MENTAL CHRONOMETRY: Beyond Reaction Time

Details of response e.xecution were examined in two classic human information processing paradigms: lexical decision and memory scanning. In the lexical decision experiment, word frequency influenced both the time of response onset and kinematic properties of the response. In the memory scanning experiment. when the probe was present in the memory set, the responses were both initiated sooner and were more forceful when the memory set consisted of two items than when it consisted of six items. When the probe was absent from the memory set, responses were initiated sooner but were less forceful when the memory set consisted of two items than when it consisted of six items. The results suggest that the amount of activation in support of a given response can modulate both the time taken to initiate the response and the force with which the response is executed. These findings bear on all models of human cognitive performance that have been developed within the mental chronometry tradition. Over 100 years ago. Donders (1868-1869/1969), a Dutch physiologist, developed a method to measure what he called "The Speed of Mental Processes." His procedure was quite elegant. Donders recorded the time people required to complete two simple tasks that were identical with the exception of a single additional mental operation inserted into one of the tasks. He argued that the difference in completion times for the two tasks could be used as an estimate of the duration of the target mental operation. In this classic work, Donders derived estimates for the durations of the mental processes involved in stimulus identification and response selection. Since Donders' original work, there has been a considerable increase in sophistication regarding the inferences that are and are not possible from response latency data (see Meyer, Osman, Irwin, & Yantis, 1988, for a review). Despite the diversity of the mechanisms proposed to account for response latency data, all information processing models make an important assumption: Functionally early processes such as stimulus identification and response selection are assumed to be completed prior to the onset of an overt response. The response is then believed to be ballistically "triggered" when such processing is completed. Thus, the primary interest in studies of mental chronometry is the influence of factors on when a response is initiated, not how the response is executed. Address correspondence to either author at Department of Psychology, Washington University, St. Louis, MO 63130. e-mail: C39823RA@WUVMD. bitnet. Recently, however, there has been evidence accumulating that characteristics of the response itself can be influenced by the same factors that affect the earlier stimulus identification and response selection processes. For example, Balota, Boland, and Shields (1989) demonstrated that the duration of a pronunciation can be influenced by some of the same manipulations that affect the time to initiate the pronunciation. Also, Osman, Kornblum, and Meyer (1986) have shown that response latency can be affected by factors that have their impact after the point in processing when response selection has already occurred (i.e.. after the "point of no return"). Finally, Coles et al. (1985) have shown that activation may grow continuously prior to the production of an overt response in a two-choice reaction time paradigm, sometimes resulting in covert muscular activity that favors the incorrect response. These results suggest that the parameters involved in the execution of a response may be affected by factors that previously were believed to influence only the earlier stimulus identification and response selection stages. The present experiments extend these recent demonstrations in two ways. First, we investigated the extent to which kinematic features of a response may be related to response latency by using an arbitrary response (a limb movement) that is not inherently tied to the factors being manipulated. This approach eliminates some limitations of previous work. For example, in the Balota et al. (1989) study described above, a linguistic variable (associative context) was shown to influence the duration of a linguistic response (naming duration). But stress, a high correlate of duration, can provide useful cues to a listener in speech. Hence, the response used in the Balota et al. study was not arbitrary with respect to the manipulation, and the findings may not generalize beyond the domain of speech production. Second, we studied a single graded response under stimulus conditions that unequivocally defined the response. In the Coles et al. (1985) study, described above, subjects responded to centrally presented target letters {H or 5) by squeezing handles with either their right or left hands. Targets could be flanked by either compatible stimuli {HHHHH) or incompatible stimuli (SSHSS). They found that squeeze responses were stronger when the flanking stimuli were compatible with the target than when they were incompatible. However, as noted above, they also found that incompatible stimuli produced partial squeeze responses in the incompatible response channel (i.e., the opposing hand). Thus, as Coles et al. claimed, their results reflect response competition processes. In the present experiments, there were no characteristics in the stimuli (e.g.. flanking incompatible letters) that simultaneously led to incompatible responses. VOL. 2, NO. 3, MAY 1991 Copyrighl © 1991 Americim Psycho!'-= > .A Soct'.-lv PSYCHOLOGICAL SCIENCE