A Theory of Interactive Parallel Processing: New Capacity Measures and Predictions for a Response Time Inequality Series.

Abstract

The authors present a theory of stochastic interactive parallel processing with special emphasis on channel interactions and their relation to system capacity. The approach is based both on linear systems theory augmented with stochastic elements and decisional operators and on a metatheory of parallel channels' dependencies that incorporates standard independent and coactive parallel models as special cases. The metatheory is applied to OR and AND experimental paradigms, and the authors establish new theorems relating response time performance in these designs to earlier and novel issues. One notable outcome is the remarkable processing efficiency associated with linear parallel-channel systems that include mutually positive interactions. The results may offer insight into perceptual and cognitive configural-holistic processing systems.When a person views a work of art in all its complexity, it seems as though all dimensions-color, form, arrangement, perspective, and sharpness of edges-are acting in league. But are they? The antithetical notions of independence and dependence have long played a role in the philosophy and science of human perception and cognition. Whether the focus is on the internal representations that support psychological experience or on the channels or systems that work with those representations, eventually the implications of assumptions about independence or dependence need to be considered. 1It is natural to contrast the general notion of independence with ideas such as a holism or a gestalt. However, there have been very few attempts to develop explicit, systematic, quantitative definitions, taxonomies, and explanations of holistic cognitive processes. The work presented here is an attempt to provide these definitions, taxonomies, and explanations, by way of (a) an emphasis on the general characteristics of information processing and (b) a concern with real-time interactions in specific (parallel) informationprocessing architectures.It is the case that certain interactive architectures have been shown to be capable of simulating specific gestaltlike phenomena (e.g., Biederman & Kalocsai, 1998;Cottrell, Dailey, Padgett, & Adolphs, 2001;Grossberg, 1991b;Mordkoff & Yantis, 1991;Rumelhart & McClelland, 1981. Although such studies provide sufficiency arguments, it is not clear what highly simple (in a sense to be defined precisely) and interactive systems can do when compared, within a common framework, with noninteractive systems.Nevertheless, the great preponderance of quantitativecomputational accounts of processing multidimensional stimuli are based on the assumption of independence among channels or items (e.g

Parallel interactive retrieval of item and associative information from event memory

“…For example, while the analytic form of the SIC is known for some classes of coactive system (Townsend & Nozawa, 1995;Fific et al, 2008;, there as yet is no general proof that covers all coactive systems, let alone those with more complex interactions of the kind we considered. While the converging evidence from our various analyses and their relation to other results in the literature (as discussed both above and below) puts our conclusions regarding memory on firm footing, our efforts also contribute to the ongoing project to understand the qualitative signatures of different cognitive systems and to develop experimental means to detect them (Townsend & Wenger, 2004;Eidels et al, 2011;Little et al, 2011;Yang et al, 2014).…”

“…In both cases, the output is a set of signals reflecting the degree to which each type of information matches or mismatches the contents of memory. Item and associative match information could be mutually facilitatory in two ways: first, by cross-talk between between two otherwise separate match accumulators (Townsend & Wenger, 2004); second, by computing a match using a compound cue that combines both item and associative information and compares them jointly to memory (Ratcliff & McKoon, 1988;Dosher & Rosedale, 1989. In the first case, item and associative matches are still processed separately but, according to our results, these processes communicate so strongly with one another that they effectively act as a single process.…”

“…Unfortunately, it is unlikely that one can experimentally distinguish strongly-interactive-but-separate processes from a single holistic process, though this does not preclude testing specific implementations of these two types of facilitatory model (Colonius & Townsend, 1997). One possible test relies on the fact that a holistic match can only yield a decision on the basis of combined evidence whereas separate channels can result in separate decisions; as a result, a holistic match should yield similar performance regardless of the decision rule whereas separate processes could yield different performance (Townsend & Wenger, 2004). The fact that associative information intrudes even when decisions need only be based on items (Tulving & Thompson, 1973;Clark & Shiffrin, 1987) and vice versa (Buchler et al, 2008) suggests that a holistic match is the more likely of these two candidates.…”

Predicting Search Performance in Heterogeneous Visual Search Scenes with Real-World Objects

“…A potential avenue for further testing this facilitative interaction between homogeneous items is through the use of the capacity coefficient (Townsend & Wenger, 2004; for an example of application, see Godwin, Walenchok, Houpt, Hout, & Goldinger, 2015), which could provide more direct evidence for the violation of the independent processing in stage one.…”

Information-processing architectures in multidimensional classification: A validation test of the systems factorial technology.

“…One possibility involves a very general class of models in which separate dimensions facilitate one another through parallel channel interactions. In this case, information might be shared, or activation on one channel might cross over and help activate a separate channel (Mordkoff & Yantis, 1991;Townsend & Wenger, 2004b). Unlike true coactivation models, this kind of processing still involves separate detection criteria operating in each channel.…”