnegligible. In fact, under these conditions, many participants failed even to identify the single relevant stimulus dimension. This poor performance is generally incompatible with all current models of unsupervised category learning.     

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. Experiments 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 Experiments 3 and 4. The model formalizes 3 explanatory principles: rapidly shifting attention with learned shifts, decreasing learning rates, and graded similarity in exemplar representation.     

Category use and category learning

Categorization models based on laboratory research focus on a narrower range of explanatory constructs than appears necessary for explaining the structure of natural categories. This mismatch is caused by the reliance on classification as the basis of laboratory studies. Category representations are formed in the process of interacting with category members. Thus, laboratory studies must explore a range of category uses. The authors review the effects of a variety of category uses on category learning. First, there is an extensive discussion contrasting classification with a predictive inference task that is formally equivalent to classification but leads to a very different pattern of learning. Then, research on the effects of problem solving, communication, and combining inference and classification is reviewed.     

Combining exemplar-based category representations and connectionist learning rules.

Adaptive network and exemplar-similarity models were compared on their ability to predict category learning and transfer data. An exemplar-based network (Kruschke, 1990a, 1990b, 1992) that combines key aspects of both modeling approaches was also tested. The exemplar-based network incorporates an exemplar-based category representation in which exemplars become associated to categories through the same error-driven, interactive learning rules that are assumed in standard adaptive networks. Experiment 1, which partially replicated and extended the probabilistic classification learning paradigm of Gluck and Bower (1988a), demonstrated the importance of an error-driven learning rule. Experiment 2, which extended the classification learning paradigm of Medin and Schaffer (1978) that discriminated between exemplar and prototype models, demonstrated the importance of an exemplar-based category representation. Only the exemplar-based network accounted for all the major qualitative phenomena; it also achieved good quantitative predictions of the learning and transfer data in both experiments.     

Blocking in category learning

Many theories of category learning assume that learning is driven by a need to minimize classification error. When there is no classification error, therefore, learning of individual features should be negligible. The authors tested this hypothesis by conducting three category-learning experiments adapted from an associative learning blocking paradigm. Contrary to an error-driven account of learning, participants learned a wide range of information when they learned about categories, and blocking effects were difficult to obtain. Conversely, when participants learned to predict an outcome in a task with the same formal structure and materials, blocking effects were robust and followed the predictions of error-driven learning. The authors discuss their findings in relation to models of category learning and the usefulness of category knowledge in the environment.     

Unitization during category learning

Five experiments explored the question of whether new perceptual units can be developed if they are useful for a category learning task, and if so, what the constraints on this unitization process are. During category learning, participants were required to attend either a single component or a conjunction of 5 components. Consistent with unitization, the conjunctive task became much easier with practice; this improvement was not found for the single-component task or for conjunctive tasks in which the components could not be unitized. Influences of component organization (Experiment 1), component contiguity (Experiment 2), component proximity (Experiment 3), and number of components (Experiment 4) on practice effects were found. Deconvolved response times (Experiment 5) showed that prolonged practice yielded faster responses than predicted by an analytic model that integrates evidence from independently perceived components.     

Effects of contrasting category,conjoint frequency and typicality on categorization

Two experiments were conducted to investigate whether (a) experience with a contrasting category, (b) conjoint frequency of dimensional values, (c) range of typicality of values, and (d) type of information administered during learning influenced subsequent test performance. Each experiment began with an observational category learning task, employing faces as stimuli. This was followed by a classification test task and by pairwise comparisons of faces. Influence of a contrasting category was studied in experiment 1 by varying frequency of values of the contrasting category, and in experiment 2 by either including or not including a contrasting category in the learning task. Results indicated that (a) categorization is influenced by experience with a contrasting category, (b) conjoint frequency enhances the importance of values to a category, (c) broad typicality range experience reduces typicality differences among exemplars of a category, whereas small range experience diminishes differences in a contrasting category, and (d) information on representativeness of exemplars does not facilitate subsequent test performance. The implications of the results for categorization models are discussed.     

Learning and transfer of category knowledge in an indirect categorization task

Knowledge representations acquired during category learning experiments are ‘tuned’ to the task goal. A useful paradigm to study category representations is indirect category learning. In the present article, we propose a new indirect categorization task called the “same”–“different” categorization task. The same–different categorization task is a regular same–different task, but the question asked to the participants is about the stimulus category membership instead of stimulus identity. Experiment 1 explores the possibility of indirectly learning rule-based and information-integration category structures using the new paradigm. The results suggest that there is little learning about the category structures resulting from an indirect categorization task unless the categories can be separated by a one-dimensional rule. Experiment 2 explores whether a category representation learned indirectly can be used in a direct classification task (and vice versa). The results suggest that previous categorical knowledge acquired during a direct classification task can be expressed in the same–different categorization task only when the categories can be separated by a rule that is easily verbalized. Implications of these results for categorization research are discussed.     

Adaptive categorization in unsupervised learning

In 3 experiments, the authors provide evidence for a distinct category-invention process in unsupervised (discovery) learning and set forth a method for observing and investigating that process. In the 1st 2 experiments, the sequencing of unlabeled training instances strongly affected participants' ability to discover patterns (categories) across those instances. In the 3rd experiment, providing diagnostic labels helped participants discover categories and improved learning even for instance sequences that were unlearnable in the earlier experiments. These results are incompatible with models that assume that people learn by incrementally tracking correlations between individual features; instead, they suggest that learners in this study used expectation failure as a trigger to invent distinct categories to represent patterns in the stimuli. The results are explained in terms of J. R. Anderson's (1990, 1991) rational model of categorization, and extensions of this analysis for real-world learning are discussed.     

Procedural learning in perceptual categorization

In two experiments, observers learned two types of category structures: those in which perfect accuracy could be achieved via some explicit rule-based strategy and those in which perfect accuracy required integrating information from separate perceptual dimensions at some predecisional stage. At the end of training, some observers were required to switch their hands on the response keys, whereas the assignment of categories to response keys was switched for other observers. With the rule-based category structures, neither change in response instructions interfered with categorization accuracy. However, with the information-integration structures, switching response key assignments interfered with categorization performance, but switching hands did not. These results are consistent with the hypothesis that abstract category labels are learned in rule-based categorization, whereas response positions are learned in information-integration categorization. The association to response positions also supports the hypothesis of a procedural-learning-based component to information integration categorization.     

Comparing supervised and unsupervised category learning

Two unsupervised learning modes (incidental and intentional unsupervised learning) and their relation to supervised classification learning are examined. The approach allows for direct comparisons of unsupervised learning data with the Shepard, Hovland, and Jenkins (1961) seminal studies in supervised classification learning. Unlike supervised classification learning, unsupervised learning (especially under incidental conditions) favors linear category structures over compact nonlinear category structures. Unsupervised learning is shown to be multifaceted in that performance varies with task conditions. In comparison with incidental unsupervised learning, intentional unsupervised learning is more rule like, but is no more accurate. The acquisition and application of knowledge is also more laborious under intentional unsupervised learning.     

The effects of category overlap on information-integration and rule-based category learning

In three experiments, we investigated whether the amount of category overlap constrains the decision strategies used in category learning, and whether such constraints depend on the type of category structures used. Experiments 1 and 2 used a category-learning task requiring perceptual integration of information from multiple dimensions (an information-integration task) and Experiment 3 used a task requiring the application of an explicit strategy (a rule-based task). In the information-integration task, participants used perceptual-integration strategies at moderate levels of category overlap, but explicit strategies at extreme levels of overlap--even when such strategies were suboptimal. In contrast, in the rule-based task, participants used explicit strategies, regardless of the level of category overlap. These data are consistent with a multiple systems view of category learning, and suggest that categorization strategy depends on the type of task that is used, and on the degree to which each stimulus is probabilistically associated with the contrasting categories.