In the current study, we examined how smooth pursuit eye movements influenced task-switching saccades. Participants alternated their gaze between a continuous pursuit and a discrete object discrimination task. The main results of our study show asymmetries in saccade reaction times (SRTs) from and to smooth pursuit depending on the smooth pursuit movement direction.
Outward SRT asymmetry
We examined whether ongoing pursuit influenced initiation of outward saccades (saccades from the pursuit to the discrimination task). The SRTs of outward saccades were shorter when the saccade target appeared in the direction of the pursuit target’s movement.
In explaining this result, we first address two basic factors that are known to influence SRTs, namely the eccentricity of the saccade target and the orbital position of the eye at saccade onset. SRTs have been shown to be a function of target eccentricity (Kalesnykas and Hallett 1994, but see also Hodgson 2002; Dafoe et al. 2007). For example, in the study by Kalesnykas and Hallett (1994), longer SRTs were measured for extremely small (<2°) and very large eccentricities (>15°). Our analysis of saccade amplitudes indicates that eccentricities at the time of saccade onset differed in the two motion conditions. Saccades in the direction of motion (to condition) were larger than saccades in the opposite direction. However, considering the pattern of results of Kalesnykas and Hallett (1994), longer rather than shorter SRTs would be expected for larger eccentricities. Another factor that directly affects SRTs is the orbital position of the eye. It has been shown that centripetal saccades, saccades from an eccentric starting positions that move the eyes back to the primary position, exhibit shorter SRTs than saccades in the opposite condition (Fuller 1996; Paré and Munoz 2001). In the current experiment, orbital positions were approximately equal (namely close to the primary position) at saccade onset. This suggests that the observed SRT asymmetries are neither linked to eye nor target position but to the target’s motion direction.
This observation is in line with previous research that explains asymmetric SRTs on the basis of an attentional bias in the direction of pursuit (Tanaka et al. 1998; Kanai et al. 2003; Khan et al. 2010). For example, Khan et al. (2010) showed shorter SRTs to targets ahead of the pursuit stimulus and longer SRTs to targets behind the pursuit target. This phenomenon has also been reported for manual response times in a detection task (van Donkelaar 1999; van Donkelaar and Drew 2002). This suggests that SRT asymmetries are not due to biomechanical compatibility between saccade and pursuit direction but rather an example for an attention shift in the direction of pursuit. Khan et al. (2010) suggest that orienting of covert attention in anticipation of the pursuit target’s motion is important such that potentially required actions can be planned ahead, in compensation for neural processing delays in perception and action.
The results of our study exclude a basic explanation for this attentional bias. In pursuit tasks that do not require a visual analysis of the target of some sort, the pursuit behavior is not completely smooth but shows discontinuities in the form of anticipatory saccades (Van Gelder et al. 1990, see also Koken and Erkelens 1992; Xia and Barnes 1999). According to Van Gelder et al. (1990, 1995a, b; Kathmann et al. 1999), pursuit is typically performed automatically to support the visual analysis of the target. Without such a visual function, attention is unnaturally focused on pursuit itself, which may explain any anticipatory behavior. Our hypothesis was that this tendency to anticipate the pursuit target’s motion could have also caused the attentional bias and reduction in SRTs in the motion direction. However, the current results speak against this assumption. SRTs were asymmetric in both pursuit conditions (ocular, oculomanual), despite clear differences in smooth pursuit behavior in regard to anticipation (see also Mather and Putchat 1983; Gauthier et al. 1988; Vercher and Gauthier 1992; Koken and Erkelens 1992). This suggests that pursuit-related attentional biasing is not merely the result of confined experimental settings and extends its relevance to more realistic conditions.
Inward SRT asymmetry
We tested whether SRT asymmetries also existed for inward saccades (saccades from the discrimination object to the pursuit target). SRTs were shorter when the saccade target moved away from the current fixation location (foveofugal) and longer when it moved toward the fixation location (foveopetal).
Like outward saccades, SRTs were shorter when the saccade moved the eye in the same direction as the pursuit target. An advantage for saccades that are compatible with the pursuit motion direction was explained for saccades from pursuit by a broad attentional bias in the direction of pursuit, which facilitates detection and processing of targets that appear in this direction (Blohm et al. 2005; Khan et al. 2010). An alternative explanation suggests that it is not a sustained bias in attention but facilitation of attention capture, which leads to reduced SRTs to sudden target onsets in the direction of pursuit (Lovejoy et al. 2009).
Neither theory sufficiently explains the current results. First, when fixating on the discrimination object prior to the inward saccade, pursuit targets were situated in the same visual hemifield at similar visual field locations in both motion conditions. A broad tuning of attention would therefore be expected to affect saccades in both conditions. Second, inward saccades were not triggered by a sudden target onset. Instead, saccades followed the discrimination task and moved the eye to the pursuit target, which was continuously present throughout the experiment. Hence, facilitation of attention capture is also unlikely to explain the obtained result.
In the remainder of this discussion, we will consider several alternative explanations, namely the influence of the discrimination task, amplitude differences, motion processing asymmetries, compatibility with early pursuit responses, and inhibition of saccades at the onset of smooth pursuit.
The amount of time spent on performing the discrimination task may explain differences in SRTs. For example, longer discrimination times may be the result of inaccurate foveation after the outward saccade. However, our analysis provides no evidence for this assumption. Discrimination performance and saccade accuracies were similar in both conditions. In addition, the current finding is corroborated by data from a different experiment, in which neither discrimination nor a saccade was required before the saccade to pursuit (Bieg et al. 2013, in preparation).
Factors that influence SRTs more directly are the eccentricity of the saccade target and the orbital position of the eye at saccade onset. However, the eccentricity differences in our experiment would predict the opposite effects on SRTs (see previous section). This suggests that the observed SRTs are primarily influenced by the motion direction of the pursuit target.
Asymmetries in the processing of motion have been observed in several experiments. But the conditions that would lead to an advantage in one or the other direction (foveofugal/foveopetal) are not clear (Naito et al. 2010). For example, in an experiment by Ball and Sekuler (1980), RTs to motion onsets of foveofugal motion were shorter. Other experiments showed an advantage for foveopetal motion (Mateeff and Hohnsbein 1988; Mateeff et al. 1991a, b; Raymond 1994; Jancke et al. 2004). One reason for these conflicting findings could be differences in the presented type of motion. Mateeff et al. (1991b) compared flow-field motion (i.e., random-dot kinematograms) stimuli and single-target motion stimuli. The latter stimulus is similar to the one that was used in the present experiment. Mateeff et al. (1991b) show that stimuli of this sort lead to processing advantages of foveopetal motion (in terms of onset detection) rather than foveofugal motion, as in our experiment (in terms of SRTs). These findings speak against an explanation on the basis of motion processing asymmetries.
Potentially related to asymmetries in motion processing are asymmetries in smooth pursuit behavior. These can be observed during the early (ca. 100 ms), open-loop pursuit response (Tychsen and Lisberger 1986; Carl and Gellman 1987). This response can occur at the onset of pursuit and moves the eyes in the direction of the pursuit target’s motion. Investigations of this response showed larger early accelerations during foveopetal motion (Tychsen and Lisberger 1986). This initial acceleration could potentially affect saccade onsets by modulating the omnipause neuron activity in the brain stem. Inhibition of these neurons is required to trigger a saccade (Scudder et al. 2002), and they also likely regulate smooth pursuit onset and gain (Missal and Keller 2002; Kornylo et al. 2003; Krauzlis 2005). Inhibition of omnipause activity due to early pursuit responses could therefore facilitate saccade triggering. With regard to the findings by Tychsen and Lisberger (1986), stronger inhibition of omnipause neurons would be expected when the target moves foveopetally, which would explain shorter SRTs in this direction. Again, this is incompatible with the results that we observed, namely shorter SRTs to foveofugal motion.
Apart from this hypothetical facilitatory connection, pursuit-related activity is known to inhibit saccades in certain conditions. Increased SRTs or even complete suppression of a saccade can be observed in foveopetal step-ramp tasks. There, the target is stepped in the opposite movement direction such that it moves across its original position after a specified time. This time is the zero-crossing or eye crossing time (Gellman and Carl 1991; de Brouwer et al. 2002a). In the case of zero-crossing times of 200 ms, the initial saccade to the target position is delayed or suppressed completely and smooth pursuit of the target commences directly (Rashbass 1961; Gellman and Carl 1991). It is currently unknown how this cancelation process affects saccades for zero-crossing times larger than 200 ms. For example, the study by Moschner et al. (1999) measured SRTs in step-ramps with 200 ms zero-crossing times. Their results show longer SRTs in foveopetal steps (ca. 400 ms) and shorter SRTs in foveofugal steps (ca. 200 ms). However, this difference in SRTs can be primarily attributed to cancelation of the initial saccade and re-planning of a new saccade in the direction of motion after zero-crossing. In contrast, SRT differences in inward saccades in our experiment cannot be attributed to cancelation and re-planning since (1) zero-crossing never actually occurred and (2) hypothetical zero-crossing times were much longer: An estimate based on the average amplitude prior to saccade onset (12.5°) divided by the pursuit target speed (max. 6.7°/s, average before onset 5.2°/s) results in zero-crossing times between 1.8 and 2.4 s.
It cannot be excluded that the same mechanisms that lead to cancelation of saccades in short zero-crossing times also influence saccade generation in longer zero-crossing times. Saccade triggering as well as cancelation are thought to depend on neuronal accumulation processes (Carpenter and Williams 1995; Hanes and Schall 1996). Importantly, there is also evidence for inhibitory links between those processes (Boucher et al. 2007). Assuming that cancelation of saccades to foveopetal motion is indeed organized by such a process network, foveopetal motion would be expected to have a stronger impact on the cancelation process gain than foveofugal motion. The inhibitory connections between the two processes can then explain increased SRTs to foveopetal motion. In this respect, it is important to point out that an asymmetry in SRTs may also be behaviorally useful. In foveofugally moving targets, computation of the exact time of zero-crossing from a motion analysis of the pursuit target becomes obsolete. Considering that a more precise motion estimate also requires more time (Bruyn and Orban 1988; Bennett et al. 2007), it would be efficient to allocate less time for the analysis of foveofugal rather than foveopetal motion, in particular because foveofugal motion moves the target out of the visual field, which poses the danger of losing track of it entirely when the saccade is triggered too late. In this respect, the ensuing reduction in saccade RTs may additionally be related to time pressure (Reddi and Carpenter 2000; Montagnini and Chelazzi 2005; Bieg et al. 2012).