Issues and trends in the debate on discrete vs. continuous processing of information

Abstract

The paper discusses various arguments and lines of evidence with respect to discrete stage and continuous flow models of choice reaction processes and in particular to the applicability of the Additive Factor Method (AFM) which has always been strongly tied to discrete processing. The paper starts with a brief outline of Miller's three different aspects of ‘discrete’ and ‘continuous’ (internal code, transmission, transformation). It is argued that application of the AFM does not require discrete processing in all respects. Discrete internal codes are essential, discrete transmission is preferable but not absolutely necessary, while continuous transformation can be easily accommodated. Most research on the AFM shows consistent and robust evidence in favor of seven successive processing stages in traditional choice reactions. Some inconsistencies can be satisfactorily explained by effects of confounded variables; yet, there also remain some inconsistencies, in particular with respect to one-dimensional processing, which suggest limits of the AFM. A critical issue is speed-accuracy trade-off (SAT). Here, the decomposition technique proposed by Meyer et al. (1988a,b) offers a promising tool for studying discrete versus continuous processing. A second critical issue, basically at stake in selective attention paradigms, is what constitutes a signal. Finally, advocates of continuous flow often refer to psychophysiological research and in particular to evoked potentials (EP). It is argued that the relations between EP and behavioral research seem insufficiently clear-cut to allow conclusions on discrete stages versus continuous flow. In conclusion, discrete stage models should not be rejected too hastily. They tend to have stronger axioms and hence allow stronger predictions. Unless more explicit constraints are specified, continuous flow models have much less predictive power.