According to one productive and influential approach to cognition, categorization, object recognition, and higher level cognitive processes operate on a set of fixed features, which are the output of lower level perceptual processes. In many situations, however, it is the higher level cognitive process being executed that influences the lower level features that are created. Rather than viewing the repertoire of features as being fixed by low-level processes, we present a theory in which people create features to subserve the representation (...) and categorization of objects. Two types of category learning should be distinguished. Fixed space category learning occurs when new categorizations are representable with the available feature set. Flexible space category learning occurs when new categorizations cannot be represented with the features available. Whether fixed or flexible, learning depends on the featural contrasts and similarities between the new category to be represented and the individual's existing concepts. Fixed feature approaches face one of two problems with tasks that call for new features: If the fixed features are fairly high level and directly useful for categorization, then they will not be flexible enough to represent all objects that might be relevant for a new task. If the fixed features are small, subsymbolic fragments (such as pixels), then regularities at the level of the functional features required to accomplish categorizations will not be captured by these primitives. We present evidence of flexible perceptual changes arising from category learning and theoretical arguments for the importance of this flexibility. We describe conditions that promote feature creation and argue against interpreting them in terms of fixed features. Finally, we discuss the implications of functional features for object categorization, conceptual development, chunking, constructive induction, and formal models of dimensionality reduction. Key Words: concept learning; conceptual development; features; perceptual learning; stimulus encoding. (shrink)

Previous neural network models of concept learning were mainly implemented with supervised learning schemes. However, studies of human conceptual memory have shown that concepts may be learned without a teacher who provides the category name to associate with exemplars. A modular neural network architecture that realizes concept acquisition through two functionally distinct operations, categorizing and naming, is proposed as an alternative. An unsupervised algorithm realizes the categorizing module by constructing representations of categories compatible with prototype theory. The naming module associates (...) category names to the output of the categorizing module in a supervised mode. In such a modular architecture, the interface between the modules can be conceived of as an “information relay” that encodes, constrains, and propagates important information. Five experiments were conducted to analyze the relationships among internal conceptual codes and simple conceptual and lexical development. The first two experiments show a prototype effect and illustrate some basic characteristics of the system. The third experiment presents a bottom‐up model of the narrowing down of children's early lexical categories that honors mutual exclusivity. The fourth experiment introduces top‐down constraints on conceptual coding. The fifth experiment exhibits how hierarchical relationships between concepts are learned by the architecture, and also demonstrates how a spectrum of conceptual expertise may gradually emerge as a consequence of experiencing more with certain categories than with others. (shrink)

We adapted the Bubbles procedure [Vis. Res. 41 (2001) 2261] to examine the effective use of information during the first 282 ms of face identification. Ten participants each viewed a total of 5100 faces sub-sampled in space–time. We obtained a clear pattern of effective use of information: the eye on the left side of the image became diagnostic between 47 and 94 ms after the onset of the stimulus; after 94 ms, both eyes were used effectively. This preference for the (...) eyes increased with practice, and was not solely due to the informativeness of the eyes for the task at hand. The bias for the eye on the left side of the image is explained in terms of hemispheric specialization. Although there were individual differences, most participants exhibited this pattern of effective use of information. An intriguing finding is that most participants displayed a clear sinusoidal modulation of effective use of attention through time with a frequency of about 10.6 Hz. (shrink)

Does the perceptual processing of faces flexibly adapt to the requirements of the categorization task at hand, or does it operate independently of this cognitive context? Behavioral studies have shown that the fine and coarse spatial scales of a face are differentially processed depending on the categorization task performed, thus suggesting that the latter can influence stimulus perception. Here, we investigated the time course of these task influences on perceptual processing by examining the visual N170 face‐sensitive Event‐Related Potential (ERP), while (...) observers categorized faces for their gender and familiarity. Stimuli were full spectrum, or filtered versions that preserved either coarse or fine scale information of the faces. Behavioral results replicated previous findings of a differential processing of coarse and fine spatial scales across tasks. In addition, the N170 amplitude was larger in the Gender task as compared to the Familiarity task for LSF faces exclusively, thus showing that task demands differentially modulated the spatial scale processing on faces. These results suggest that the diagnosticity of scale‐specific cues in categorization tasks can modulate face processing. (shrink)

The relationship between perceptual categorization and organization processes in 3‐ to 4‐month‐old infants was explored. The question was whether an invariant part abstracted during category learning could interfere with Gestalt organizational processes. Experiment 1 showed that the infants could parse a circle in accord with good continuation from visual patterns consisting of a circle and a complex polygon. In Experiments 2 and 3, however, this parsing was interfered with by a prior category familiarization experience in which infants were presented with (...) visual patterns consisting of a pacman shape and a complex polygon. Part 1 of Experiments 2 and 3 showed that the infants recognized the pacman as familiar, and Part 2 demonstrated that the representation of the pacman blocked the subsequent parsing of the circle. The results suggest that a cognitive system of flexible feature creation can override organizational principles with which a perceptual system may come pre‐equipped. (shrink)

Pylyshyn argues against representations with pictorial properties that would be superimposed on a scene. We present evidence against this view, and a new method to depict pictorial properties. We propose a continuum between the top-down generation of internal signals (imagery) and the bottom-up signals from the outside world. Along the continuum, superstitious perceptions provide a method to tackle representational issues.

Pylyshyn acknowledges that cognition intervenes in determining the nature of perception when attention is allocated to locations or properties prior to the operation of early vision. I present evidence that scale perception (one function of early vision) is cognitively penetrable and argue that Pylyshyn's criterion covers not a few, but many situations of recognition. Cognitive penetrability could be their modus operandi.

The origin of features from nonfeatural information is a problem that should concern all theories of object categorization and recognition, not just the flexible feature approach. In contrast to the idea that new features must originate from combinations of simpler fixed features, we argue that holistic features can be created from a direct imprinting on the visual medium. Furthermore, featural descriptions can emerge from processes that by themselves do not operate on feature detectors. Once acquired, features can be decomposed into (...) component features if required by other categorizations. We therefore argue that it is not necessary to separate holistic and componential approaches to representations, because the latter is a development of the former. The requirements for representational flexibility outstrip the performance of any existing computational models, but specific mechanisms of feature creation are discussed and evaluated. Challenges for feature creation mechanisms are discussed together with the constraints (perceptual, statistical, functional, and task) they will need to satisfy. (shrink)