THE DYNAMICS OF CAPACITY ALLOCATION IN SINGLE AND DUAL TASK PERFORMANCE

AARON BARTELL, Purdue University

Abstract

This thesis examined the performance tradeoff functions (i.e., Performance Operating Characteristics, or "POCs") between simultaneously executed (i.e., dual) tasks. The tasks employed in the thesis experiment were reciprocal tapping (T) and arithmetic digit-naming (D). Performance tradeoffs in timesharing conditions were induced with instructions ("75%T,25%D"; "50%T,50%D"; and "25%T,75%D") to differentially allocate resources between the component dual tasks. Each experimental task was also performed alone (single task condition) as a control measure. A family of POCs was generated by varying the level of difficulty (in tapping) or complexity (in digit-naming) as well as the pacing combination of the experimental tasks. All the POCs resulting from the experimental manipulations described above were simulated with formal mathematical realizations of the central capacity model (e.g., Broadbent, 1971, and Kahneman, 1973) as well as the multiple capacity model (e.g., Navon & Gopher, 1979). The "best" fit of the experimental data obtained from an empirically derived model structure was extremely efficient (generating a chi-square fit of 3.694 with df = 47: p (DOT=) 1) and required only one parameter to be estimated. This most efficient empirical structure, originally generated as a weak version of the central capacity model (in which the assumption of inefficient resource utilization is permitted), was shown to also be consistent with several other types of model formulations: a strong version of the central capacity model (in which efficient resource utilization is assumed) as well as strong and weak versions of the multiple capacity model. Performance Operating Characteristics were assumed to be linear in the a priori specification of most realizations of the central and multiple capacity models. However, the POCs were empirically found to exhibit several nonlinearities (e.g., a local plateau indicating unused capacity in dual task conditions as well as an absolute performance plateau for one of the unpaced single tasks). In addition to the empirical finding of POC nonlinearities, subjects appeared to only nominally follow the differential resources allocation instructions discussed earlier (e.g., performance tradeoffs between dual tasks undershot the range demanded by experimental instructions). Also, it appears that subjects did not utilize total capacity within any unpaced (or paced) single task condition (thus implying that single task performance should not necessarily be considered as a limiting case of a POC). Finally, maximum performance levels were estimated in all model formulations to be lower for the paced variant than unpaced variant of each experimental task, thus indicating that the processes necessary to pace a task are probably capacity demanding. The implications of these empirical findings for timesharing research in general are discussed in the last chapter of this thesis.

Degree

Ph.D.

Subject Area

Psychology|Experiments

COinS