The Influence of Task Instruction on Action Coding: Constraint Setting or Direct Coding? Dorit Wenke and Peter A. Frensch Humboldt University at Berlin In 3 experiments, the authors manipulated response instructions for 2 concurrently performed tasks. Specifically, the authors’ instructions described left and right keypresses on a manual task either as left versus right or as blue versus green keypresses and required either “left” versus “right” or “blue” versus “green” concurrent verbalizations. When instructions for responses on the 2 tasks were in terms of location (Experiment 1) or color (Experiments 2a and 2b), then compatible responses on the tasks were faster than incompatible responses. However, when the verbal task required “left” versus “right” responses but instructions for manual keypresses referred to blue versus green (Experiments 3a and 3b), then no response compatibility effects were observed. These results suggest that response labels used in the instruction directly determine the codes that are used to control responding. Human beings are able to seemingly effortlessly and rapidly translate arbitrary task instructions into behavior. Thus, most hu- mans are perfectly capable of lifting a finger at the onset of a tone, for example, when they are instructed to do so. Although most researchers would probably agree that instructions somehow de- termine how such a task is performed, it is perhaps surprising that relatively little is known about how exactly task instructions come to control behavior. In the present line of research, we are concerned with simple binary stimulus–response (S-R) instructions involving spatially organized keypress responses, such as, for example, “When you see a square on the screen, then press the blue key; when you see a circle on the screen, then press the red key.” Our main goal is to understand which role the specific response labels given in verbal instructions (e.g., “blue” and “red”) play in the control of the instructed behavior. There exist at least two prominent theoretical positions regard- ing this question. On the one hand, it is conceivable that response labels in instructions directly influence response coding. For ex- ample, a simple S-R instruction (e.g., “When you see a square on the screen, then press the blue key”) might set up a link between the stimulus and the response components of the instruction by activating and linking their corresponding mental representations (i.e., “square” and “blue” in the example). The specific motor codes that are needed to perform the instructed response might then primarily be accessible via the mental representation of the response label (i.e., “blue”). Such a view is consistent with the intentional-feature-weighting hypothesis, for instance, as recently proposed by Hommel and colleagues (Hommel, Mu ¨sseler, As- chersleben, & Prinz, 2001). According to the intentional-feature- weighting hypothesis, the mental representation that controls responding contains both instructed response labels and nonin- structed features; for the control of action, instruction response labels are weighed more heavily than are noninstruction features. Alternatively, task instructions might only set up general con- straints on how actions can be coded. According to this view, the response labels used in the instructions (e.g., “blue” and “red”) are not incorporated into the mental representations that control re- sponding. Rather, responses are controlled by the codes that best discriminate between the response alternatives possible in the context of a given task instruction. For spatially organized key- press responses, the spatial-coding hypothesis that is widely ac- cepted in the (spatial) stimulus–response compatibility (SRC) lit- erature (e.g., De Jong, Liang, & Lauber, 1994; Lu, 1997; Roswarski & Proctor, 2003) can be considered to represent this general position. According to the spatial-coding hypothesis, man- ual keypress responses are coded in terms of relative left–right key location whenever the spatial dimension allows discriminating between responses and regardless of the specific response labels provided in the instruction. In the present research, our goal was to pit the intentional- feature-weighting and the spatial-coding hypotheses against each other empirically, that is, to assess which of the two alternative hypotheses better captures how responding is controlled in situa- tions where instructions for binary S-R behavior require keypress responding. For the most part, the existing empirical evidence that speaks to this issue involves so-called (spatial) compatibility ef- fects. Compatibility effects refer to the finding that responding is easier (faster and less error prone) in compatible conditions than in incompatible conditions. For instance, keypress responses are faster when stimulus position and response location are compatible (i.e., when they correspond), even if stimulus position is irrelevant for responding—a phenomenon known as the Simon effect (for reviews, see Lu & Proctor, 1995; Simon, 1990). If participants code their responses as instructed—as is proposed by the direct coding view, in general, and the feature-weighting Dorit Wenke and Peter A. Frensch, Department of Psychology, Hum- boldt University at Berlin, Berlin, Germany. We thank Martin Eimer, Sander Los, and Robert Proctor for helpful comments on earlier versions of this article and Ina-Kristin Saure for collecting and coding the data. Correspondence concerning this article should be addressed to Dorit Wenke, Department of Psychology, Humboldt University at Berlin, Ru- dower Chaussee 18, D-12489, Berlin, Germany. E-mail: dorit.wenke@ psychologie.hu-berlin.de Journal of Experimental Psychology: Copyright 2005 by the American Psychological Association Human Perception and Performance 2005, Vol. 31, No. 4, 803– 819 0096-1523/05/$12.00 DOI: 10.1037/0096-1523.31.4.803 803