One of the central processing stages linking perception and action is response selection (e.g. Lien & Proctor, 2002). Response selection can be described as a mechanism translating stimulus into response representations in a way, which allows adaptive behavior with respect to the given situation and the individual goals. In the two studies of this dissertation, a dynamic, control-related (Study I) and a structural, representation-related (Study II) aspect of response selection were studied thoroughly. Therefore, the effects of response category repetitions under task shift conditions were analyzed.In the first study, it should be answered, whether response execution, selection or merely activation in the previous task are necessary for observing the respective repetition effects, and to what extent these effects are modulated by the actual risk of possible accidental response reexecutions. The answers to these questions served to test, whether learning or reconfiguration mechanisms suffice to explain response-only repetition effects, or whether an additional inhibitory control mechanism needs to be assumed. The results of this first study show, that the respective repetition effects a) depend only on response activation processes and b) are modulated by adaptive adjustment mechanisms. These results suggest the assumption of an inhibitory control mechanism, which is triggered by potential response conflicts.The second study focused on the response categories, which are relevant for response selection, and the factors determining the choice of a specific category type. It should be answered, whether the participants represent the responses according to the relevant stimulus dimensions, or according to some abstract, response-immanent features. The results of the second study show that the responses are generally represented according to abstract response features (e.g. index / middle finger). Furthermore, the instructions proved to be crucial as to which of several equally possible category types are indeed used for response representation, but only if the alternative category type (e.g. left / right) is not induced by overlapping implicit stimulus features.Altogether the results of the present studies support a) the notion that response selection consists of several subprocesses (automatic response activation, intentional S-R-translation and control mechanisms) and b) models postulating an additional layer of abstract response categories besides the stimulus category layer (e.g. Zhang, Zhang & Kornblum, 1999).