Human information processing efficiency in a dual-task paradigm

Gladstones, Regan, and Lee (1989) have presented a dual-task paradigm which they claim enabl JS a strict test of the single-channel hypothesis. In this paradigm, subjects perform two serial choice reaction time (RT) tasks, first separately, and then simultaneously, at their maximum sustainable rates of information processing. The tasks are statistically independent, forced-paced, and do not permit subjects to group signals into chunks. In two experiments, Gladstones et al. found that the time taken by subjects to perform two tasks simultaneously was approximately equal to the sum of the times required to perform each task separately. Thus, derived rates of information processing in bits per second were no higher for dual-tasks than for single-tasks (averaged). This finding was unaffected by (a) whether both tasks were two-choice or both were three­ choice, (b) whether inputs and/or outputs involved the same or different modalities, or (c) levels of stimulus-response (S-R) compatibility. Gladstones etal. presented these findings as strong support for the single-channel hypothesis. The aim of the present study was to explore the application of their paradigm in a range of other situations to test the generality of their findings. Six experiments are reported. The first three experiments, Experiments lA, IB, and IC, were designed to test some critical assumptions underlying the paradigm. The tasks in each of these experiments were identical, each requiring subjects to respond manually to the colour of a small circular light. Experiment lA attempted to replicate one of the findings obtained by Gladstones et al. (1989) for two non-redundant three-choice tasks and to confirm this finding over more than one level of task difficulty. Task difficulty was manipulated by varying the number of stimulus alternatives. When both tasks conveyed three, four, or five stimulus alternatives, there was no overlap between them, that is, the time taken by subjects to perform the tasks simultaneously was no less than the sum of the times required to perform each task in isolation. This finding replicated and confirmed (over two additional levels of difficulty) the findings of no overlap obtained by Gladstones et al. There was considerable overlap, however, between two-choice tasks. This conflicted with a finding reported by Gladstones et al. Evidence suggested that the two-choice tasks in Experiment lA were integrated into a single four-choice task. The aim of Experiment IB was to determine why there was overlap between two­ choice tasks in the previous experiment but not in the Gladstones et al. (1989) study. It was hypothesized that task integration may have been discouraged in the latter study because the mappings of colours to fingers for two-choice tasks in that study were different (mappings were the same in Experiment lA). The two-choice condition in Experiment lA was replicated. Two further conditions were included in which the colour­ to-finger mappings for tasks were different. Considerable overlap occurred in all three conditions, although significantly less overlap occurred in the latter two conditions. The results suggested that in all conditions tasks were integrated into a single four-choice task. It was hypothesized that the spatial arrangement of the task response keys in Experiments lA and IB, but not in the Gladstones et al. study, may have facilitated the spontaneous integration of two-choice tasks. Experiment 1C sought to replicate the finding of no overlap obtained for four-choice tasks in Experiment lA. The sequential redundancy of tasks was systematically manipulated in order to determine its effects on degree of overlap. No overlap was found when both were non-redundant, replicating the findings obtained for four-choice tasks in Experiment lA. Similarly, but unexpectedly, there was no overlap between tasks when one was perfectly predictable. Evidently, the particular combination of experimental controls in the present paradigm prevented subjects from utilizing sequential redundancy. Complete overlap occurred between tasks when both were perfectly predictable (the time taken by subjects to perform the tasks simultaneously was no longer than the time required to perform either task separately). Here, tasks were integrated into one. The last three experiments, experiments 2A, 2B, and 2C, tested the generality of the Gladstones etal. (1989) findings. Experiment 2A sought to replicate, using the present paradigm, a finding of almost complete overlap obtained by Schvaneveldt (1969). Subjects responded with hand and voice to the position and colour, respectively, of an illuminated light. As in Experiment lA, the difficulty of the tasks was manipulated by varying the number of stimulus alternatives. Almost complete overlap was found in all conditions, replicating Schvaneveldt's main finding. This conflicted with the findings of no overlap obtained in Experiment 1A. More overlap occurred between five-choice tasks than between two-choice tasks. There was, however, no significant variation in the interresponse interval (IRI), implying that response grouping had occurred. Experiment 2B was designed to eliminate response grouping. Component tasks were similar to those in Experiment 2A. However, rather than responding in dual-task trials to two attributes (position and colour) of an illuminated light, subjects responded to the single attribute (position or colour) of each of two physically separated lights. As in previous experiments, task difficulty was manipulated by varying the number of stimulus alternatives. Response grouping was eliminated. However, considerable overlap was still found in all conditions (although this did not vary with task difficulty). These findings, like those in the previous experiment, conflicted with the findings of no overlap obtained in Experiment lA. The aim of Experiment 2C was to determine to what extent separation of response modalities in Experiments 2A and 2B may have contributed to the differences in degree of overlap observed between each of those experiments and Experiment lA (in which response modalities were the same). The experiment also sought to determine whether differences in levels of S-R compatibility and stimulus-central processing-response (S-C­ R) compatibility of tasks between Experiment 2A and Experiment lA contributed to differences in the degree of overlap between those two experiments. Two main findings emerged from the experiment. First, the use of identical rather than different modalities of response resulted in significantly less overlap when subjects responded to two attributes of the same stimulus but not when they responded to the single attribute of two physically separated stimuli. This implied that the superior dual-task performances achieved in Experiment 2A, but not those achieved in Experiment 2B, were due in part to the separation of response modalities. Secondly, as in Experiment 2A, almost complete overlap occurred when subjects responded to two attributes (position and colour) of the same stimulus. However, when neither task was highly S-R compatible or S-C-R compatible, there was no overlap between tasks, implying that these two factors contributed in part to the superior dual-task performances observed in Experiment 2A. Together, the findings from the present study and from the Gladstones etal. (1989) study suggest that if two independent and non-redundant tasks (a) are sufficiently dissimilar, (b) generate little or no mutual confusion, and (c) are such that at least one task is highly S-R (and S-C-R) compatible, then some processing in the two tasks will occur in parallel. If, on the other hand, two independent and non-redundant tasks (a) are identical, (b) are dissimilar but involve highly abstract or arbitrary mappings of stimuli to responses, or (c) generate massive mutual confusion, processing will regress to a purely serial, single-channel, mode. Existing accounts of dual-task performance seem unable to account for this complex pattern of findings. In the final chapter of this thesis a hybrid single-channel and multi-channel model of dual-task performance is proposed to account for these and other relevant findings.

Author dc:creator

• Regan, Michael Arthur

Language dc:language.iso

en_AU

Dc Identifier Other

b18142229

OAI identifier oai:identifier

oai:openresearch-repository.anu.edu.au:1885/11217