Covert spatial attention allows us to select important and/or behaviourally relevant visual inputs by enhancing signals arising from attended locations and suppressing signals from unattended locations [1
] without actually moving the eyes to that location. Despite many advances in understanding the cognitive processes involved in spatial attentional selection, an enduring issue is the mechanism by which attention is moved from one location to another. It is generally agreed that the orienting of spatial attention can occur in an automatic ‘exogenous’ mode in response to salient external events (e.g., the flashing lights of an emergency services vehicle) or a controlled ‘endogenous’ mode in response to the observer’s goals (e.g., systematically scanning the road ahead to check for hazards) [2
], and that these systems are partially dissociable [3
]. It is also widely accepted that eye movements (‘overt’ shifts of attention) are preceded by a covert shift of attention to the saccade goal, known as ‘pre-saccadic attention’. However, there is a long-running debate concerning the relationship between the mental process involved in covert orienting of attention (i.e., attending to things that are not being gazed at), and those involved in overt orienting of attention (i.e., orienting the eye to the stimulus of interest) [4
]. One proposal, originally known as the Oculomotor Readiness Hypothesis (OMRH) [5
] and later as Premotor Theory (PMT) [6
], proposed a complete functional overlap between spatial attention and oculomotor control. OMRH/PMT is often used as shorthand to refer to the general idea that covert attention is, in some way, linked to the oculomotor system. However, this usage does not do full justice to the OMRH/PMT theory, which makes clear and testable predictions about the precise relationship between oculomotor control and covert spatial attention. More specifically, OMRH/PMT holds that the programming of a saccade is both necessary and sufficient for covert orienting of attention [7
Despite being the original proponents of OMRH, Klein and colleagues concluded that endogenous attention was in fact independent of saccade programming [5
], although they speculated that OMRH may still hold for exogenous attention. Subsequently, a number of other proposals suggesting differing degrees of overlap between attention and saccade control have been put forward [9
]. Following the work of Klein and colleagues, we have pursued the idea that the relationship between covert attention and saccade programming may indeed be dependent on the mode of orienting, such that OMRH/PMT was only true when the exogenous mode of orienting was engaged [4
]. In this review we outline the main lines of argument for and against OMRH/PMT as a theory of endogenous covert orienting, then explain why we believe that OMRH/PMT is false for endogenous covert orienting, but remains tenable as a theory of exogenous, covert orienting.
2. The Case for OMRH/PMT
The case for OMRH/PMT draws on three main lines of evidence. Firstly, there is clear evidence that saccadic eye movements are preceded by a mandatory ‘pre-saccadic’ shift of attention [12
] and a more efficient distractor suppression at non-saccade goals [19
]. This pre-saccadic attentional facilitation is clearly tied to the programming of an eye movement, as the effect grows larger with proximity with saccade onset [20
] and occurs even when the participant expects the probe to appear opposite the saccade goal, implying that programming an eye movement is sufficient to trigger a shift of covert attention [13
]. Furthermore, shifts of attention appear to affect the trajectory of saccadic eye movements, consistent with the idea that shifts of attention activate a saccade plan [16
Secondly, eye movements and covert shifts of attention appear to activate similar networks of brain areas, including the Frontal Eye Fields (FEF), the Lateral Intraparietal cortex, and the Superior Colliculi (SC) [24
](see Figure 1
), and lesions to these brain areas are associated with deficits of both covert orienting and saccade control [30
]. Moreover, electrical stimulation of FEF neurons in non-human primates elicited fixed-vector saccadic eye movements, and subthreshold stimulation of the same neurons significantly enhanced perceptual discrimination, even though the monkey was still centrally fixating [37
]. Using a similar methodology, Moore and colleagues also demonstrated that stimulation of FEF modulated the sensitivity of neurons in V4, an area of the visual cortex that codes for colour, orientation and spatial frequency, and whose visual receptive fields overlap with the motor field [39
]. The effect of FEF microstimulation on neural responses in V4 was analogous to that observed when the monkey endogenously attended the location [39
]. These data suggest a causal role for saccade programming in covert attention, as predicted by OMRH/PMT.
A third line of argument draws on studies in which eye movements are impaired, experimentally restricted, or experimentally modulated. For example, Craighero, Carta and Fadiga [41
] observed that patients with a palsy of the VIth
cranial nerve were unable to covertly orient attention only when viewing stimuli with their palsied eye, suggesting that the endogenous shift of attention was impaired when viewing with the damaged eye but not when viewing with the intact one. In line with this study, Craighero, Nascimben and Fadiga [42
] used an eye abduction paradigm (see Figure 2
), where saccadic eye movement programming is disrupted by forcing healthy participants to rotate the eye by 40° into the temporal hemifield. In their experiment participants were presented with a classical Posner cueing task in which a central predictive cue (i.e., a bar attached to the fixation square indicating left or right) indicated in 70% of the cases the accurate position of the upcoming target, which could be either in the nasal hemispace (i.e., at a position that can be reached by a saccadic eye movement) or in the temporal hemispace (i.e., outside a position reachable by a saccadic eye movement). Visual acuity remained unaffected but the attentional benefits typically observed with valid cues were reduced when stimuli were presented in the temporal/eye movement restricted hemispace but not when presented in the nasal hemispace. The authors concluded that, consistent with Premotor theory, covert orienting of attention is subject to the limitations of the saccadic system such that attention cannot be deployed at a location that cannot become the goal of a saccadic eye movement. This led to the proposal that covert attention and saccadic eye movements share the same ‘stop limit’, which is the range of eye movements, also referred to as Effective Oculomotor Range (EOMR).
Other studies have used the saccadic adaptation technique to dissociate the perceived position of a saccade target from the actual endpoint of the eye movement. In saccadic adaptation tasks the participant makes a saccade to a peripheral stimulus, but during the saccade the stimulus jumps to a new position (double-step task) [43
]. At the start of the experiment the participant initially moves to the original stimulus position then, unconsciously, makes corrective eye movements towards the second stimulus position. However, over the course of many trials they adapt the amplitude of the saccade to ensure it lands at the final position of the stimulus rather than its original position (for a review, see [44
]). OMRH/PMT predicts that saccadic adaptation should also result in the adaptation of covert shifts of attention, such that the locus of attention should be at the final stimulus position, not the starting position. To test whether attention focus is shifted towards the saccade target or the final eye position, Ditterich et al. [45
] asked participants to make a saccade towards a peripheral location and, before the first saccade onset, they briefly flashed a discrimination target at one of four possible locations. The discrimination performances were compared before and after the saccadic adaptation. Prior to adaptation, discrimination performance was best at the goal of the saccade. After adaptation, optimal discrimination performance was still observed at the goal of the first saccade, and not at the endpoint of the adapted saccade. This result is not consistent with OMRH/PMT, and Ditterich et al. concluded that the attentional focus is always directed to the primary target position and not to the saccade landing position [45
]. However, Collins and colleagues argued that this conclusion was premature, given that the magnitude of the adaptation effects observed by Ditterich. was somewhat small. In two subsequent studies using more effective adaptation protocols they showed that saccadic adaptation does indeed produce adaptations of pre-saccadic attention [46
] and that pre-saccadic displacement of attention would be shifted both to the position of the saccadic target and to the landing position of the adapted saccades [48
]. In a recent study, Habchi and colleagues claimed that saccadic adaptation leads to changes in the allocation of covert attention, although these changes appear to be due to a more general bias towards the side of adaptation, rather than a modulation of covert orientation per se [49
]. Overall, the evidence is consistent with the claim that saccadic adaption is associated with adaptations of pre-saccadic attention, which has been interpreted as evidence for OMRH/PMT.
Further evidence for OMRH/PMT is the finding that covertly attending a location produces a change in the trajectory of saccades, such that they deviate from the intended location [22
]. Trajectories of vertical and oblique saccades are never completely straight but curvilinear, even when aiming at an isolated target [50
], and it has been suggested that saccade curvature is determined by mechanisms situated in the final pathway of eye movement generation [52
]. In addition to this natural tendency, other objects presented in the visual scene can influence the magnitude and direction of saccade curvature. Several authors have found that presenting an irrelevant distractor stimulus near a saccade target affects the saccade curvature [22
]. In some instances, saccades can curve towards the irrelevant stimulus, as in visual search tasks [56
], when the location of the saccade target is highly unpredictable, or for short-latency saccade [57
], but in other cases, there is a tendency to deviate from the position of the distractor, particularly when saccade latencies are long [55
], whether the saccade is reflexive or voluntarily triggered [53
]. These trajectory deviations are typically attributed to competition between saccade plans associated with the target and distractor, and evidence that covert attention can also cause trajectory deviations [22
] is therefore often cited as evidence for OMRH/PMT.
To briefly summarize, OMRH/PMT argues that covert orienting of attention depends on the activation of a saccade plan. Consistent with this hypothesis, there is a mandatory orientation of attention to saccade goals; covert and overt attention activate overlapping brain areas and damage to these areas causes problems with both overt and covert orienting. For example, ophthalmoplegic patients have deficits of covert attention that seem to mirror their ocular deficit. Moreover, modulating the gain of saccades also modulates the gain of pre-saccadic shifts of attention, and covertly attending a peripheral location affects the metrics of overt saccades, such that their trajectories are deviated away from the attended location. Altogether, these studies seem to offer clear evidence for a tight coupling between attention and oculomotor control.
4. An Oculomotor Readiness Hypothesis of Exogenous Orienting (OREO)
On the basis of these studies, we argue that the data are most consistent with an Oculomotor Readiness Hypothesis that is specific to Exogenous Orienting (OREO). On a theoretical level, the relationship between attention and eye movements can be understood in terms of Biased Competition, such that activation of the motor system exerts a powerful biasing influence on competitive interactions in the visual system [112
]. In Biased Competition, the locus of attention arises from a stimulus-driven competition between signals relating to stimulus salience (e.g., their brightness, size, contrast, orientation), which can be biased by goal-driven factors such as the goals of the observer. The competition takes places in a topographic map of space, called a priority map ([113
]. The cortical substrates of the priority map are thought to lie in the posterior parietal cortex a region that has dense reciprocal connections with areas known to be directly involved in saccade control such as Frontal Eye Field (FEF) and Superior Colliculus (SC) (for a review, see [114
]) When a location is activated in the priority map the activation is passed downstream to oculomotor structures, such as the SC, which represent the prioritized location as the goal of a potential movement. These oculomotor signals are then fed back into the priority map, thus further biasing activity in favour of the activated location [115
]. This reciprocal feedback loop will typically produce very rapid selection of a peripherally cued location, which will facilitate target detection, producing the rapidly developing perceptual advantage typically associated with exogenous attention. When the oculomotor system malfunctions, or when targets appear at locations that cannot become the goal of a saccade, the motor system exerts a much weaker influence on the biased competition. If a target is associated with a persistently large salience signal (e.g., in a feature search task in which the search array remains visible until a response is made), the absence of reciprocal reinforcement from the oculomotor system should slow selection of the feature singleton but will not necessarily prevent its selection. This is exactly the pattern we observed, such that placing a salient feature beyond the EOMR delayed, rather than abolished the capture of attention by the singleton [93
]. If salience signal is transient (as in the peripheral cueing task), the absence of reinforcement from the oculomotor system reduces the chance of the cued location ‘winning’ the competition before the signal decays, and therefore reduces the probability of observing an exogenous shift of attention to the cued location. We can therefore understand the relationship between exogenous attention and saccade programming in terms of oculomotor inputs that bias competition on the priority map in favour of the saccade endpoint. The demotion of the oculomotor system from being the sole arbiter of the locus of attention to being one of many potential influences on the process of biased competition is a key difference between OREO and OMRH/PMT. Importantly it does not deny the possibility that exogenous orienting can be driven by other inputs, such as stimulus salience [116
]. Rather, OREO holds that optimally efficient exogenous orienting relies on activation of a saccade plan, and when this activation is disrupted exogenous orienting becomes slower and less reliable.
OREO makes some clear and testable predictions about the interaction between covert, exogenous orienting and saccade programming. Firstly, exogenous orienting should always be associated with the activation of a saccade plan. Secondly, inability to plan a saccade should disrupt exogenous orienting. Thirdly, factors that affect the properties of saccadic eye movements (e.g., their latency, amplitude and direction) should also affect the speed and accuracy of covert exogenous orienting. MacLean et al. [102
] tested the first prediction using a variant of the dual task procedure developed by Klein and Pontefract [8
]. Contrary to the predictions of OREO, they observed no reduction in saccade latency at peripherally cued locations and concluded that exogenous orienting was not associated with saccade programming. However, this conclusion is premature, as MacLean et al. used a SOA of 250 ms, allowing ample time for the suppression of saccade programming following a shift of attention. Indeed, the authors concede that their results are more similar to those of Belopolsky and Theeuwes [9
], who previously argued that maintenance of attention was independent of saccade programming. The MacLean study also utilises a very high proportion of ‘no-go’ trials, where a cue appears but no saccade is permitted, and as we have already noted, a high proportion no-go trials can mask saccadic priming effects caused by peripheral cues [9
]. We examined the third prediction by using instrumental conditioning of eye movements [117
]. If exogenous orienting depends on activation of the oculomotor system, then one might predict that a manipulation that modulates saccade latencies should also affect covert exogenous attention. In our first experiment we found that rewarding eye movements to a specific spatial location reliably reduced saccade latencies to that location, and that this conditioning persisted for 180 trials once rewards were removed. However, in a second experiment this modulation of the oculomotor system had no effect on the magnitude of covert, exogenous orienting or Inhibition of Return. McCoy and Theeuwes [118
] report a similar result in a study in which participants learned to made saccades to a location associated with a large reward. As with our study, the high-value location was associated with shorter saccade latencies. However, this oculomotor facilitation did not translate into enhanced performance at the rewarded location in a subsequent task that measured perceptual discrimination at the rewarded location while the eyes remained at fixation. These findings may seem hard to reconcile with the third prediction of OREO, but it is important to note that OREO predicts that reducing the latency of a saccade should lead to a reduction in the rise-time of attention (i.e., the speed at which attention is oriented to the cued location) rather than the absolute magnitude of the cueing effect. Thus, in our view, none of these studies offers a strong test of the predictions of OREO. In contrast, McFadden, Khan and Wallman [119
] reported that it was possible to elicit adaptation of exogenous, covert orienting, which was accompanied by an adaptation of subsequent eye movements, suggesting that the adaptation of exogenous attention relied on changes in the oculomotor plans elicited by the peripheral onset. It is not known whether endogenous, covert attention can be adapted in the same way, but such a study would provide a good test of OMRH/PMT and OREO, and the former theories predict an effect of adaptation of endogenous attention on saccade amplitude, whereas OREO does not.