|
Primary Reviewer: Jessica Goldstone, Robert L.; Barsalou, Lawrence W (1998) Reuniting perception and conception. Cognition, 65, 231-262. Purpose The purpose of this paper was to describe ways in which conceptual thought is grounded in perceptual similarity. Historically, there has been a division between conceptual and perceptual systems suggesting that perceptual similarities are not used when creating categories. Goldstone and Barsalou state that conceptual processing shares computational resources with perception, suggesting that a common representational and processing system underlies both. The authors discuss two approaches for reuniting perception and conceptualization. The first is the eliminative view in which human knowledge contains no non-perceptual representations. The primary existence of the eliminative view is to serve as an existence proof that one can develop a fully functional symbolic system that is inherently perceptual. Earlier approaches assumed that aspects of perceptual states become stored in memory to form concepts. The eliminative view assumes that perceptual states become amodal symbols. Amodal symbols are assumed to bear arbitrary relations to perception and to their referents in the world, much like the relationship between words and their referents. The agnostic view differs from the eliminative view in its strictness. It suggests that human knowledge has major perceptual components and may or may not contain non-perceptual components. The agnostic view states that these amodal symbols may exist, but are not always necessary. This second approach, the agnostic view, is the main concentration of the paper. Their position in the agnostic view stresses the evolution of concepts from perception through processes that can eventually achieve abstract end states. There is strong evidence for this idea in studies that show that children rely on perceptual representations when developing their concept of numbers. Goldstone and Barsalou start by discussing the efficiency of perceptual representations versus the complications of amodal representations. The alternative to computing with perceptually based representations is to reason from symbolic representations that either completely remove perceptual information or start with a symbolic representation. However, representations that preserve physical properties are likely to be more efficient than symbolic representations because they do not require external constraints such as inference rules to ensure proper inferences. Therefore, the authors conclude, it makes sense to preserve physical perceptual properties rather than to try to mimic rules and constraints in amodal representations. Processing similarity is another area in which the authors believe the agnostic view is more efficient than the eliminative view. They break similarity down into two parts, overall similarity and learned similarity. Sekuler and Abrams investigated overall similarity and showed that people are faster to respond that two objects are identical along all of their elements than that two objects have a single common element. This tendency to process similarity holistically has its origins in perception. The distinction between perceptual and conceptual processes is related to the associative/rule distinction. Associative mechanisms process bursts of features made available automatically through parallel processing in perception, or automatic activation in long-term memory. Rule mechanisms process individual features selectively through serial processing in working memory using limited attentional resources, which the authors liken to amodal processing. Rapid categorization is possible when multiple perceived properties can be matched in parallel to potentially corresponding properties in category knowledge. Goldstone and Barsalou claim that similarity is more efficiently processed by perceptual systems than amodal symbolic systems based on two facts. First, the processes that allow parallel processing without a capacity limit are found in perceptual systems. Second, perceptual systems often blend multiple sources of information together without individuating sources. Learned perceptual similarity has been studied in several ways. One area of categorical perception focuses on discriminations involving pairs of stimuli that straddle category boundaries. In these studies, the results consistently show that these distinctions are more easily made than are discriminations involving stimuli that fall within the same category. Other studies have found that perceptual processing is slower to change than higher-level processing. Task specific needs will not typically modify perceptual systems permanently but if permanent changes are made, they may change the perceptual properties that are noticed. For instance, if you are a short person and often need a boost to reach high shelves, you may eventually come to notice the feature of being a stool-like object in items not normally considered stools. Perception, but not necessarily conception, is helpful in that it can be sensitive to contextual factors and tune into an organism’s needs. Goldstone and Barsalou then go on to discuss perceptual simulation in conceptual tasks. They say that perceptual simulation (thinking of an image when you’re not required to) is central to conceptual processing. Even when subjects are asked to not use mental imagery, they spontaneously produce images rather than access amodal representations to solve conceptual problems. Wu showed this in subjects that simulate the referents of nouns and noun phrases to list features and that the feature’s visibility in real-world referents predicted its likelihood of being produced. Finally, the authors mention parallels between perceptual and conceptual processing. They say that properties associated with abstract cognition are often present in perceptual systems. They believe that perceptual systems have mechanisms that are useful for more abstract cognition and provide insights into how higher-order cognition may operate. They highlight six properties that they feel the two systems share: selectivity, blurring and filtering, structure and binding, differentiation, subcategories, and dimensions, cross modal matching, and productivity. Selectivity- Cognitive and perceptual processes are both characterized by selective attention to relevant properties and selective inhibition of irrelevant properties. Many properties of perceptual attention make it a promising candidate for subserving situations where more abstract conceptual highlighting of properties is needed. Evidence exists for the cooption of perceptual selective attention by cognitive selective attention in such cases as schizophrenia where patients show impairments on visual tasks before cognitive impairments arise. This is also supported by studies of field dependent personalities that find that selective attention operations are shared by perceptual and conceptual processes. Blurring and filtering- Blurring is any process that permits processing of detailed information. Abstraction is typically accomplished by distilling the essence from its superficial trappings. This is accomplished by developing a schema that is tuned to the essence while the irrelevant aspects are blurred over or ignored. Blurring can operate even when schema cannot be formulated. Visual agnosias provide support for categorizing by blurring. Amodal abstractions must represent or not represent a property, but perceptual attention allows for partially representing a property. Structure and binding- Binding of objects to roles establishes a structure in propositions beyond representations such as features. For example, in perception, elements from the left and right eye are bound together to create a cohesive scene. Most perceptually-motivated models of binding provide a dynamic and temporally limited mechanism for establishing structures. In these models, a one-to-one binding is rare. Perception requires structured binding as much as does abstract cognition. Mechanisms of perceptual binding may better explain cases of intermediate degrees of binding than do amodal propositions. Differentiation, subcategories, and dimensions- Differentiation is a cognitive process in which people divide a category into subcategories. To dimensionalize is to break apart dimensions that were originally fused and is a perceptual process. The primary reason to believe that cognitive differentiation and perceptual dimensionalization share fundamental mechanisms is that it is hard to make a distinction between dimensions (features) and concepts. Cross-modal matching- Cross-modal matching is people’s natural tendency to link distinct sensory domains such as loudness, pitch, and color together. An ability to reason conceptually may borrow from perceptual processes that underlie cross-modal matching. Productivity- Productivity is the combining of small units to create larger structures which produces the creativity usually associated with amodal systems. However, perceptual representations can support an equivalent operation called spatio-temporal juxtaposition. The authors summarize these six parallels by saying that many properties of abstract cognition, when explored from the perspective of processes that could furnish them, are also found in perceptual systems. Both abstract cognition and perceptual processes are abstractive, have an argument structure, permit analysis into components, and allow for productivity. These similarities arise not through coincidence, but because they share important mechanisms. Finally, two distinctions are discussed. The perceptual/conceptual distinction is the idea that there may be a continuum from perceptual to conceptual with how much and what sort of top-down processing has been done to bottom-up input information varying. The authors endorse this view. They do not endorse the perceptual/abstract distinction because they believe it is misleading to equate perceptual processes to superficial processes. They believe that perceptual aspects are at least cues for the abstraction that is built and are often never removed from the abstraction. Conclusions Goldstone and Barsalou endorse the perceptual/conceptual distinction but they feel that being conceptual is being perceptual. They do not endorse the perceptual/abstract distinction because they deny the requirement that abstract knowledge be amodal. The authors believe that perceptual processes vary in how much they transform the original sensory input, but concrete aspects of the input are rarely, if ever, discarded completely. Instead, there is a continuum of distance from sensory input, with amodal symbols constituting an ideal endpoint. They believe that amodal symbols should be kept in models for two reasons. The first is that if there is some necessary function of intelligence that amodal symbols can accomplish that perceptual ones cannot, then they should be maintained. The second is that amodal symbols may deliver certain conceptual functions more efficiently than perceptual symbols. Their bottom line is that perceptual mechanisms underlie conceptual processing to a considerable degree. Points for Discussion One of the important premises that Goldstone and Barsalou build from in this paper is that perceptual systems process in parallel, but conceptual systems process serially. From this, they reason that perceptual systems process faster than conceptual systems and are therefore, inherently more efficient. Do you feel that they have supported this claim or simply pulled the processing divisions from thin air? Goldstone and Barsalou state that one of the advantages of the perceptual system is that it is adaptable to context and a person’s needs. They say that the perceptual system can be modified through practice to search features in ways that are temporally necessary. Is there any evidence given for why people cannot search their amodal representations in a similar way or does this relate to their claim that the conceptual system processes serially? To believe or not to believe this statement… “The primary reason to believe that cognitive differentiation and perceptual dimensionalization share fundamental mechanisms is that it is hard to make a distinction between dimensions and concepts.” Does this statement seem sufficiently reasoned?
|
|
This page was last updated: |