Serial vs. Parallel Processing: Sometimes They Look like Tweedledum and Tweedledee but They Can (And Should) be Distinguished

A number of important models of information processing depend on whether processing is serial or parallel. However, many of the studies purporting to settle the case use weak experimental paradigms or results to draw conclusions. A brief history of the issue is given along with examples from the literature. Then a number of promising methods are presented from a variety of sources with some discussion of their potential. A brief discussion of the topic with regard to overall issues of model testing and applications concludes the paper. Somewhat informally, serial processing means strictly sequential, without overlap of the successive processing times on objects or distinct subsystems. In a standard type of serial system, each object takes the same average amount of time to process and the next object begins processing only when the previous one is completed. On the other hand, parallel processing signifies simultaneous processing on several objects or subsystems at the same time, although processing may finish on different objects at different times. In either type of operation, both individual and overall processing times may be random. That is, the durations required for processing an item or performing an operation may vary from trial to trial. This paper is about testing parallelism vs. seriality. The question as to whether and when people can perform perceptual or mental operations in parallel began to receive experimental treatment in the late 19th century, although not under these names. It was a natural question for the emerging discipline of psychology because it is inherently related to the capacity of mind and how that capacity is allocated to sundry cognitive and perceptual endeavors. Perhaps Hamilton (1859) was the first to attempt an empirical, if hardly experimental, answer to the question. One of his techniques was to toss several dice on his desk and try to assess "instantaneously" the number of dots showing. The intent was to determine the number of objects that could be apprehended simultaneously (i.e., in parallel) by human consciousness. This interest reappeared in various guises in the emerging psychological laboratories of the world. It remains a lively research topic today in diverse areas of pure and applied cognitive psychology. The longevity of the topic is probably due to its fundamental importance in describing how mental operations take place. As intimated above, this topic is also closely connected with the issue of capacity; that is, to what extent mental processing of some type suffers when the number of things to do mentally or the difficulty of the cognitive operations increases. For instance, standard serial processing with each successive subtask taking the same average duration is of limited capacity with respect to the overall total processing time required for an increasing number of subtasks. That is, the overall reaction time for all the subtasks increases, the more tasks there are to do. However, the same serial processing is of unlimited capacity on individual items in the sense that the average item processing duration per item is constant regardless of the total number of items to be done. Parallel processing can be either limited or unlimited capacity on either the individual item or on the whole set, the difference depending on the type of parallel system in question (e.g., Townsend, 1974a, Townsend & Ashby, 1983). l This paper will advocate the view that contemporary research on the parallel-serial question often uses methodology or logic that was shown to be faulty or at least precarious twenty or more years ago. Also, it is argued that few investigations take advantage of more powerful techniques of testing the dichotomy that have been developed since 1968. Several pertinent examples will be given from the literature in perception and cognition. In the following section, a number of promising methods that can be mathematically demonstrated to separate large classes of parallel vs. serial models will be collated from the literature and explained. HISTORICAL PERSPECTIVE A very brief, limited survey of the recent history of the issue 1. We are presently concerned strictly with models that may be referred to as traditional serial or parallel models. Introductions to certain other types of models may be found in Meyer, Irwin, Osman & Kounios (1988) and Ratcliff (1988) along with many references. More technical information on the traditional and some of the more recent models can be found in Luce (1986) and Townsend & Ashby (1983). Correspondence and offprint requests to: James T. Townsend, Department of Psychology, Indiana University, Bloomington, IN 47405. 46 Copyright © 1990 American Psychological Society VOL. 1, NO. 1, JANUARY 1990 PSYCHOLOGICAL SCIENCE