Mechanisms of inhibition of return: Brain, behavior, and computational modeling

Abstract

Inhibition of return (IOR) is a cognitive phenomenon whereby reaction times (RTs) are slower to cued relative to uncued targets at cue-target onset asynchronies (CTOAs) greater than approximately 300 ms. One important theory of IOR proposes that there are two mutually exclusive forms of IOR, with an attentional/perceptual form arising when the oculomotor system is actively suppressed, and a motoric form arising when it is engaged (Taylor & Klein, 2000). Other theories propose that IOR is the result of multiple, additive neural mechanisms (Abrams & Dobkin, 1994). Here, we have performed computational simulations and empirical investigations in an attempt to reconcile these two competing theories. Using a dynamic neural field (DNF) model of the intermediate layers of the superior colliculus (iSC), we have modeled both a sensory adaptation mechanism of IOR, and a motoric mechanism resulting from the aftereffects of saccadic eye movements. Simulating these mechanisms, we replicated behavior and neurophysiology in a number of variations on the traditional cue-target paradigm (Posner, 1980). Predictions driven by these simulations have led to the proposal of many behavioral and neuroimaging experiments which further examine the plausibility of a 2-mechanisms theory of IOR. Contrary to our original predictions, we demonstrated that saccades are biased away from cued targets in a paired target saccade averaging paradigm, even at short CTOAs. In paradigms thought to recruit both sensory and motoric mechanisms, we robustly demonstrated that there are at least two independent, additive mechanisms of IOR when tasks require saccadic responses to targets. When similar paradigms were tested with manual responses to targets, additivity effects did not hold, implying that the motoric mechanism of IOR does not transfer from the oculomotor to skeletomotor systems. Furthermore, across numerous experiments using event-related potential (ERP) techniques, we have demonstrated that P1 component reductions are neither necessary, nor sufficient, for the behavioral exhibition of IOR. We propose that a comprehensive framework for behavioral IOR must include (at least) four independent neural mechanisms, differentially active depending on circumstances, including sensory adaptation, saccadic aftereffects, local inhibition, and cortical habituation.