Action-based selection across feature dimensions: Multidimensional vector models of stimulus-response compatibility

From a functional point of view, cognition is a mechanism by which human uses information in their environments in order to actualize adaptive behavior to sustain their existence. The capacity of the system is limited, so only a subset of environmental information can be processed at a time. Thus, selection of information must occur. This selection mechanism is named selective attention. The first chapter of the thesis reviews studies on selective attention across feature dimensions. Two types of the selection mechanisms are commonly distinguished; controlled processes that are intentionally initiated and monitored, and automatic processes that operate without the actor\u27s intention. Although automatic processes are thought to be independent of the actor\u27s intention, the review revealed that automatic processes are conditioned by the actor\u27s intention. More specifically, selection across feature dimensions is implemented in accordance to what actions are prepared. This action-based selection plays a central role in psychological phenomena such as Stroop interference and stimulus-response (S-R) compatibility. On the other hand, it is believed that controlled processes become automatized only if the operations are extensively practiced for a long period. However, recent studies provide evidence that certain controlled processes can be automatized without extensive training. The second chapter reports a series of experiments that investigated this issue. It is demonstrated that arbitrary S-R mappings can be implemented automatically without one\u27s intention. Consequently, stimulus features that are irrelevant to performing the current task activate associated responses and facilitate or interfere with responding to the task-relevant features, resulting in the cross-dimensional response congruity effect. The experiments indicated that active maintenance of S-R mappings is not necessary for automatic implementation of task-defined S-R mappings. Instead, retrieval of episodic memory contributes to the automatic implementation of task-defined S-R mappings. Finally, in the third chapter, a modeling framework for selection across feature dimensions is developed, named the multidimensional vector (MDV) model. The framework realizes the idea of action-based selective attention by specifying how response properties determine weights of feature dimensions in processing stimuli. Five experiments tested assumptions underlying the framework. The MDV models provided excellent accounts of the experimental data, validating the MDV framework