DISCRETE VERSUS CONTINUOUS MODELS OF VISUAL PERCEPTION

点击翻译按钮获取中文摘要

One of the most basic assumptions in cognitive psychology is that speeded decision making or "choice reaction time" (RT) tasks are carried out using a sequence of distinct mental processes or "stages" like perception, decision, and response execution. Traditionally, the field has attempted to identify and describe such stages, specifying each in terms of its input information, operations performed, and output produced. Recently, many researchers have begun to ask how these stages fit together to perform complex tasks, and how they communicate with one another. Two quite different types of models, commonly referred to as discrete and continuous, have been suggested (e.g., McClelland, 1979; Sternberg, 1969). It is crucial to determine which of these models better describes performance in human information processing tasks. The development of sophisticated models for normal cognitive processes is a prerequisite for understanding cognitive deficits, including those resulting from aging, injury, and disease. At present, such model development is hindered by fundamental uncertainty about whether discrete or continuous models more accurately describe normal cognition. Although both models are compatible with available evidence about human performance, they are fundamentally different in operation, and they prescribe very different strategies for conducting new experiments and interpretating the results. The proposed research will investigate two critical differences between discrete and continuous models. One is that discrete models require stages to operate in a strictly sequential manner. Each stage begins when it receives input, processes until it is finished, and then transmits a single output to the next stage in the processing sequence. Conversely, continuous models allow successive stages to operate at the game time (i.e., with some "temporal overlap"), even when the output of one stage is the input to the next. A stage transmits preliminary partial output to its successor, thereby allowing the later stage to begin before the earlier one is finished. The second critical difference concerns the nature of the outputs transmitted from stage to stage. According to discrete models, outputs are discrete ("all-or-none") activations of categorical internal representations. According to continuous models, however, stage outputs vary gradually in strength or "activation level", and possibly in other respects as well. The immediate aims of this research are to develop methods for finding out whether (1) the different stages used in performing choice RT tasks operate in strict sequence, and (2) the output of each stage is all-or-none rather than graded. The proposed methods use convergent measures of response time, response force, and psychophysiological measures derived from EEG activity, and will be used to examine human information processing in a number of scientifically important cognitive tasks, especially those involving attention.