Cognitive control of saccadic eye movements

Abstract

The saccadic eye movement system provides researchers with a powerful tool with which to explore the cognitive control of behaviour. It is a behavioural system whose limited output can be measured with exceptional precision, and whose input can be controlled and manipulated in subtle ways. A range of cognitive processes (notably those involved in working memory and attention) have been shown to influence saccade parameters. Researchers interested in the relationship between cognitive function and psychiatric disorders have made extensive use of saccadic eye movement tasks to draw inferences as to the cognitive deficits associated with particular psychopathologies. The purpose of this review is to provide researchers with an overview of the research literature documenting cognitive involvement in saccadic tasks in healthy controls. An appreciation of this literature provides a solid background against which to interpret the deficits on saccadic tasks demonstrated in patient populations.

Introduction

In everyday life we typically make around three saccadic eye movements every second (e.g. Rayner, 1998). We are rarely conscious of these movements (in the sense that we are conscious of other movements we make such as reaching to grasp an object) and, subjectively at least, they appear to involve little or no cognitive effort on our part. Given these credentials saccadic eye movements would not appear to be a particularly promising tool with which to study cognition. In fact, over the last 30 years, researchers have demonstrated that saccadic eye movements are influenced by a wide range of cognitive processes, including those involved in attention, working memory, learning, long term memory and decision making. Moreover, as Carpenter (1994) argues, the oculomotor system provides researchers with “a microcosm of the brain”—one whose sensory input can be precisely controlled and manipulated, and whose limited motor output can be measured with exceptional accuracy with current eye tracking equipment.

It is now widely accepted that the patterns of cognitive dysfunction associated with psychiatric disorders are not simply unfortunate sequelae, but represent information processing deficits that lie at the very heart of the disorders. Furthermore, future pharmacological and psychological intervention strategies will be informed by a more sophisticated understanding of the aetiological role of neurocognitive dysfunction in psychiatric disorders. The combination of a close relationship with cognitive processes and considerable practical advantages in terms of ease of measurement and manipulation mean that the saccadic system will continue to provide researchers with a powerful tool with which to understand neurocognitive processes central to a variety of psychopathological disorders (see e.g. Hutton & Ettinger, 2006).

Studies using psychiatric and neurologically impaired populations and functional neuroimaging techniques have made considerable contributions to our understanding of the role of cognitive processes in saccadic eye movements, and this research is reviewed elsewhere in this special issue (McDowell et al. and Mueri & Nyffeler). This review will focus on studies using healthy human participants, and the cognitive processes involved in performing saccadic tasks that are typically used with psychiatric populations (see Fig. 1).

Electrophysiological studies in non-human primates have demonstrated that it takes around 40 ms for a signal to be transmitted from the retina to the superior colliculus (see glossary), and about 20 ms for stimulation of the same region of the superior colliculus to trigger a saccadic eye movement to a specific location in space (Carpenter, 1981). In reality, however, the typical latency of a “reflexive” saccade (one made towards a sudden onset target) in humans is around 200 ms, and the variability around this average is large. Why do saccades take more than three times longer than they could?

As Carpenter has so elegantly argued (Carpenter, 1981, Carpenter, 2001), saccadic eye movements have such long latencies because we need to work out not just where to look, but whether it is worth our looking there at all (given all the possible places we could be looking). A 15° saccade typically takes around 50 ms, during which much visual processing is suppressed (Ross, Burr, & Morrone, 1996). The more saccades made, the less time there is for fixations (see glossary)—during which the image remains stable enough on the retina for us to “see”. We need, therefore, some means of determining whether the cost any given saccade is worth it given our current goals and limited processing resources (although certain saccadic eye movements such as square wave jerks (see glossary) and micro-saccades are probably not subject to such a cost-benefit analysis). In other words, in most instances saccadic latencies can be viewed as decision times, and, as with any decision process, the outcome can depend on a large number of influences, many of which could be termed “cognitive”.

The LATER (Linear Approach to Threshold with Ergodic Rate) model of saccade generation (Carpenter, 1981, Carpenter and Williams, 1995) provides a useful starting point with which to consider the influences of cognitive processes on saccade generation. According to this model, a decision signal rises linearly from a baseline level (S₀) at a rate (r) until it reaches a threshold value (S_T) at which point the saccade is triggered. The baseline level (S₀) is considered to reflect expectation—the prior probability that a target is present and should be looked at. The rate of rise (r) reflects the supply of information, and the threshold at which the saccade is triggered (S_T) can be considered to reflect “urgency”. Cognitive processes could potentially influence all three of these parameters and therefore mediate how quickly, or even whether, a saccade to a specific location is triggered.

The superior colliculus, in conjunction with other subcortical structures, is perfectly capable of determining the position of a target and generating an accurate saccade towards its location (Moschovakis, 1996, Schall, 1995). It does not, however, operate in isolation. It receives connections (typically inhibitory) from a wide range of cortical areas (including the parietal cortex and frontal regions such as the frontal eye fields (FEF) and supplementary eye fields (SEF) which themselves receive projections from V1 and other areas of visual cortex (see Johnston & Everling in this volume). This wider cortical network, and its interactions with other cortical areas such as the dorsolateral prefrontal cortex (DLPFC; see glossary), can therefore influence the “decision” as to whether a saccade should be made. In other words saccade generation involves a trade off between “bottom up” signals that concern basic stimulus properties such as position, size and luminance, and “top down” signals that reflect the current goals and intentions of the observer. In this sense the extent to which a saccade can ever be said to be truly “reflexive” is debateable, and for the remainder of this review, the term prosaccade (see glossary) will be used to refer to saccades made towards targets.

The review itself is divided into three main sections. The first considers cognitive factors that have been found to influence visually guided saccades made towards targets—most notably attentional processes. The second considers experimental tasks in which the saccades are triggered on the basis of more central cues (although, as argued above, a binary distinction between visually and centrally guided saccades is an over simplification). The final section considers antisaccades (see glossary)—a powerful and popular task in which the peripheral cue and task instructions compete for the control of behaviour.