Journey to the center of the category: the dissociation in amnesia between categorization and recognition

Premise typicality as feature inference decision-making in perceptual categories

Making property inferences for category instances is important and has been studied in two largely separate areas—categorical induction and perceptual categorization. Categorical induction has a corpus of well-established effects using complex, real-world categories; however, the representational basis of these effects is unclear. In contrast, the perceptual categorization paradigm has fostered the assessment of well-specified representation models due to its controlled stimuli and categories. In categorical induction, evaluations of premise typicality effects, stronger attribute generalization from typical category instances than from atypical, have tried to control the similarity between instances to be distinct from premise–conclusion similarity effects, stronger generalization from greater similarity. However, the extent to which similarity has been controlled is unclear for these complex stimuli. Our research embedded analogues of categorical induction effects in perceptual categories, notably premise typicality and premise conclusion similarity, in an attempt to clarify the category representation underlying feature inference. These experiments controlled similarity between instances using overlap of a small number of constrained features. Participants made inferences for test cases using displayed sets of category instances. The results showed typicality effects, premise–conclusion similarity effects, but no evidence of premise typicality effects (i.e., no preference for generalizing features from typical over atypical category instances when similarity was controlled for), with significant Bayesian support for the null. As typicality effects occurred and occur widely in the perceptual categorization paradigm, why was premise typicality absent? We discuss possible reasons. For attribute inference, is premise typicality distinct from instance similarity? These initial results suggest not.

Breaking the perceptual-conceptual barrier: Relational matching and working memory

Cognitive, comparative, and developmental psychologists have long been interested in humans' and animals' ability to respond to abstract relations, as this ability may underlie important capacities like analogical reasoning. Cross-species research has used relational matching-to-sample (RMTS) tasks in which participants try to find stimulus pairs that "match" because they both express the same abstract relation (same or different). Researchers seek to understand the cognitive processes that underlie successful matching performance. In the present RMTS paradigm, the abstract-relational cue was made redundant with a first-order perceptual cue. Then the perceptual cue faded, requiring participants to transition from a perceptual to a conceptual approach by realizing the task's abstract-relational affordance. We studied participants' ability to make this transition with and without a working-memory load. The concurrent load caused participants to fail to break the perceptual-conceptual barrier unless the load was abandoned. We conclude that finding the conceptual solution depends on reconstruing the task using cognitive processes that are especially reliant on working memory. Our data provide the closest existing look at this cognitive reorganization. They raise important theoretical issues for cross-species comparisons of relational cognition, especially regarding animals' limitations in this domain.

Computational and empirical simulations of selective memory impairments: Converging evidence for a single-system account of memory dissociations

Current theory has divided memory into multiple systems, resulting in a fractionated account of human behaviour. By an alternative perspective, memory is a single system. However, debate over the details of different single-system theories has overshadowed the converging agreement among them, slowing the reunification of memory. Evidence in favour of dividing memory often takes the form of dissociations observed in amnesia, where amnesic patients are impaired on some memory tasks but not others. The dissociations are taken as evidence for separate explicit and implicit memory systems. We argue against this perspective. We simulate two key dissociations between classification and recognition in a computational model of memory, A Theory of Nonanalytic Association. We assume that amnesia reflects a quantitative difference in the quality of encoding. We also present empirical evidence that replicates the dissociations in healthy participants, simulating amnesic behaviour by reducing study time. In both analyses, we successfully reproduce the dissociations. We integrate our computational and empirical successes with the success of alternative models and manipulations and argue that our demonstrations, taken in concert with similar demonstrations with similar models, provide converging evidence for a more general set of single-system analyses that support the conclusion that a wide variety of memory phenomena can be explained by a unified and coherent set of principles.

Formal models in animal-metacognition research: the problem of interpreting animals' behavior

Ongoing research explores whether animals have precursors to metacognition-that is, the capacity to monitor mental states or cognitive processes. Comparative psychologists have tested apes, monkeys, rats, pigeons, and a dolphin using perceptual, memory, foraging, and information-seeking paradigms. The consensus is that some species have a functional analog to human metacognition. Recently, though, associative modelers have used formal-mathematical models hoping to describe animals' "metacognitive" performances in associative-behaviorist ways. We evaluate these attempts to reify formal models as proof of particular explanations of animal cognition. These attempts misunderstand the content and proper application of models. They embody mistakes of scientific reasoning. They blur fundamental distinctions in understanding animal cognition. They impede theoretical development. In contrast, an energetic empirical enterprise is achieving strong success in describing the psychology underlying animals' metacognitive performances. We argue that this careful empirical work is the clear path to useful theoretical development. The issues raised here about formal modeling-in the domain of animal metacognition-potentially extend to biobehavioral research more broadly.

Ecology, Fitness, Evolution: New Perspectives on Categorization

Categorization's great debate has weighed single-system exemplar theory against the possibility of alternative processing systems. We take an evolutionary perspective toward this debate to illuminate it in a new way. Animals are crucial behavioral ambassadors to this area. They reveal the roots of human categorization, the basic assumptions of vertebrates entering category tasks, and the surprising weakness of exemplar memory as a category-learning strategy. These results have joined neuroscience results to prompt important changes in categorization theory. Categorization's great debate is ending. Categorization is served by multiple systems of process and representation.

Task variations and attention shifts in young children’s category learning

Two studies examined the conditions under which 6-year-old children succeeded in discovering prototypical information within ill-defined categories for fictitious animals that had salient individuating properties. Following either incidental or intentional learning of a single category, children attended to both prototypical and instance-specific features when judging the category membership of new examples (Experiment 1). When the same category was contrasted with a similar category in a sorting-with-feedback procedure, children relied on prototypical features in categorisation despite the fact that instance-specific features dominated their recognition-memory judgements (Experiment 2). The results show young children to be capable of shifting their attention to different kinds of category attributes according to the conditions of category formation and the nature of the assessment task.

An interplay of fusiform gyrus and hippocampus enables prototype- and exemplar-based category learning

The aim of the present study was to examine the contributions of different brain structures to prototype- and exemplar-based category learning using functional magnetic resonance imaging (fMRI). Twenty-eight subjects performed a categorization task in which they had to assign prototypes and exceptions to two different families. This test procedure usually produces different learning curves for prototype and exception stimuli. Our behavioral data replicated these previous findings by showing an initially superior performance for prototypes and typical stimuli and a switch from a prototype-based to an exemplar-based categorization for exceptions in the later learning phases. Since performance varied, we divided participants into learners and non-learners. Analysis of the functional imaging data revealed that the interaction of group (learners vs. non-learners) and block (Block 5 vs. Block 1) yielded an activation of the left fusiform gyrus for the processing of prototypes, and an activation of the right hippocampus for exceptions after learning the categories. Thus, successful prototype- and exemplar-based category learning is associated with activations of complementary neural substrates that constitute object-based processes of the ventral visual stream and their interaction with unique-cue representations, possibly based on sparse coding within the hippocampus.

Stages of Abstraction and Exemplar Memorization in Pigeon Category Learning

It has been proposed that human category learning consists of an early abstraction-based stage followed by a later exemplar-memorization stage. To investigate whether similar processing stages extend to category learning in a nonverbal species, we applied a prototype-exception paradigm to investigating pigeon category learning. Five birds and 8 humans learned six-dimensional perceptual categories constructed to include prototypes, typical items, and exceptions. We evaluated the birds' and humans' categorization strategies at different points during learning. Early on in both species, prototype performance improved rapidly as exception performance remained below chance, indicating an initial mastery of the categories' general structure. Later on, exception performance improved selectively and dramatically, indicating exception-item resolution and exemplar memorization. Abstraction- and exemplar-based formal models reinforced these interpretations. The results suggest a psychological transition in pigeon category learning from abstraction- to exemplar-based processing similar to that found in humans.

Refining the visual-cortical hypothesis in category learning

Participants produce steep typicality gradients and large prototype-enhancement effects in dot-distortion category tasks, showing that in these tasks to-be-categorized items are compared to a prototypical representation that is the central tendency of the participant's exemplar experience. These prototype-abstraction processes have been ascribed to low-level mechanisms in primary visual cortex. Here we asked whether higher-level mechanisms in visual cortex can also sometimes support prototype abstraction. To do so, we compared dot-distortion performance when the stimuli were size constant (allowing some low-level repetition-familiarity to develop for similar shapes) or size variable (defeating repetition-familiarity effects). If prototype formation is only mediated by low-level mechanisms, stimulus-size variability should lessen prototype effects and flatten typicality gradients. Yet prototype effects and typicality gradients were the same under both conditions, whether participants learned the categories explicitly or implicitly and whether they received trial-by-trial reinforcement during transfer tests. These results broaden out the visual-cortical hypothesis because low-level visual areas, featuring retinotopic perceptual representations, would not support robust category learning or prototype-enhancement effects in an environment of pronounced variability in stimulus size. Therefore, higher-level cortical mechanisms evidently can also support prototype formation during categorization.

Similarity, typicality, and category-level matching of morphed outlines of everyday objects

During visual object categorisation, a match must be found between the input image and stored information about basic-level categories. Graf [2002 Form, Space and Object (Berlin: Wissenschaftlicher Verlag Berlin)] suggested the involvement of analogue transformational, shape-changing processes in aligning the memory representation of the category with the perceptual representation of the current stimulus. Here we compare the predictions of alignment models with those of exemplar-based models, using morphing between four exemplar outlines within each of eleven categories. Overall, with increasing transformational distance between two exemplars of the same category, reaction times to decide whether they belong to the same category in a sequential matching paradigm increased, while rated similarity between the two exemplars decreased. However, in contrast to alignment accounts, exemplar-based accounts can correctly predict the observed dissociation between typicality and categorisation time, and allow the observed deviations from sequential additivity and nonlinear relations between transformational distance and rated similarity. Discussion of integrations of exemplar-based theories with neglected processes, such as information accumulation, response competition, response priming, and gain-modulation leads to a view of the recognition process from input to response, which increases the validity and scope of modern exemplar-based categorisation and recognition models.

Psychological Distance and Group Judgments: The Effect of Physical Distance on Beliefs about Common Goals

The present research examined the consequences of physical distance on beliefs about common goals, which have been implicated in judgments of entitativeness (“groupness”) of social entities. A central feature of task groups is the degree to which its members are driven by common goals. According to construal level theory, as stimuli are removed psychologically (e.g., physically), people construe stimuli in more abstract terms, focusing more on central features of stimuli. Adopting this framework, four studies demonstrated that people are more likely to assume the behavior of task group members is driven by common goals for physically distant rather than near groups. This effect occurred when perceived identification and similarity to others were held constant. Implications for intergroup relations are discussed.

The comparative psychophysics of complex shape perception

The authors compared the complex shape perception of humans and monkeys. Members of both species participated in a Same-Different paradigm in which they judged the similarity of shape pairs that could be variations of the same underlying prototype. For both species, similarity gradients were found to be steep going out from the transformational center of psychological space. In contrast, similarity gradients were found to be flat going from the periphery in toward the center of psychological space. These results show that there are important common principles in the shape-perception and shape-comparison processes of humans and monkeys. The same general organization of psychological space is obtained. The same quantifiable metric of psychological distance is applied. Established methods for creating controlled shape variation have the same effect on both species' similarity judgments. The member of the to-be-judged pair of shapes that is peripheral in psychological space controls the strength of the perceived similarity of the pair. The results have broader implications for the comparative study of perception and categorization.

When parameters collide: a warning about categorization models

Similarity-choice (S-C) models of categorization contain two principal mathematical transformations: an exponential-decay similarity function and a choice rule. However, there is a tension between the psychological processes that models emulate and the mathematics they use to do so. To illustrate this, I show that in these models an unappreciated interaction occurs between the mathematical transformations so that the stages of the model essentially cancel each other out. The result is that the model's output reflects its input linearly. This cancellation phenomenon has potentially serious implications regarding the interpretation and use of S-C models. The phenomenon also raises questions about the simplification and psychological grounding of categorization models. Modelers broadly might benefit from an internal analysis of their models, such as that described here.

Prototype-distortion category learning: a two-phase learning process across a distributed network

This paper reviews a body of work conducted in our laboratory that applies functional magnetic resonance imaging (fMRI) to better understand the biological response and change that occurs during prototype-distortion learning. We review results from two experiments (Little, Klein, Shobat, McClure, & Thulborn, 2004; Little & Thulborn, 2005) that provide support for increasing neuronal efficiency by way of a two-stage model that includes an initial period of recruitment of tissue across a distributed network that is followed by a period of increasing specialization with decreasing volume across the same network. Across the two studies, participants learned to classify patterns of random-dot distortions (Posner & Keele, 1968) into categories. At four points across this learning process subjects underwent examination by fMRI using a category-matching task. A large-scale network, altered across the protocol, was identified to include the frontal eye fields, both inferior and superior parietal lobules, and visual cortex. As behavioral performance increased, the volume of activation within these regions first increased and later in the protocol decreased. Based on our review of this work we propose that: (i) category learning is reflected as specialization of the same network initially implicated to complete the novel task, and (ii) this network encompasses regions not previously reported to be affected by prototype-distortion learning.

Correlations of cortical activation and behavior during the application of newly learned categories

Large individual differences are commonly observed during the early stages of category learning in both functional MRI (fMRI) activation maps and behavioral data. The current investigation characterizes this variability by correlating the volume of activation with behavioral performance. Healthy subjects were trained to classify patterns of random dots into categories. Training was carried out using a 4-choice categorization task with feedback. Functional MRI was performed prior to any training and then following each of 3 training sessions. The fMRI sessions involved the presentation of 3 separate paradigms which required the skill imparted by the training to determine whether two patterns of dots belonged to the same category. Contrasts between the 3 paradigms allowed the examination of the effects of training and of familiarity with the task. For fMRI performed with those materials used during training, increases in the volume of activation were observed initially. As behavioral performance continued to improve, reductions in activation were observed across regions involved in visuospatial processing and spatial attention. These reductions in activation were observed only for those materials used in training and only after high levels of performance were achieved. The magnitude of these reductions in activation correlated with each individual's own rate of learning. The present data support the observation that at least two stages of cortical activation underlie the use of newly learned categories. The first, recruitment of nearby tissue, is observed as initial increases in the volumes of activation. These initial stages of recruitment are followed by specialization across the same network which is observed as a reduction in activation with continued improvements in behavioral performance.

Is categorization performance really intact in amnesia? A meta-analysis

Most published studies of category learning in amnesia have reported intact categorization performance. These results have been used to challenge single-system accounts of categorization and recognition, in which a single representational system mediates performance in these two tasks. Many of the published studies, however, have shown a numerical advantage for controls over amnesics and often have had low statistical power. A meta-analysis was conducted to assess whether this numerical advantage is significant when the data are pooled across studies. This analysis indicates that amnesic subjects do, in fact, show deficits in categorization tasks, which is consistent with single-system exemplar model predictions.

Visual Search and the Collapse of Categorization

Categorization researchers typically present single objects to be categorized. But real-world categorization often involves object recognition within complex scenes. It is unknown how the processes of categorization stand up to visual complexity or why they fail facing it. The authors filled this research gap by blending the categorization and visual-search paradigms into a visual-search and categorization task in which participants searched for members of target categories in complex displays. Participants have enormous difficulty in this task. Despite intensive and ongoing category training, they detect targets at near-chance levels unless displays are extremely simple or target categories extremely focused. These results, discussed from the perspectives of categorization and visual search, might illuminate societally important instances of visual search (e.g., diagnostic medical screening).

Specific-Token Effects in Screening Tasks: Possible Implications for Aviation Security

Screeners at airport security checkpoints perform an important categorization task in which they search for threat items in complex x-ray images. But little is known about how the processes of categorization stand up to visual complexity. The authors filled this research gap with screening tasks in which participants searched for members of target categories in visual displays. The authors found that when targets were sampled with replacement and repetition, participant screeners relied on recognizing familiar targets and had great difficulty using category-general knowledge. The authors observed a "heartbeat" in detection performance--it improved while test images repeated but dropped sharply when unfamiliar targets from the same categories appeared. This reliance on familiarity illuminates the processes of categorization under conditions of visual complexity and suggests limits on those processes. This reliance also has implications for the training and evaluation of screeners in the field.

False prototype enhancement effects in dot pattern categorization

Results from the classic dot pattern distortion paradigm have sometimes yielded prototype enhancement effects that could not be accounted for by exemplar models of categorization. However, in these experiments the status of the prototype was confounded with certain stimulus-specific properties as well as with the frequency of presentation of the prototype during testing. In two mock-subliminal experiments, participants made categorization judgments to patterns that were generated as prototypes, low-level distortions, or high-level distortions. The participants rated the prototypes as being more likely to be members of a category, although no patterns were presented during training, and there was no objective category structure. In two other experiments, greater prototype enhancement effects were observed when the prototype and low-level distortions were presented with greater frequency during transfer. These results suggest that classic prototype enhancement effects may not be due to the abstraction of a prototype at time of original learning, but rather to other factors not formalized in extant models.

The cognitive neuroscience of category learning

Recently, a multidisciplinary approach has provided new insights into the mechanisms of category learning. In this article, results from theoretical modeling, experimental psychology, clinical neuropsychology, functional neuroimaging, and single-cell studies are reviewed. Although the results are not conclusive, some general principles have emerged. Areas localized in the sensory neocortex are responsible for the perceptual representation of category exemplars, whereas lateral and anterior prefrontal structures are necessary for the encoding of category boundaries and abstract rules. The prefrontal cortex may influence categorical representation in the sensory neocortex via top-down control. The neostriatum is important in stimulus-response mapping, and the orbitofrontal cortex/ventral striatum are related to stimulus-reward associations accompanying category learning. Many category learning tasks can be performed implicitly. In conclusion, category learning paradigms provide a unique opportunity to investigate cognitive processes such as perception, memory, and attention in a systematic and interactive manner. Category learning tasks are suitable for mapping damaged brain systems in clinical populations.

As easy to memorize as they are to classify: The 5–4 categories and the category advantage

Recently, it has been suggested that some categories commonly used in category learning research are eliciting primarily item-level memorization strategies. A new measure of generalization, the category advantage, was introduced and used to test performance on the popular "5-4" categories. To estimate a category advantage, performance on a standard category learning task is compared with performance in an identification task, where participants learn a unique response to each stimulus. Once corrected for differences in chance expectancy, the advantage shown for the category learning task represents the degree to which participants capitalize on the natural similarity structure of the categories. In Experiment 1, the category advantage measure was validated on structured and unstructured categories. In Experiments 2 and 3, the 5-4 categories failed to produce a category advantage when tested with either of two stimulus types, suggesting that these categories elicit predominantly memorization.

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.

Exemplar theory's predicted typicality gradient can be tested and disconfirmed

One of exemplar theory's central predictions concerns the shape of typicality gradients. The typicality gradient it predicts is a consequence of its exemplar-based comparisons and appears no matter how the theory is evaluated. However, this predicted typicality gradient does not fit the empirical typicality gradients obtained in an influential version of the dot-distortion category task, and this is true even when the exemplar model is made more flexible and mathematically powerful. Thus, exemplar theory is disconfirmed in this domain of categorization. In contrast, prototype theories are consistent with the empirically obtained typicality gradients.

The effect of 'masking' on picture naming

It is frequently assumed that because compared to nonliving things, living things are less familiar, have lower name frequency, and are more visually complex, this makes them more difficult to name by patients and normal subjects. This has also been implicitly accepted as an explanation for the greater incidence of living thing disorders. Patient studies do not, however, typically contain any premorbid data and so, we do not know that the same variables would have necessarily predicted their 'normal' performance. To examine this issue, we measured picture-naming latencies in normal subjects presented with unmasked and masked versions of the same line drawings. In accord with other recent studies, living things were named faster than nonliving things. Furthermore, contrary to some theories of category naming, the living thing advantage persisted regardless of whether stimuli were undegraded, degraded or the density of degradation. Finally, multiple simultaneous regression analyses showed that one visual variable (Euclidean Overlap) and one linguistic variable (Age of Acquisition) predicted naming latencies across all masked and unmasked conditions. Other variables either had no predictive value (Contour Overlap; Name Frequency; Category); predicted only high masking (Visual Complexity; Familiarity), or normal and low masking (Number of Phonemes). These findings imply that the more commonly documented deficits for living things do not reflect an exaggeration of the normal profile (be it with masked or unmasked stimuli) or the influence of the same variables that affect normal naming.