Academic journal article
By Mayr, Ulrich
Psychonomic Bulletin & Review , Vol. 13, No. 5
Schneider and Logan (2006) recently showed that cue-switch and task-switch costs are sensitive to the relative probability of cue switches versus task switches. From this they concluded that taskswitch costs reflect priming of cue-cue transitions rather than actual task-switching operations. However, because this design confounded probability of specific cue transitions with probability of task switches, the results could also reflect task-switch-level adjustments. The present experiment (N = 80) pits the critical prediction of the cue-priming account, namely that costs for high-probability cue-cue transitions are smaller than for low-probability cue-cue transitions, against the main prediction of the switch-probability account, namely that switch probability, irrespective of specific cue-cue transitions, determines switch costs. Whereas the cue-priming prediction was rejected, a specific version of the probability account-that subjects are sensitive to the probability of a task switch, given a cue switch-was fully confirmed. Thus, tasks are in fact the critical representational units that determine task-switch cost.
Switch costs as assessed in the task-switching paradigm are typically interpreted as a reflection of the processing demands involved in changing task-specific cognitive configurations (for a recent review, see Monsell, 2003). Recently, however, for one variant of the task-switching paradigm, the so-called cuing paradigm, this interpretation of switch costs has been called into question. In the cuing paradigm, tasks vary randomly on a trial-by-trial basis and each task is indicated by a cue, usually visual. The cuing paradigm allows tight control over when subjects can start preparing for an upcoming task and therefore has proven useful for examining how proactive control influences switch costs (see, e.g., Meiran, 1996). However, in two parallel papers, Logan and Bundesen (2003) and Mayr and Kliegl (2003) noted that, in this paradigm, task changes are completely confounded with cue changes. They therefore introduced a new variant using a 4:2 mapping between cues and tasks. This allows three types of transitions to be examined: no switch, cue switch without task switch, and task switch. The contrast between the no-switch and the cue-switch conditions yields the costs associated with changing the cue (henceforth referred to as cue-switch costs). In both Logan and Bundesen (L&B) and Mayr and Kliegl (M&K) these costs were surprisingly large, suggesting that a substantial component of the traditional switch costs had nothing to do with changing cognitive configurations per se. Where the results of M&K and L&B differed was with respect to the second contrast, that between the cue-switch and the task-switch transitions (henceforth called true task-switch cost). Whereas this component was substantial in M&K, it was nonexistent in L&B; in some of the succeeding papers by Logan and coworkers, the true task-switch costs were present, but only very small (see, e.g., Logan & Bundesen, 2004). On the basis of the finding of substantial task-switch costs, M&K concluded that there is true task switching going in the cuing paradigm, whereas L&B suggested that subjects form compounds between cues and stimuli that allow disambiguating response selection. According to this view, switching occurs between these compounds and not between the actual tasks. Whether or not there is true task switching in the cued task-switching paradigm is a question of considerable theoretical relevance: At stake is to what degree task sets are a mandatory, organizing principle in human action control.
One noteworthy difference between the procedure used by L&B and the one used by M&K was in switch probabilities. In L&B, all cues and stimuli were selected randomly without constraints. With four cues mapped to two tasks, there were four different no-switch cue-cue transitions and four different cue-switch cue-cue transitions, but eight different possible task-switch cue-cue transitions (see Table 1 for a listing of all specific cue transitions possible with a 4:2 cue-task mapping). …