Developing a deep understanding of animal cognition in tasks such as category learning demands that one first achieve an appreciation of an animal’s sensory/perceptual/memory world. In this project, we report work that, for the first time, derives a nonhuman high-dimensional psychological-scaling representation for a set of visual objects and uses the representation to predict complex forms of category learning in a nonhuman species. Specifically, we pursue the question of whether pigeons can acquire multiple hard-to-discriminate rock-image categories as defined in the geologic sciences. We test a formal computational model of associative learning on its ability to account quantitatively for pigeons’ category learning performance. A prerequisite for applying the model is to embed the rock images in a pigeon psychological similarity space. We achieve that goal by modeling pigeons’ performance in an independently conducted same-different discrimination task involving the identical set of to-be-categorized rock images. The models provide a unified and accurate quantitative account of intricate sets of same-different and categorization-confusion data in this high-dimensional rock-categories domain. The psychological similarity space derived for pigeons resembles to a surprising degree one previously derived for humans, but with some notable exceptions, which are crucial to explaining pigeons’ detailed patterns of categorization performance.

We propose and implement an approach for deriving multidimensional scaling (MDS) solutions for objects from diverse everyday-object categories. The goal is for the MDS solutions to capture relative similarities between pairs of objects both within and across the categories. For example, if the members of the category apples are more similar to one another than are the members of the category lamps, then the MDS solution for the apples will be more compressed overall than the MDS solution for lamps. To achieve this goal, the key idea is that, rather than collecting similarity-judgment data one category at a time, we alternate in random fashion across trials the category from which the similarity-judgment data are collected. We hypothesize that if similarity-judgment data are collected one category at a time, observers may recalibrate their judgment scale with respect to each individual category, which could cause loss of information of overall discriminability relations across the different categories. By using the alternating-category approach, observers may be able to maintain a more nearly constant judgment scale across the different categories. We combine the alternating-category procedure with the use of metric forms of MDS that produce MDS solutions in which differences in overall discriminability relations across categories are maintained. We provide preliminary evidence of the success of the approach by showing that, when used as input to a simple computational model of recognition memory, the derived MDS solutions predict reasonably well the false-alarm rates associated with the different categories observed in an old-new recognition experiment.

We conduct tests of a hybrid-similarity exemplar model on its ability to account for the context-dependent memorability of items embedded in high-dimensional category spaces. According to the model, recognition judgments are based on the summed similarity of test items to studied exemplars. The model allows for the idea that “self-similarity” among objects differs due to matching on highly salient distinctive features. Participants viewed a study list of rock images belonging to geologically defined categories where the number of studied items from each category was manipulated. Following study, the participants’ old-new recognition memory performance was tested. We also manipulated across experiments the nature of the encoding task used during the study phase: Experiment 1 used a category-description matching task, whereas Experiment 2 used more neutral encoding instructions. Hit rates were markedly lower in Experiment 2 than in Experiment 1 and participants relied less on the presence of distinctive features for recognizing old items in the second experiment. With a minimum of parameter estimation, the hybrid-similarity model provided good accounts of a wide variety of fundamental benchmark phenomena across the two experiments. These included changing levels of memorability due to contextual effects of category size, within- and between-category similarity, and the presence of distinctive features. However, the hybrid model and a variety of extensions of the model fell short in accounting for the variability in hit rates within the class of old target items themselves. We discuss future directions for potentially improving upon the current predictions from the model.

Classic studies of human categorization learning provided evidence that high-variability training in the prototype-distortion paradigm enhances subsequent generalization to novel test patterns from the learned categories. More recent work suggests, however, that when the number of training trials is equated across low-variability and high-variability training conditions, it is low-variability training that yields better generalization performance. Whereas the recent studies used cartoon-animal stimuli varying along binary-valued dimensions, in the present work we return to the use of prototype-distorted dot-pattern stimuli that had been used in the original classic studies. In accord with the recent findings, we observe that high-variability training does not enhance generalization in the dot-pattern prototype-distortion paradigm when the total number of training trials is equated across the conditions, even when training with very large numbers of distinct instances. A baseline version of an exemplar model captures the major qualitative pattern of results in the experiment, as do prototype models that make allowance for changes in parameter settings across the different training conditions. Based on the modeling results, we hypothesize that although high-variability training does not enhance generalization in the prototype-distortion paradigm, it may do so when participants learn more complex category structures.

Categorization and old-new recognition memory are closely linked topics in the cognitive-psychology literature and there have been extensive past efforts at developing unified formal modeling accounts of these fundamental psychological processes. However, the existing formal-modeling literature has almost exclusively used small sets of simplified stimuli and artificial category structures. The present work extends this literature by collecting both categorization and old-new recognition judgments on a large set of high-dimensional stimuli that form real-world category structures: namely, a set of 540 images of rocks belonging to the geologically-defined categories igneous, metamorphic and sedimentary. Participants first engaged in a learning phase in which they classified large sets of training instances into these real-world categories. This was followed by a test phase in which they classified both training and novel transfer items into the learned categories and also judged whether each item was old or new. We attempted to model both the classification and recognition test data at the level of individual items. Ultimately, the categorization data were well fit by both an exemplar and clustering model, but not by a prototype model. Only the exemplar model was able to provide a reasonable first-order account of the old-new recognition data; however, the standard version of the model failed to capture the variability in hit rates within the class of old-training items themselves. An extended hybrid-similarity version of the exemplar model that made allowance for boosts in self-similarity due to matching distinctive features yielded much improved accounts of the old-new recognition data. The study is among the first to test cognitive-process models on their ability to account quantitatively for old-new recognition of real-world, high-dimensional stimuli at the level of individual items.

To compare the discrimination performance of 6-year-old children for optotypes from six paediatric visual acuity tests and to fit Luce's Biased Choice Model to the data to estimate the relative similarities and bias for each optotype. Full data sets were collected from 20 typically developing 6-year-olds who had passed a vision screening. They were presented with single optotypes labelled 6/12 at a distance of 9 m and were asked to identify the optotype using a matching task containing all optotypes from the relevant test. The data were combined to form a confusion matrix for each test and a biased choice model was fitted to the data. Median correct performance varied from 40

Though individual categorization or decision processes have been studied separately in many previous investigations, few studies have investigated how they interact by using a two-stage task of first categorizing and then deciding. To address this issue, we investigated a categorization-decision task in two experiments. In both, participants were shown six faces varying in width, first asked to categorize the faces, and then decide a course of action for each face. Each experiment was designed to include three groups, and for each group, we manipulated the probabilistic contingencies between stimulus, category assignments, and decision consequences. For each group, each participant received three different sequences of category response, category feedback, decision response, and decision feedback. We found that participants were only partially responsive in the appropriate directions to the contingencies assigned to each group. Comparisons of results from different sequences provided evidence for empirical interference effects of categorization on decisions. The empirical interference effect is defined as the difference between the probability of taking a hostile action in decision-alone conditions and the total probability of taking a hostile action in categorization-decision conditions. To test competing accounts for multiple empirical results, including two-stage choice probabilities and empirical interference effects, we compared a quantum cognition model versus a two-stage exemplar categorization model at both aggregate and individual levels. Using a Bayesian information criterion, we found that the quantum model provided an overall better model fit than the exemplar model. Although both models predicted empirical interference effects, the exemplar model was able to generate probabilistic deviation by incorporating category information of the first stage into the feature representation of the subsequent decision stage, while the quantum model produced interference effect by superposition, measurement, and quantum entanglement.

Single-system exemplar models of categorization and old-new recognition have been challenged on grounds of dissociations observed between the tasks. Statistical dissociations have been reported that show low correlations between categorization and recognition. Behavioral dissociations have been reported in which amnesics with poor recognition memory perform normally on categorization tasks. Brain-imaging dissociations have been reported that indicate that separate neural networks are activated when participants engage in categorization vs. recognition. Whereas the usual interpretation is that separate memory systems govern categorization and recognition, this chapter reviews work that argues that single-system exemplar models provide natural accounts of the dissociation findings.

A classic issue in the cognitive psychology of human category learning has involved the contrast between exemplar and prototype models. However, experimental tests to distinguish the models have relied almost solely on use of artificially-constructed categories composed of simplified stimuli. Here we contrast the predictions from the models in a real-world natural-science category domain-geologic rock types. Previous work in this domain used a set of complementary methods, including multidimensional scaling and direct dimension ratings, to derive a high-dimensional feature space in which the rock stimuli are embedded. The present work compares the category-learning predictions of exemplar and prototype models that make reference to this derived feature space. The experiments include conditions that should be favorable to prototype abstraction, including use of multiple large-size categories, delayed transfer testing, and real-world category structures. Nevertheless, the results of the qualitative and quantitative model comparisons point toward the exemplar model as providing a far better account of the observed results. Evidence is also provided that participants do not rely on all-or-none rote memories for the stored exemplars but rather use remembered exemplars as a basis for generalizing to novel transfer items from the learned categories. Limitations and directions of future work are discussed. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

An important question in the cognitive-psychology of category learning concerns the manner in which observers generalize their trained category knowledge at time of transfer. In recent work, Conaway and Kurtz (Conaway and Kurtz, Psychonomic Bulletin Review 24:1312–1323, 2017) reported results from a novel paradigm in which participants learned rule-described categories defined over two dimensions and then classified test items in distant transfer regions of the category space. The paradigm yielded results that challenged the predictions from both exemplar-based and logical-rule-based models of categorization but that the authors suggested were as predicted by a divergent auto-encoder (DIVA) model (Kurtz, Psychonomic Bulletin Review 14:560–576, 2007, Kurtz, Psychology of learning and motivation, Academic Press, New York, 2015). In this article, we pursue these challenges by conducting replications and extensions of the original experiment and fitting a variety of computational models to the resulting data. We find that even an extended version of the exemplar model that makes allowance for learning-during-test (LDT) processes fails to account for the results. In addition, models that presume a mixture of salient logical rules also fail to account for the findings. However, as a proof of concept, we illustrate that a model that assumes a mixture of strategies across subjects—some relying on exemplar-based memories with LDT, and others on salient logical rules—provides an outstanding account of the data. By comparison, DIVA performs considerably worse than does this LDT-exemplar-rule mixture account. These results converge with past ones reported in the literature that point to multiple forms of category representation as well as to the role of LDT processes in influencing how observers generalize their category knowledge.

A fundamental component of human categorization involves learning to attend selectively to relevant dimensions and ignore irrelevant ones. Past research has shown that humans can learn flexible strategies in which the attended dimensions vary depending on the region of feature space in which classification takes place. However, region-specific selective attention (RSA) is often challenging to learn. Here, we test the hypothesis that RSA is facilitated when individual categories are embedded within single regions of stimulus space rather than dispersed across multiple regions. We conduct an experiment that varies across conditions whether categories are embedded within regions, but in which the same RSA strategy would benefit performance across the conditions. To evaluate the hypothesis, we use measures of overall performance accuracy as well as comparisons among formal computational models that do and do not make allowance for RSA. We find strong support for the hypothesis among the upper-median-performing participants in the tested groups. However, even in the condition that promotes the learning of RSA, performance is considerably worse than in comparison conditions in which a single set of dimensions can be attended throughout the entire stimulus space.

Participants gave recognition judgments for short lists of pictures of everyday objects. Pictures in a given list were an equal mixture of three types that varied according to the way they were used as targets and foils earlier in the same session. Under consistent-mapping (CM), targets and foils never switch roles; under varied-mapping (VM), targets and foils switch roles randomly across trials; whereas all-new (AN) items are novel on each trial of the experiment. Past research has shown that markedly enhanced performance occurs in CM conditions, leading to conclusions that item-response learning takes place in CM, perhaps automatically. However, almost all past research has compared CM, VM, and AN performance in between-blocks designs in which participants may adopt different cognitive strategies and criterion settings across the conditions. The present mixed-list design holds constant the strategy and criterion settings that are used for CM, VM, and AN items, and produced patterns of performance dramatically different than those observed in pure-list control conditions. We develop an extended version of an exemplar-based random-walk model of probe recognition to account for the major qualitative effects in the data. The data and the modeling provide evidence for strong item-response learning for CM foils but weak item-response learning for CM targets. We consider possible explanations for these effects in our General Discussion. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Teaching natural-science categories is highly challenging because the objects in such categories are composed of numerous complex dimensions that need to be perceived, evaluated, and integrated. Furthermore, the boundaries separating such categories are often fuzzy. A technique that has been proposed and investigated for enhancing the teaching of natural-science categories is feature highlighting, in which diagnostic features for identifying category members are explicitly described and illustrated. Using rock classification in geology as an example target domain, the present study further investigated the potential benefits of feature highlighting and also of providing causal explanations for the highlighted features. The authors found that feature highlighting did not always lead to improved generalization to novel members of the taught categories. However, robust beneficial effects were seen when the categories were relatively confusable ones and the stated diagnostic features were highly valid for distinguishing among the categories. Finally, at least under the present conditions, supplementing the highlighted features with causal explanations of the reasons for their occurrence did not further enhance the participants' rock-classification learning and generalization. Although the teaching of causal explanations is fundamental to science education, clear evidence that causal explanations enhance classification-learning per se in this domain remains to be demonstrated. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

In a novel version of the classic dot-pattern prototype-distortion paradigm of category learning, Homa et al. (2019) tested a condition in which individual training instances never repeated, and observed results that they claimed severely challenged exemplar models of classification and recognition. Among the results was a dissociation in which participants classified transfer items with high accuracy in the no-repeat condition, yet in old-new recognition tests showed no ability to discriminate between old and new items of the same level of distortion from the prototype. In addition, speed of classification learning was no faster in a condition in which a small set of training instances was repeated continuously compared with the no-repeat condition. Here we show through computer-simulation modeling that exemplar models naturally capture the classification-recognition dissociation in the no-repeat condition, as well as a wide variety of other qualitative effects reported by Homa et al. (2019). We also conduct new conceptual-replication experiments to investigate their reported null effect of repeated versus nonrepeated training instances on speed of classification learning. In contrast to Homa et al. (2019) we find that speed of learning is substantially faster in the repeat condition than in the no-repeat condition, precisely as exemplar models predict. The exemplar model also captures a wide variety of transfer effects observed following the completion of category learning, including the classification-recognition dissociation observed across the repeat and no-repeat conditions. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Category learning is a core component of course curricula in science education. For instance, geology courses teach categorization of rock types. Using the educationally authentic rock categories, the current project examined whether category learning at a broad-level (BL; igneous, sedimentary, and metamorphic rocks) could be enhanced by learning category information at a more specific-level (SL; e.g., diorite under igneous, breccia under sedimentary, etc.). Experiments 1 and 2 showed that SL training was inferior to BL training when participants were required to respond at the BL regardless of whether BL and SL category labels were presented simultaneously during classification training or SL categories were learned initially followed by training on the specific-broad level name associations. However, Experiments 3 and 4 showed that SL training was as good as BL training when the training was more extensive and participants were allowed to respond at the trained level. By considering confusion matrices (i.e., probabilities that instances in a given category was erroneously classified as belonging to other categories), we conjectured that between-SL category similarity, specifically the degree to which similar-looking SL categories belong to the same BL category, is an important factor in determining the efficacy of SL training. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Many successful formal models of human categorization have been developed, but these models have been tested almost exclusively using artificial categories, because deriving psychological representations of large sets of natural stimuli using traditional methods such as multidimensional scaling (MDS) has been an intractable task. Here, we propose a novel integration in which MDS representations are used to train deep convolutional neural networks (CNNs) to automatically derive psychological representations for unlimited numbers of natural stimuli. In an example application, we train an ensemble of CNNs to produce the MDS coordinates of images of rocks, and we show that the ensemble can predict the MDS coordinates of new sets of rocks, even those not part of the original MDS space. We then show that the CNN-predicted MDS representations, unlike off-the-shelf CNN representations, can be used in conjunction with a formal psychological model to predict human categorization behavior. We further show that the CNNs can be trained to produce additional dimensions that extend the original MDS space and provide even better model fits to human category-learning data. Our integrated method provides a promising approach that can be instrumental in allowing researchers to extend traditional psychological-scaling and category-learning models to the complex, high-dimensional domains that exist in the natural world.

An important goal in cognitive and mathematical psychology is to scale up the application of computational models of human classification learning to real-world, naturalistic domains. Application of the models, however, requires the derivation of the complex, multidimensional “feature spaces” in which the to-be-classified objects are embedded and to which the formal models make reference. In recent work, using rock classification in the geologic sciences as an example target domain, we used multidimensional scaling (MDS) of similarity-judgment data as an approach to deriving the feature space (Nosofsky et al., Behavior Research Methods 50:530–556, 2018c). However, subsequent work involving the modeling of independent sets of classification-learning data led us to the hypothesis that the MDS solution had many “missing dimensions” that were crucial to categorization performance (Nosofsky et al., Psychonomic Bulletin & Review 26:48–76, 2019). In the present work, we conduct a “search for the missing dimensions” in an effort to develop a more comprehensive feature-space representation for the rock stimuli. By supplementing the original MDS solution with the missing dimensions, we achieve dramatically improved accounts of varied sets of classification-learning data in this domain. We outline future steps for continuing and expanding the work to meet the goal of achieving meaningful computational modeling of human classification in naturalistic object domains.

BACKGROUND:Most science categories are hierarchically organized, with various high-level divisions comprising numerous subtypes. If we suppose that one's goal is to teach students to classify at the high level, past research has provided mixed evidence about whether an effective strategy is to require simultaneous classification learning of the subtypes. This past research was limited, however, either because authentic science categories were not tested, or because the procedures did not allow participants to form strong associations between subtype-level and high-level category names. Here we investigate a two-stage response-training procedure in which participants provide both a high-level and subtype-level response on most trials, with feedback provided at both levels. The procedure is tested in experiments in which participants learn to classify large sets of rocks that are representative of those taught in geoscience classes.RESULTS:The two-stage procedure yielded high-level classification performance that was as good as the performance of comparison groups who were trained solely at the high level. In addition, the two-stage group achieved far greater knowledge of the hierarchical structure of the categories than did the comparison controls.CONCLUSION:In settings in which students are tasked with learning high-level names for rock types that are commonly taught in geoscience classes, it is best for students to learn simultaneously at the high and subtype levels (using training techniques similar to the presently investigated one). Beyond providing insights into the nature of category learning and representation, these findings have practical significance for improving science education.