Control by action representation and input selection (CARIS): a theoretical framework for task switching.

Control by action representation and input selection (CARIS) is a modeling framework for task-switching experiments, which considers action-related effects as critical constraints. It assumes that control operates by choosing control parameter values, representing input selection and action representation. Competing CARIS models differ in whether (a) control parameters are determined by current instructions or represent a perseveration, (b) current instructions apply to the input selection and/or to action representation. According to the chosen model (a) task execution results in a default bias in favor of the executed task thus creating perseverative tendencies; (b) control counteracts these tendencies by applying a transient momentary bias whose locus (input selection or action representation) changes as a function of task preparation time; (c) this happens because the task-cue (e.g., SHAPE) initially attracts attention to the immediately available cue-information (e.g., target shape) and then attracts it to inferred or retrieved information (e.g., "circle" is related to the right key press).