FORMATTING EFFECTS ON THE USE OF COMPUTER-GENERATED ALPHANUMERIC DISPLAYS: THE MODERATING EFFECTS OF TASK CHARACTERISTICS (SEARCH, SCANNING, HUMAN PERFORMANCE, PREDICTION, FORMAT)
SCHWARTZ, DAVID ROBERT
Doctor of Philosophy
Tullis (1983, 1984) identified and quantified six conceptually distinct dimensions of alphanumeric display formats. On the basis of these dimensions, he derived regression equations which predicted performance and user perceptions in a simple search task. The current study sought to extend his findings to other common tasks performed with visually coded information--where the use of several pieces of information from predictable display locations was required. Further, complexity and visual monitoring load were manipulated to study the effects of the Tullis dimensions in a broader task context so that any boundary conditions might be identified. While there did not appear to be any clear effects of format once information was extracted from the display, the time taken to extract the information did vary with the format dimensions. Thus, their importance was not diluted by the predictability of the displays. The relative merits of alternative formats in the Scanning task, however, were not predicted successfully by the Tullis equations. This prediction failure was explained post hoc as being due to an enhancement and reversal of the effect of the Item Uncertainty dimension when display locations are known by users. Further research designed to test this hypothesis directly will be necessary to determine whether an alternative prediction system might be useful for such task situations. In contrast, the Tullis equations did predict the perceived difficulty associated with alternative formats fairly well in all task conditions. Subjects' perceptions seemed to be based largely on the ease with which information was extracted regardless of other task considerations (such as their actual level of performance). Since the regression equations were based on user perceptions of the ease of search, their predictions appear to be robust to major differences in task requirements. Finally, it was suggested that subjects' apparent inability to evaluate their performance accurately under alternative designs should serve as a warning to system designers. That is, where performance is the paramount design criterion, empirical human performance research should be conducted in the absence of a validated prediction system such as the one developed by Tullis for search.