On the Limits of Advance Preparation for a Task Switch: Do People Prepare All the Task Some of the Time or Some of the Task All the Time? Mei-Ching Lien Oregon State University Eric Ruthruff, Roger W. Remington, and James C. Johnston National Aeronautics and Space Administration Ames Research Center This study investigated the nature of advance preparation for a task switch, testing 2 key assumptions of R. De Jong’s (2000) failure-to-engage theory: (a) Task-switch preparation is all-or-none, and (b) preparation failures stem from nonutilization of available control capabilities. In 3 experiments, switch costs varied dramatically across individual stimulus–response (S-R) pairs of the tasks—virtually absent for 1 pair but large for others. These findings indicate that, across trials, task preparation was not all-or-none but, rather, consistently partial (full preparation for some S-R pairs but not others). In other words, people do not prepare all of the task some of the time, they prepare some of the task all of the time. Experiments 2 and 3 produced substantial switch costs even though time deadlines provided strong incentives for optimal advance preparation. Thus, there was no evidence that people have a latent capability to fully prepare for a task switch. People are capable of performing an enormous range of tasks. To perform tasks relevant to current goals, it is necessary to impose top-down executive control over cognitive processing rather than simply reacting to the most salient stimulus. Although such executive control is clearly possible, it is subject to some limitations. One important limitation is that switching from one task to another incurs a cost in time and/or accuracy. Following the discovery of this switch cost by Jersild (1927), considerable re- search has been carried out to determine its implications for the nature of executive control (see Monsell & Driver, 2000). One classical method used to study task switching is to compare performance in pure-task blocks (no switching required) with performance in alternating-task blocks (switching required). Re- sponse times (RTs) are typically longer in alternating-task blocks than in pure-task blocks, producing a switch cost (e.g., Allport, Styles, & Hsieh, 1994; Jersild, 1927; Spector & Biederman, 1976). A drawback of this method, noted by Rogers and Monsell (1995; see also Los, 1996), is that participants have to keep two stimulus– response (S-R) mappings in mind throughout the alternating-task blocks but not the pure-task blocks. Consequently, the cost of switching tasks is confounded with the cost of keeping two task sets active. To avoid this confound, Rogers and Monsell (1995) developed an alternating-runs paradigm (e.g., AABBAABB . . .) in which task-switch trials and task-repetition trials are intermixed within the same block. Another common method used to study task switching is to present a task cue prior to each trial, in an otherwise unpredictable task sequence, and then examine how the switch cost is affected by the preparation time (e.g., Meiran, 1996). Despite the variety of paradigms used to investigate task switch- ing, studies have generally yielded similar results. One of the most robust findings is that switch costs are larger when each stimulus is appropriate for performance of both tasks (dual-affordance or bivalent stimuli) than when each stimulus is appropriate for only one task (single-affordance or univalent stimuli). In a dual- affordance condition, each stimulus has the potential to automat- ically activate responses from both the currently relevant and the currently irrelevant task set. Thus, executive control is needed to prevent application of the irrelevant task set or, failing that, to resolve the conflict between multiple activated responses. Dual-affordance switch costs are robust, remaining substantial even when (a) the task identity is provided in advance of each trial (e.g., Lien, Schweickert, & Proctor, 2003; Sohn & Carlson, 2000) and (b) the time interval between each response and the subsequent stimulus onset (response–stimulus interval; RSI) is relatively long (e.g., Meiran, 1996, 2000; Rogers & Monsell, 1995). The existence of a residual switch cost when there is ample opportunity to prepare in advance for a new task indicates that executive task- control ability is limited in some way. Theoretical Interpretations of Residual Switch Costs Several alternative, but not mutually exclusive, hypotheses have been proposed to explain why a residual switch cost occurs even with ample opportunity for preparation. Allport and his colleagues (Allport et al., 1994; Allport & Wylie, 2000) have attributed costs on task-switch trials to task-set activation carried over from the previous trial (task-set inertia) that cannot easily be suppressed by top-down preparation for a new task. Mayr and Keele (2000) Mei-Ching Lien, Department of Psychology, Oregon State University; Eric Ruthruff, Roger W. Remington, and James C. Johnston, National Aeronautics and Space Administration (NASA) Ames Research Center, Moffett Field, California. This research was supported by funding from the National Research Council, by NASA Grant NCC 2-1325, and by the Human Measures and Performance Project of NASA’s Airspace Systems Program. We thank Richard Carlson and Stefanie Schuch for their comments on a draft of this article. Correspondence concerning this article should be addressed to Mei- Ching Lien, Department of Psychology, Oregon State University, Corval- lis, OR 97331. E-mail: mei.lien@oregonstate.edu Journal of Experimental Psychology: Copyright 2005 by the American Psychological Association Human Perception and Performance 2005, Vol. 31, No. 2, 299 –315 0096-1523/05/$12.00 DOI: 10.1037/0096-1523.31.2.299 299