Smooth pursuit eye movements are used to continuously track slowly moving visual objects. A peculiar property of the smooth pursuit system is the nonlinear increase in sensitivity to changes in target motion with increasing pursuit velocities. We investigated the role of the frontal eye fields (FEFs) in this dynamic gain control mechanism by application of transcranial magnetic stimulation. Subjects were required to pursue a slowly moving visual target whose motion consisted of 2 components: a constant velocity component at 4 different velocities (0, 8, 16, and 24 deg/s) and a superimposed high-frequency sinusoidal oscillation (4 Hz, ±8 deg/s). Magnetic stimulation of the FEFs reduced not only the overall gain of the system, but also the efficacy of the dynamic gain control. We thus provide the first direct evidence that the FEF population is significantly involved in the nonlinear computation necessary for continuously adjusting the feedforward gain of the pursuit system. We discuss this with relation to current models of smooth pursuit.
Gain modulation is a major computational principle of sensorimotor transformations. Meaningful changes in the sensory environment or variations of the internal state during sensorimotor processing demand an adaptive reweighting of sensory inputs. A prominent example of gain modulation occurs in visually guided reaching, during which neurons in the posterior parietal cortex show an approximate Gaussian tuning with respect to retinal image location that is modulated by gaze position (Andersen and Mountcastle 1983; Andersen et al. 1985; Salinas and Thier 2000). This mechanism allows for a gaze-independent representation of the target in world-centered coordinates by downstream neurons that form the motor command for reaching movements (Salinas and Thier 2000). A variety of other tasks also require gain modulation mechanisms, for example, postural control during upright stance (Jeka et al. 2000; Peterka and Loughlin 2004; Ravaioli et al. 2005), target selection (Maunsell 1995; Salinas 2005), and the control of the focus of attention (Colby and Goldberg 1999; McAdams and Maunsell 1999; Treue and Martinez-Trujillo 1999). The frontal eye fields (FEFs) are one prominent cortical area implicated in gain modulation mechanisms for saccadic target selection and visual spatial attention (for review, see Awh et al. 2006; Knudsen 2007). Moore and Armstrong (2003) demonstrated that visual responses in area V4 are enhanced by electrical microstimulation of retinotopically corresponding sites in the FEF.
Similarly, smooth pursuit eye movements (SPEM, for review see Lisberger et al. 1987; Krauzlis 2004) exhibit dynamic gain control, resulting in higher sensitivity to retinal image motion with increasing pursuit velocities (Schwartz and Lisberger 1994; Keating and Pierre 1996; Krauzlis and Miles 1996). During fixation, the eyes are largely insensitive to brief perturbations of a stationary target. If, however, the target's trajectory is perturbed during an ongoing SPEM, a clear motor response to that perturbation can be observed (Schwartz and Lisberger 1994). The extent to which perceived target velocity or ongoing eye/gaze velocity contribute to gain control is not yet fully determined. We will therefore refer to the controlling factor as the tracking velocity (see Discussion). It is assumed that this nonlinear behavior results from a cortical amplification of the primary motion detection signal (Keating and Pierre 1996; Tanaka and Lisberger 2001).
Several cortical areas are involved in the generation of the motor commands for SPEM, which serve very different aspects, such as target selection, pursuit initiation and maintenance, or prediction of the target trajectory. In particular, the medial superior temporal area (MST) and the pursuit area of the FEFs contribute crucially to the control of SPEM. They both project to the brainstem premotor structures via different parallel pathways (see e.g., Krauzlis 2004 for review). However, the exact site of the dynamic gain control mechanism has not yet been ascertained. Electrical stimulation studies (Tanaka and Lisberger 2001) and modeling approaches (Nuding et al. 2006, 2008) suggest the FEFs as a promising candidate for dynamic gain control.
Recent studies have indicated that TMS over the FEFs influences the visuo-motor transmission for smooth pursuit, particularly during pursuit initiation (Drew and van Donkelaar 2007) and pursuit of predictable target motion (Gagnon et al. 2006). The aim of the present study was to determine the role of the FEFs in the dynamic regulation of the gain during ongoing smooth pursuit.
The smooth pursuit system is in essence a closed-loop negative feedback system that zeros the retinal error velocity between target and eye. For this class of systems, an amplification of the retinal error signal (i.e., increase in gain in the feedforward sensorimotor pathway from the retina to the eye muscles) leaves the eye velocity during pursuit of constant velocity targets unchanged (due to the negative feedback control), but boosts the responses to high-frequency perturbations of the target trajectory (Keating and Pierre 1996; Churchland and Lisberger 2002). We therefore probed the pursuit system at different constant tracking velocities (carrier, with 0, 8, 16, 24 deg/s) with superimposed high-frequency sinusoidal oscillations (hum, with 4 Hz, ±8 deg/s velocity amplitude) to determine the dependency of the feedforward gain on the ongoing tracking velocity. A similar stimulus has been successfully applied in the quantification of the gain control in monkeys (Keating and Pierre 1996). There, the modulation of the response to the hum during different carrier velocities was evaluated and a linear relationship between response modulation and carrier velocity was observed (this relationship was also observed in responses to impulse-like perturbations in humans, see Churchland and Lisberger 2002). Consequently, the efficacy of the dynamic gain control can be defined as the slope of the measured response modulation at different tracking velocities.
Two possible effects of the stimulation of the FEFs can be distinguished. First, stimulation may yield a change in the feedforward gain of the pursuit system, similar to the electrical stimulation studies by Tanaka and Lisberger (2001). This global change in gain would affect the pursuit system independently of ongoing tracking velocity and would result in a vertical shift of the measured response modulation. In this case, the FEFs could be interpreted to operate as a constant gain element for the conditions considered here. However, other gain modulation mechanisms may still be accomplished by the FEFs (e.g., gain increase due to attentional mechanisms or other cognitive processes). Second, if the FEFs were directly involved in forming the nonlinear signal to dynamically adjust the feedforward gain, interference with them would yield an interaction with tracking velocity. In other words, the slope of the measured response modulation versus tracking velocity would change.
Nine subjects, (4 males, 5 females, mean age 28 years, all right handed) took part in the study. Seven subjects had prior experience of TMS and all gave their informed consent before participating in the study, which conformed to the standards set by the Declaration of Helsinki. Exclusion criteria conformed to current guidelines for repetitive TMS research were applied (Wassermann 1996), including exclusion of any subjects with a family history of epilepsy, or other neurological disorders, the presence of metal in the head or if they were taking medication known to affect seizure thresholds. One subject had to be excluded from the statistical analysis (see below). This study was covered by ethical approval from the UCL/UCLH Ethics Committee (TMS ID 1144 001).
The task was presented using an SMI Eyelink I system, modified for use with TMS, which recorded eye position information at 250 Hz throughout experimental trials. Each trial began with a correction of any eye drift and then followed the time course as described below. A Magstim 200 Super Rapid Stimulator was used to deliver TMS at 60% of machine output over FEFs and the vertex Cz. TMS over Cz produces no significant changes in any of the pursuit characteristics for either predictable or unpredictable target motion, and is therefore typically used as the control condition in eye movement research (see e.g., Drew and van Donkelaar 2007). The fixed stimulation level used was based on our previous studies of FEF with TMS (Muggleton et al. 2003; O'Shea et al. 2004; Campana et al. 2007; Juan et al. 2008). We used a fixed percentage output because motor cortex excitability does not provide a good guide to TMS thresholds in other cortical areas (Stewart et al. 2001) and minor corrections for cortical distance have not proven useful in studies of areas that do not have a measurable threshold of activation. Effective stimulation of FEF has also been found at lower stimulation levels (Grosbras and Paus 2003) than those used here. Stimulation was delivered via a 50-mm figure-of-8 coil held clamped in position with the coil windings anterior to the handle and the handle at approximately 45° from the horizontal and slightly medially (see schematics in Fig. 1).
Sites were localized in each subject using a frameless stereotaxy system (Brainsight, Rogue Research, Montreal, Canada). Each subject was coregistered with their own structural magnetic resonance imaging (MRI) scan acquired on a Siemens 1.5T Vision MRI scanner with a standard circularly polarized head coil. The T1-weighted axial anatomical scans had a 1 × 1 × 2 mm resolution and were acquired by using an magnetization prepared rapid gradient echo (MPRAGE) sequence (time echo = 4 ms, time repetition = 9.7 ms, T1 = 300 ms, 128 partitions, field of view 250 × 250 × 256 mm). The FEF locations were identified on the MRI scans, and the scalp positions overlying the sites to be stimulated were marked on a cloth swimming cap which was worn throughout the test session. The location of the control site, the vertex (Cz), was also marked on the cap by taking the midpoints of the distance between the nasion and inion and the left and right pre-auricular points (Jasper 1958). Briefly, identification of the stimulation sites on the structural scans was achieved by the following procedure. Each individual MRI was normalized against a standard template using the FSL registration tool FLIRT (FMRIB, Oxford), which is a robust and accurate affine tool based around a multistart, multiresolution global optimization method and is described in detail by Jenkinson and Smith (2001). This resulted in a matrix which described the transformation applied to the structural scan to produce the normalized brain. This was then reverse applied to the coordinates for the FEFs (right FEF: 35, −7, 52; left FEF: −36, −9, 48, see e.g., Gagnon et al. 2006) to obtain the location of these sites in the original structural scan. These locations were then marked on the MRI scans in the Brainsight system (cf. Fig. 2).
We want to stress that this method of localizing the FEFs is neither the only one nor the most accurate. There are several alternative possibilities to determine the individual location, like functional imaging, behavioral probing (e.g., saccadic changes), or taking into account the individual sulcal anatomy. Taking into consideration the fact that the FEFs vary between subjects, individual functional MRI (fMRI) scans with exactly the same task to be used in the TMS study is the optimal method. However, we have used the method described here as it provides the advantage of minimizing the number of TMS pulses delivered (which would be elevated using a task to localize) and also avoiding the need to scan each subject individually. Furthermore, several studies now show that anatomical, fMRI and coordinate transform methods result in locations that agree well and we have confirmed that localization, for example by anatomical methods, results in effects on saccade latencies (see e.g., O'Shea et al 2007). The effect of “missing” the FEF would be to reduce the chances of seeing our predicted effect. Furthermore, the results presented in this paper are consistent with successful stimulation of the target area, and there are no other oculomotor areas in the direct vicinity of the FEF that contribute to smooth pursuit.
Subjects were instructed to pursue a continuously moving target (cross, size 0.5°) with their eyes. The target was presented centrally on a PC monitor (refresh rate 100 Hz, viewing distance 57 cm from a chin rest). The velocity profile of the target consisted of a constant component (carrier) at 4 different velocities (0, 8, 16, and 24 deg/s). A high-frequency sinusoidal oscillation (hum) with 4 Hz and ±8 deg/s velocity amplitude was superimposed. Three example trials of a 16 deg/s constant velocity condition are shown in Figure 2.
Each trial started with a 1-s fixation period at the center of the screen. The target then began to hum and to move either left- or rightward, or stayed at the center of the screen. In one fifth of all trials, the target moved at a constant velocity (16 deg/s) without superimposed hum. All parameters were completely randomized. The duration of the pursuit period during each trial was fixed to 4.5 s and when the target reached the maximum excursion of ±16 deg, the movement direction reversed. After the target crossed the center of the screen, TMS was applied at 10 Hz for 500 ms over the site being tested in that session, and the first pulse of the TMS train was synchronized with the 0° phase of the high-frequency hum (cf. Fig. 2). Each subject performed 3 sessions (carried out on different days), where TMS stimulation occurred over left FEF, right FEF, and Cz. Each session consisted of 14 blocks with 30 trials each. Because the 5 conditions were selected randomly prior to each trial, conditions occurred on average 84 times per session (obeying a normal distribution with a standard deviation of 7.6 in the range [63, 103]). The task was programmed in C++ (MS Visual Studio 6, Microsoft) utilizing the Eyelink software development kit.
All data were analyzed using custom analysis routines written in MatLab (The Mathworks Inc., Natick, MA) on a Linux workstation. Eye position data were filtered with a Gaussian lowpass (cut-off frequency 15 Hz) and 3-point differentiated to obtain the velocity traces. An iterative algorithm was applied on the eye velocity to separate saccades and slow-phase components (SPC). An estimate of the SPC of the eye-velocity trace is iteratively improved in 2 steps. In the first step, the SPC estimate is initialized to zero. The difference between the eye-velocity trace and the current SPC estimate serves as an estimate of the fast phase component (FPC). Saccades are detected when the FPC exceeds a threshold. This velocity threshold decreased from step to step (100, 20 deg/s). The beginning and end of each saccade are defined by the points in time before and after peak velocity when the FPC becomes zero. The SPC estimate for the next step is then computed by linear interpolation of the eye velocity across saccades and subsequent Gaussian low-pass filtering. The cut-off frequency of this low-pass is 1 and 10 Hz in the first and second step, respectively. Details of this procedure have also been described elsewhere (Ladda et al. 2007). The main advantage of this algorithm is its capability of separating very small saccades from the slow phase even during pursuit at high velocities.
Each subject's mean eye-velocity trace during the 500-ms TMS stimulation was calculated for each condition, that is, velocity and site of stimulation (cf. Fig. 2). Regions in the estimated slow-phase velocity traces that corresponded to saccades were cut out and excluded from averaging. Although subjects were instructed not to blink during stimulation, there were some trials which were contaminated by eyelid twitches. These were detected offline and excluded from averaging.
A delay of 50 ms between TMS and its effect was assumed; therefore the analysis window was shifted by 50 ms with respect to the TMS pulse onset. Similar delays can be found in recent TMS studies (Gagnon et al. 2006; Drew and van Donkelaar 2007).
In order to obtain an index for the response modulation, the mean eye-velocity traces were then transformed using the Fast Fourier Transform (FFT). The observation period of 500 ms corresponded to a frequency domain resolution of 2 Hz. Notwithstanding this coarse resolution, a clear peak at the stimulus frequency of 4 Hz was found that served as a measure for the response modulation. This procedure is very similar to fitting sinusoidal oscillations of 4 Hz in amplitude and phase to the mean eye-velocity traces, but the FFT-method also reveals whether there is a prominent 4 Hz component in the eye movement response at all, or if other frequencies are significantly present. As can be seen from Figure 3, there is a clear peak only at the hum frequency of 4 Hz (apart from the constant velocity component at 0 Hz).
All results were analyzed using a repeated-measures analysis of variance (RMANOVA) design and tested for significance by means of a group analysis. Subject S3 had to be excluded from the analysis due to 2 reasons. First, many saccades and eyelid twitches during TMS stimulation led to a very noisy estimate of the mean eye velocity resulting in a multiple peaked spectrum where no clear peak at 4 Hz could be identified. Second, estimating the gain by ignoring the low signal-to-noise ratio produced artifacts in the gain control curve (e.g., drop of the gain at 24 deg/s below the 8 deg/s level).
We either tested for effects on the average response modulation (as defined by the amplitude of the 4-Hz Fourier component during hum condition, see above), or the mean eye velocity during the 500 ms analysis window of the control condition (constant velocity target at 16 deg/s without superimposed hum). In case of violation of the sphericity assumption, we crosschecked significance with a multivariate approach.
To test for effects of FEF stimulation on the slope of the measured response modulation, we further applied an analysis of covariance (ANCOVA) which checks for significant differences between the linear regression parameters of the FEF and Cz stimulation conditions using a t-test. Prior to the ANCOVA, the mean response modulation of each subject was removed to eliminate effects which are due to large offset differences between subjects (this is comparable to the repeated-measures approach).
All results are based on measurements of either the response modulation at different tracking velocities (0, 8, 16, 24 deg/s) or the mean eye velocity during a constantly moving target at 16 deg/s without additional hum during stimulation of the FEFs or the control site Cz. An example for the differences in the eye movement response during stimulation of the FEF versus the control condition is depicted for subject S6 in Figure 4. Here, the mean responses during TMS for a carrier velocity of 24 deg/s are shown. The 500-ms TMS duration corresponds to 2 periods of the high-frequency oscillation. We also checked for effects on the phase of the response modulation, that is, response latency, but found neither significant main effects of TMS nor interactions between TMS and tracking velocity.
Quantifying Dynamic Gain Control
In order to confirm our method of measuring the dynamic gain control function, we first analyzed the changes in response modulation with respect to carrier velocity during stimulation of the control site Cz. Figure 5 displays the data for all 8 subjects as well as the group mean. A clear increase of the measured response modulation for increasing carrier velocities can be observed (F3,21 = 54.456, P < 0.001). This result confirms our method of quantifying the dynamic gain control, as this increase is in agreement with behavioral studies in monkeys (Keating and Pierre 1996) and humans (Churchland and Lisberger 2002).
Unilateral FEF ablation affects visually driven pursuit in both contralateral and ipsilateral directions (Keating 1991). Nevertheless, Shi et al. (1998) found that contralateral pursuit after inactivation was much less impaired than ipsilateral. In our experiment, the target motion during TMS was contralateral or ipsilateral, respectively, in approximately 50% of all trials. To test for directionality effects, we analyzed the response modulation with respect to stimulation side (FEF left and right), direction of target motion, and all different carrier velocities. This RMANOVA design utilizing 3 factors yielded no significant main effect for the direction of target motion. However, we found an effect of the side of FEF stimulation (F1,7 = 6.389, P = 0.04). The effect of left FEF stimulation on response modulation was about 10% larger than of right FEF stimulation. More importantly, we did not find any significant interaction effects between these factors. We also analyzed the mean eye velocity during pursuit of a constantly moving target at 16 deg/s without superimposed hum (control condition) and found neither significant main effects nor interactions between TMS site and direction of target motion. For further analysis, we pooled the data from left and right FEF stimulation and leftward and rightward target motion. Note that the pooling of the directions solely results in an attenuation of any directionality effects of eye velocity or response modulation.
TMS Effects on Steady-State Eye Velocity
The FEFs are substantially involved in the initiation and maintenance of SPEM. Lesions therein reduce the steady-state gain for predictive and visually guided pursuit eye movements drastically (Lynch 1987; Keating 1991; MacAvoy et al. 1991; Shi et al. 1998). We therefore assumed that disrupting neuronal activity by means of TMS stimulation of the FEFs will cause a slight decrease in steady-state gain during visually guided pursuit. In approximately one fifth of all trials, subjects had to pursue a constantly moving target at 16 deg/s without superimposed hum. We determined the mean eye velocity during stimulation of the FEFs and Cz for each subject. We found that the averaged mean eye velocity over all measured subjects was slightly reduced from 16.1 to 15.4 deg/s during stimulation of the FEFs as compared with the control condition. This small but highly significant (F1,7 = 17.210, P < 0.005) reduction corresponds to about 4.3% decrease in steady-state gain for visually guided SPEM.
TMS Effects on Response Modulation
As stated in the introduction, an unequivocal effect of disrupting the dynamic gain control mechanism is the reduction of the oscillatory component of the eye movement response during pursuit of a constantly moving target with a superimposed high-frequency hum. Two possible effects can be distinguished: 1) reduction of the overall gain, that is, a downward shift of the gain control curve and 2) a reduction of the gain control mechanism itself, resulting in a reduced slope of the gain control curve.
We first evaluated the effect of TMS on the oscillatory component of the eye movement response. As expected, we found a highly significant main effect (F1,7 = 17.445, P < 0.005) during stimulation of the FEFs as compared with the control condition. Stimulation of the FEFs reduced the average amplitude of the 4-Hz component to about 74% of its value. These data are presented for one subject (S6) in Figure 6.
More interestingly, if the FEFs are involved in the dynamic regulation of the gain, a disruption of neuronal activity should decrease the efficacy of the gain control mechanism. That is, a reduced effect of tracking velocity on the response modulation. From inspection of the single-subject data in Figure 6, it is already noticeable that TMS of the FEFs has an impact on the efficacy of dynamic gain control. In the population of all measured subjects, the group analysis yielded significant interaction effects between TMS site (FEF L/R, Cz) and tracking velocity (F3,21 = 4.527, P < 0.05). A more detailed ANCOVA (see Methods) revealed that TMS of the FEFs reduced the efficacy of the dynamic gain control by approx. 23.7% (T60 = 2.096, P < 0.05). Figure 7 shows the measured response modulation, averaged over all subjects, for stimulation of the FEFs and the control site Cz.
This work addressed the effect of TMS of the FEFs on SPEMs. Using a simple paradigm consisting of a constantly moving target at different carrier velocities with a superimposed high-frequency sinusoidal oscillation, we were able to provide straightforward evidence that the FEFs are directly involved in the dynamic regulation of gain during SPEMs.
By measuring the modulation of the eye movement response, we quantified the effect of applying TMS to the FEFs. There are 2 other TMS studies that deal with similar issues. Gagnon et al. (2006) analyzed smooth pursuit of predictive target motion and showed the importance of the FEF for transforming predictive signals into motor commands. In the other study, Drew and van Donkelaar (2007) were interested in the contribution of the FEFs to pursuit initiation. Our goal was to elaborate the role of the human FEFs in the dynamic regulation of the gain during SPEMs.
We first quantified the effect of FEF stimulation on the steady-state velocity gain during pursuit of a constantly moving target. Although the decrease in steady-state velocity was small, it was highly significant and comparable with previous results (Gagnon et al. 2006).
It has been shown in monkeys that electrical stimulation of the FEF enhances the responses to impulse-like perturbations of a stationary target (Tanaka and Lisberger 2001). (The opposing effects of response enhancement during electrical stimulation on the one hand and the reduction during TMS on the other hand are addressed below.) This means that the contribution of the FEFs to the pursuit response is functionally located at a position where an increase in activity yields an increase in the feedforward gain of the visual-motor transformation for pursuit.
Here, we showed that disrupting neuronal activity in FEF resulted not only in a decrease in the overall gain, but also in an attenuation of the efficacy of the dynamic gain control mechanism itself (as shown by the reduced slope of the measured response modulation in Fig. 7). This implies that the FEFs are not solely in a position to regulate the overall gain, but rather that they are directly involved in the dynamic regulation of the gain during smooth pursuit at different tracking velocities.
Analyzing the response gains with respect to carrier velocity implies that perceived target velocity is the primary controlling factor for smooth pursuit gain control. Another possibility is that an efference copy of the eye motor command and therefore eye velocity regulates the gain during smooth pursuit, as has been suggested by several studies (Keating and Pierre 1996; Carey and Lisberger 2004). However, as pursuit gain is close to one, the difference during intact steady-state pursuit is marginal. The situation looks different when considering the TMS condition. If TMS would merely affect steady-state eye velocity, which in turn changes the feedforward gain, analyzing the response gains re target velocity would yield spurious effects on the slope of the measured response modulation. Hence, the decrease in slope during TMS could theoretically result from a decrease in the steady-state proportion of eye velocity. An observed decrease in efficacy of x percent during TMS could therefore be caused by a mean eye-velocity reduction of x percent. However, as explained above, the average decrease in mean eye velocity amounts to approximately 4.3%, whereas the decrease in efficacy of the dynamic gain control is 23.7%. Therefore, the effect of TMS on the mean eye velocity is only marginal.
The estimated decrease in the efficacy of the gain control mechanism of about 23.7% is based on the evaluation of linear regression parameters. The underlying assumption, namely, that the gain depends linearly on perceived target or eye velocity is based on behavioral measurements in monkeys (Keating and Pierre 1996) and humans (Churchland and Lisberger 2002) and is clearly reflected in our dataset (Fig. 7). Furthermore, Ladda et al. (2007) derived an optimality constraint for simple velocity controllers (in essence, the SPEM system) that pursue targets undergoing a natural velocity scaling. In order to make its position error velocity-invariant, the servo gain must be adjusted proportionally to the tangential target velocity.
Dynamic gain control can be considered a major computational principle of sensorimotor transformations. The underlying neuronal substrate of gain modulation might be the so-called gain fields, which are found in several cortical areas, for example, in parietal cortex (Salinas and Thier 2000), frontal cortex (Boussaoud et al. 1993), and occipital cortex (Trotter and Celebrini 1999). For example, low-frequency repetitive TMS over the posterior parietal cortex during smooth pursuit impairs the steady-state gain, which could be explained by a desynchronization of gain field units that serve for estimating the target position in space (Hutton and Weekes 2007). It is conceivable that neurons in the FEFs also show a gain field-like behavior with respect to eye velocity and retinal image motion (Nuding et al. 2006). If this holds true, the measured effect of TMS on the response modulation can be readily understood as a disruption or desynchronization of the neuronal activities in the FEF gain field population that leads to a global inhibition of the gain control mechanism. In contrast, electrical stimulation of the FEFs can be seen as a local excitation of gain field units which increase the driving signal and therefore cause an increase in the feedforward gain (Tanaka and Lisberger 2001). The FEF gain control mechanism for pursuit exhibits similarities to the one reported for spatial visual attention and saccadic eye movements (Moore and Armstrong 2003). It modulates the sensitivity to changes in retinal image velocity during ongoing pursuit depending on current target velocity, and may thus be understood as an “attentional” mechanism within the pursuit system.
Bundesministerium für Bildung und Forschung (01GQ0440); Deutsche Forschungsgemeinschaft (KA 2284/2-1); The Royal Society to V.W.; and The Medical Research Council to N.M.
Conflict of Interest: None declared.