Peri-saccadic orientation identification performance and visual neural sensitivity are higher in the upper visual field

Visual neural processing is distributed among a multitude of sensory and sensory-motor brain areas exhibiting varying degrees of functional specializations and spatial representational anisotropies. Such diversity raises the question of how perceptual performance is determined, at any one moment in time, during natural active visual behavior. Here, exploiting a known dichotomy between the primary visual cortex and superior colliculus in representing either the upper or lower visual fields, we asked whether peri-saccadic orientation identification performance is dominated by one or the other spatial anisotropy. Humans (48 participants, 29 females) reported the orientation of peri-saccadic upper visual field stimuli significantly better than lower visual field stimuli, unlike their performance during steady-state gaze fixation, and contrary to expected perceptual superiority in the lower visual field in the absence of saccades. Consistent with this, peri-saccadic superior colliculus visual neural responses in two male rhesus macaque monkeys were also significantly stronger in the upper visual field than in the lower visual field. Thus, peri-saccadic orientation identification performance is more in line with oculomotor, rather than visual, map spatial anisotropies. Significance statement Different brain areas respond to visual stimulation, but they differ in the degrees of functional specializations and spatial anisotropies that they exhibit. For example, the superior colliculus both responds to visual stimulation, like the primary visual cortex, and controls oculomotor behavior. Compared to the primary visual cortex, the superior colliculus exhibits an opposite pattern of upper/lower visual field anisotropy, being more sensitive to the upper visual field. Here, we show that human peri-saccadic orientation identification performance is better in the upper compared to the lower visual field. Consistent with this, monkey superior colliculus visual neural responses to peri-saccadic stimuli follow a similar pattern. Our results indicate that peri-saccadic perceptual performance reflects oculomotor, rather than visual, map spatial anisotropies.

Introduction present in the time interval above, it was considered a saccade trial, and we 3 9 8 assessed saccade time relative to stimulus onset time for evaluating time courses of 3 9 9 neural suppression. 4 0 0 Spatial frequency tuning curves (i.e., responses for each given spatial 4 0 1 frequency) were described previously (Hafed and Chen, 2016; Chen and Hafed, 4 0 2 2017), but in this, study we analyzed how saccades influenced these curves 4 0 3 differently when the RF was either in the upper or lower visual field. To test the effect 4 0 4 of saccadic suppression in the upper and lower visual fields, we computed a 4 0 5 measure of "normalized firing rate". First, we calculated, for each trial, the peak firing 4 0 6 rate between 30 and 150 ms after stimulus onset. Then, for each neuron and spatial 4 0 7 frequency condition, we averaged the peak firing rate in trials in which no saccades 4 0 8 were detected. This value was then normalized by dividing the averages of each 4 0 9 spatial frequency condition by the preferred spatial frequency response of that 4 1 0 neuron, giving as a result the average tuning curve when no saccades were present. 4 1 1 Similarly, for each neuron and spatial frequency, we averaged all the trials in which 4 1 2 the gabor stimulus was presented 40 to 100 ms after saccade onset. The average 4 1 3 peak firing rate at each spatial frequency condition was then normalized by the peak 4 1 4 firing rate for the preferred spatial frequency response of the trials with no saccades. 4 1 5 Doing so, values lower than one indicated suppression of neural activity because of 4 1 6 saccade generation. 4 1 7 To summarize the time courses of saccadic suppression of SC visual bursts in 4 1 8 the upper and lower visual fields (e.g. Fig. 9 in Results), we selected all the trials in 4 1 9 which the gabor stimulus was presented between -50 to 140 ms relative to saccade 4 2 0 onset. We then smoothed the data by applying a running average window of 50 ms 4 2 1 on the normalized peak firing rate (relative to the baseline firing rate of for that spatial 4 2 2 frequency) and by moving the average time window in steps of 10 ms. This analysis 4 2 3 was performed only for the lower spatial frequency grating (0.56 cpd), which was the 4 2 4 one used in the behavioral experiment reported above. To statistically test the 4 2 5 difference between the upper and lower visual fields, we ran a series of two-sample 4 2 6 independent t-tests at each bin of the two curves, and we set a threshold of 0.01 to 4 2 7 determine whether a p-value was low enough to reject the null hypothesis. We first asked whether human orientation identification performance around the time 4 3 3 of saccades is different for upper or lower visual field peri-saccadic stimuli. In a first 4 3 4 experiment (Experiment 1; diffuse attention condition), subjects generated 4 3 5 approximately 12 deg horizontal saccades to the right or left of central fixation ( Fig.  4 3 6 1A). At different times relative to saccade onset, a brief flash lasting approximately 4 3 7 16.7 ms was presented. The flash was centered horizontally at the midpoint between 4 3 8 the initial fixation target location and the final desired saccade endpoint (that is, 4 3 9 halfway along the intended saccade vector), and it consisted of two vertically-aligned 4 4 0 image patches (each at 3 deg above or below the screen center). One patch was the 4 4 1 target to be detected by the subjects, and it was either a horizontal or vertical gabor 4 4 2 grating. The other patch was an irrelevant distractor without inherent orientation 4 4 3 information (it was a superposition of two orthogonal gabors, with the total pattern 4 4 4 tilted by 45 deg; Materials and methods). Across trials, the oriented patch was 4 4 5 placed either above (upper visual field target location) or below (lower visual field 4 4 6 target location) the horizontal meridian, and the other patch was at the vertically-4 4 7 symmetric position. There was also an equal likelihood of horizontal and vertical 4 4 8 patches at each of the two locations. The subjects were instructed to report the 4 4 9 orientation of the target flash (horizontal or vertical), and we assessed whether their 4 5 0 performance differed as a function of target location. 4 5 1 Across 20 subjects, we found that peri-saccadic orientation identification 4 5 2 performance was consistently better for upper visual field target locations when intervals long before or after the eye movements, performance was close to ceiling 4 5 7 levels. However, in the critical peri-saccadic interval in which saccadic suppression 4 5 8 was to be expected (Latour, 1962;Matin, 1974; Binda and Morrone, 2018), we found 4 5 9 that the proportion of correct trials was significantly higher in the upper visual field 4 6 0 than in the lower visual field (red asterisks; GLMM, main effect of target gabor 4 6 1 location, p<0.01; see Materials and methods). Therefore, peri-saccadic orientation 4 6 2 identification performance was significantly better in the upper visual field, unlike 4 6 3 known lower visual field superiority of perceptual performance in the absence of  interval also included trials where the gabor grating was projected on screen while the eyes started 4 8 9 moving towards the landing fixation point). We have added a visual representation of a single bin as a 4 9 0 small grey box in panels B and C, centered at the time of -7 ms, to clearly showcase the temporal 4 9 1 extent of the bin.

9 2
Therefore, the retinal conditions of the flashes were similar for upper and 5 7 8 lower visual field targets, meaning that the results of Fig. 1 were not trivially 5 7 9 explained by systematically different retinotopic stimulation between conditions. were closest to the fovea because the flash was always midpoint along the saccade path and timed to 5 8 6 frequently occur peri-saccadically. However, and most critically, the distance to the fovea was not 5 8 7 different for upper and lower visual field targets (compare yellow and blue curves in each panel).

8 8
Therefore, the results of Fig. 1 were not due to a visual acuity benefit for upper visual field targets due 5 8 9 to retinal eccentricity. Error bars denote s.e.m. (C, D) Similar analysis but for the retinal slip of the 5 9 0 images during their onset (that is, the displacement of the gabor during its presentation). Because the 5 9 1 eye was moving during a saccade, the grating slipped in position on the retina. However, once again, 5 9 2 such retinal slip was the same for upper (yellow) and lower (blue) visual field targets in both 5 9 3 experiments. Note that the gray shaded regions with center at time -7 ms are similar to those shown 5 9 4 in Fig. 1, and they illustrate the extent of our binning window for one example sample point. They also 5 9 5 explain why the retinal slip may have appeared to start increasing even before saccade onset (this 5 9 6 was only a consequence of our temporal binning windows and not because of erroneous saccade 5 9 7 onset detection). 5 9 8 5 9 9 Finally, gabors orthogonal to the executed saccade can potentially be harder 6 0 0 to resolve than parallel gabors (Castet et al., 2002;Schweitzer and Rolfs, 2020). 6 0 1 Hence, a difference between the proportion of horizontal and vertical gabors in the 6 0 2 lower and upper visual field target conditions could potentially also bias our results. 6 0 3 While this is unlikely given our balanced experimental design (Materials and 6 0 4 methods), to control for this possibility, we additionally analyzed the proportion of 6 0 5 horizontal gabors along the peri-saccadic interval and across the tested experimental 6 0 6 conditions; we did this using the same time course techniques adopted for 6 0 7 orientation identification performance ( Fig. 1), saccadic amplitude ( Fig. 2C, D), 6 0 8 distance of gabor from the fovea, and displacement of gabor on retina (Fig. 3). In 6 0 9 Experiment 1, the proportion of trials in which a horizontal gabor was presented 6 1 0 hovered around 50% in the interval from -100 ms to around 90 ms, as could be 6 1 1 expected from a balanced two-alternative forced choice task, with no discernible 6 1 2 difference between trials presented in the upper or lower visual field. In Experiment 6 1 3 2, the proportion of trials in which a horizontal gabor was presented hovered around 6 1 4 33%, compatible with a three-alternative forced choice task. Also in this case, no 6 1 5 systematic difference between trials presented in the upper of lower visual field could 6 1 6 be observed. Thus, our results in Fig. 1 cannot be trivially accounted for by a 6 1 7 systematic occurrence of more targets that are easier to discriminate in the upper 6 1 8 versus lower visual field. 6 1 9 To summarize, our results so far demonstrate that peri-saccadic orientation 6 2 0 identification performance is significantly higher in the upper rather than lower visual 6 2 1 field. These observations complement prior work by Knöll et al. (Knöll et al., 2011), 6 2 2 who have documented the topography of saccadic suppression along the horizontal 6 2 3 meridian. These authors found that peri-saccadic suppression occurred in a 6 2 4 retinotopic frame of reference, with a divisive reduction of sensitivity that was 6 2 5 constant across the retinal eccentricity dimension. To our knowledge, peri-saccadic 6 2 6 orientation identification performance in the upper and lower visual fields, around the 6 2 7 time of a horizontal eye movement, has not been reported previously. Perhaps the strongest evidence that better upper visual field peri-saccadic 6 3 3 orientation identification performance was a robust phenomenon emerged when we 6 3 4 gave our subjects, within contiguous blocks of trials, valid prior knowledge about the 6 3 5 upcoming target location. Specifically, in approximately one quarter of all trials in 6 3 6 each experiment (Materials and methods), the subjects were explicitly told that the 6 3 7 current block of trials had primarily only upper visual field targets (with 97% 6 3 8 probability). Similarly, in another one quarter of the trials, the subjects were informed 6 3 9 that the current block of trials had primarily lower visual field target locations (with 6 4 0 97% probability). We called these blocked trials the focused attention trials. In both 6 4 1 cases, orientation identification performance in the peri-saccadic interval was still 6 4 2 higher in the upper visual field than in the lower visual field. This result is shown in 6 4 3 Fig. 4. That is, even when the subjects fully knew in advance that a target was going 6 4 4 to appear in the lower visual field, their peri-saccadic orientation identification 6 4 5 performance to such a target was still lower than the orientation identification 6 4 6 performance for targets in the upper visual field. Note also that eye movement  Of course, the results of Fig. 4 were not entirely only a negative result with 6 6 0 respect to the blocking manipulation of target position. For example, when we 6 6 1 compared orientation identification performance long before saccade onset (-200 to -6 6 2 70 ms from saccade onset) in the diffuse and focused attention conditions, both 6 6 3 experiments were suggestive of a perceptual benefit when prior knowledge about 6 6 4 target location was provided. For example, in Experiment 1, the subjects exhibited 6 6 5 88% average correct rates with prior knowledge of target location (focused attention 6 6 6 trials) when compared to 86% average correct rates without prior knowledge 6 6 7 (p=0.055, paired t-test). In Experiment 2, the average correct rates were 91% and 6 6 8 88% in the diffuse and focused attention trials, respectively (p=0.017, paired t-test). 6 6 9 Therefore, the lack of influence of advanced prior knowledge on peri-saccadic 6 7 0 orientation identification performance alluded to above (Fig. 4) was primarily 6 7 1 restricted to the peri-saccadic interval. 6 7 2 It is important to note here that while it is not possible to perform a direct 6 7 3 comparison between performance in Experiment 1 and Experiment 2 (one was a 6 7 4 two-alternative and the other was a three-alternative forced choice paradigm), we did 6 7 5 collect both diffuse and focused attention trials within each of these experiments 6 7 6 (Materials and methods). Therefore, we also had an opportunity to compare peri-6 7 7 saccadic focused and diffuse attention effects within each experiment. For this 6 7 8 reason, we plotted orientation identification performance within the +/-20 ms interval 6 7 9 around saccade onset, summarizing this comparison (Fig. 5). In both experiments, 6 8 0 peri-saccadic orientation identification performance was significantly higher in the 6 8 1 upper rather than the lower visual field, irrespective of attention condition. Moreover, 6 8 2 within each experiment, the effect size for the focused and diffuse attention 6 8 3 conditions was similar, and no interaction between target location and attention were also made when we calculated d' based on all trials within each experiment in 6 9 0 the intervals shown in Fig. 5 (Materials and methods). 6 9 1 We also note here that m-AFC tasks force observers to give a response on 6 9 2 every trial, and authentic correct responses might be mixed up with lucky guesses 6 9 3 arising in cases in which the observer could not gather or use sensory information to 6 9 4 give an informed response. For future measurements it could be interesting to allow 6 9 5 observers to give "don't know" responses instead of producing a guess on trials in 6 9 6 which they have no basis for responding, and explicitly model those guess 6 9 7 responses, while fitting the authentic judgements (García-Pérez and Alcalá-6 9 8 Quintana, 2019; Reynolds et al., 2021). In the current measurements, we were fully 6 9 9 cognizant of the need to decrease noise in the data. That is why we ran our second 7 0 0 experiment, employing a 3-AFC task and limiting the issues associated with line). In both cases, peri-saccadic perceptual performance was significantly higher in the upper rather 7 1 4 than the lower visual field. It is important to note that within each experiment, the effect size for the 7 1 5 focused and diffuse attention conditions was similar, and no interaction between target location and 7 1 6 attention condition was observed.

1 7
Orientation identification performance is not higher in the upper visual field during 7 1 8 fixation 7 1 9 In a control experiment, we next explicitly tested whether our results could arise 7 2 0 simply during fixation as well, in which case there would be nothing special about the 7 2 1 peri-saccadic results reported above. We required participants to keep fixating the 7 2 2 center of the screen while gabors were presented in the upper or lower visual field 7 2 3 (on the vertical meridian). Thus, retinotopically, the gabors were at a similar location 7 2 4 to that in the peri-saccadic interval at maximal perceptual suppression (Figs. 2, 3). 7 2 5 This provided a better comparison than testing orientation identification performance 7 2 6 in the main experiments long before saccade onset, since the gabors were not on 0 neurons with either upper or lower visual field RF's. Therefore, the eye movement 8 0 1 characteristics were similar regardless of whether we recorded upper or lower visual 8 0 2 field SC neurons. hotspot locations from our recorded population, expressed as a direction from the horizontal meridian.

0 7
Approximately half of the neurons had RF hotspots in the upper visual field (yellow), and the rest had When we then inspected the neurons' visual responses themselves, we 8 2 6 observed consistently higher SC firing rates in the upper visual field neurons than in 8 2 7 the lower visual field neurons for peri-saccadic stimuli. Consider, for example, the response was suppressed in association with microsaccades, as expected but such 8 3 7 suppressed response was still robust and peaking above 200 spikes/s. On the other 8 3 8 hand, the neuron in Fig. 7D represented a lower visual field location (its RF map is 8 3 9 shown in the inset). Not only was its baseline visual response (in the absence of (Hafed and Chen, 2016), but its peri-saccadically suppressed response (dark curve) 8 4 2 was also more strongly affected by the eye movements. In other words, the neuron 8 4 3 experienced stronger saccadic suppression than the neuron in the upper visual field, 8 4 4 consistent with our perceptual results above. Thus, if anything, the spatial anisotropy 8 4 5 in the SC in terms of upper versus lower visual field neural sensitivity (Hafed and 8 4 6 Chen, 2016) was amplified even more during peri-saccadic intervals. 8 4 7 We confirmed this by isolating a measure of saccadic suppression and 8 4 8 confirming that it was stronger for lower rather than upper visual field SC neurons. 8 4 9 Across the population, we normalized each neuron's activity by its strongest no-8 5 0 microsaccade visual response to any of the five different spatial frequencies that we 8 5 1 tested (Chen and Hafed, 2017); that is, we picked the spatial frequency that evoked 8 5 2 the strongest peak response, and we normalized all trials' firing rate measurements 8 5 3 by this value (Materials and methods). We then normalized each neuron's peri-8 5 4 saccadically suppressed visual response using the very same normalization factor, tuning curves than the lower visual field neurons in the peri-saccadic interval, and 8 6 5 they were suppressed less than the lower visual field neurons at the low spatial 8 6 6 frequencies. For example, at the lowest spatial frequency (0.56 cpd), there was 8 6 7 weaker saccadic suppression in the upper visual field neurons ( Fig. 8A) than in the 8 6 8 lower visual field neurons (Fig. 8B); this is evidenced by the larger difference 8 6 9 between the blue and dark blue curves in Fig. 8B than between the yellow and dark 8 7 0 yellow curves in Fig. 8A (p=0.038, two-sample t-test). 8 7 1 At higher spatial frequencies, the saccadic suppression effect was expectedly 8 7 2 weakened overall (Chen and Hafed, 2017), but this weakening again happened 8 7 3 more so for the upper visual field neurons than for the lower visual field neurons (for 8 7 4 example, the difference between the curves at 2.22 and 4.44 cpd was smaller in the 8 7 5 upper visual field, panel A, than in the lower visual field, panel B). Coupled with the 8 7 6 fact that the neurons were themselves more sensitive in the upper visual field in the 8 7 7 no-microsaccade trials (Hafed and Chen, 2016) (e.g. Fig. 7), this suggests that there 8 7 8 was a consistently higher firing rate in the SC visual bursts in the upper visual field 8 7 9 for peri-saccadic stimuli. yellow) and peri-saccadically (dark). Error bars denote 95% confidence intervals. In both curves, we 8 8 7 normalized each neuron's firing rate to the peak visual response for the preferred spatial frequency 8 8 8 (Materials and methods). Lower spatial frequencies experienced more suppression for peri-saccadic in the lower visual field (Fig. 7), but they also experience stronger saccadic 1 0 4 5 suppression in the peri-saccadic intervals (Figs. 8,9). Therefore, the already strong 1 0 4 6 disparity in neuronal visual sensitivity between the upper and lower visual fields in 1 0 4 7 the SC (Hafed and Chen, 2016) is rendered even stronger peri-saccadically.  (  2  0  2  3  )  U  s  e  a  n  d  m  i  s  u  s  e  o  f  c  o  r  r  e  c  t  i  o  n  s  f  o  r  m  u  l  t  i  p  l  e  t  e  s  t  i  n  g  .  M  e  t  h  o  d  s  i  n  1  1  3  8  P  s  y  c  h  o  l  o  g  y  8  :  1  0  0  1  2  0  .  1  1  3  9  G  a  r  c  í  a  -P  é  r  e  z  M  Á  ,  A  l  c  a  l  á  -Q  u  i  n  t  a  n  a  R  (  2  0  1  9  )  T  h  e  d  o  '  s  a  n  d  d  o  n  '  t  s  o  f  p  s  y  c  h  o  p  h  y  s  i  c  a  l  m  e  t  h  o  d  s  1  1  4  0  f  o  r  i  n  t  e  r  p  r  e  t  a  b  i  l  i  t  y  o  f  p  s  y  c  h  o  m  e  t  r  i  c  f  u  n  c  t  i  o  n  s  a  n  d  t  h  e  i  r  d  e  s  c  r  i  p  t  o  r  s  .  T  h  e  S  p  a  n  i  s  h  1  1  4  1  j  o  u  r  n  a  l  o  f  p  s  y  c  h  o  l  o  g  y  2  2  :  E  5 R  o  s  s  J  ,  M  o  r  r  o  n  e  M  C  ,  B  u  r  r  D  C  (  1  9  9  7  )  C  o  m  p  r  e  s  s  i  o  n  o  f  v  i  s  u  a  l  s  p  a  c  e  b  e  f  o  r  e  s  a  c  c  a  d  e  s  .  N  a  t  u  r  e  1  2  3  0  3  8  6  :  5  9  8  -6  0  1  .  1  2  3  1  R  o  s  s  J  ,  M  o  r  r  o  n  e  M  C  ,  G  o  l  d  b  e  r  g  M  E  ,  B  u  r  r  D  C  (  2  0  0  1  )  C  h  a  n  g  e  s  i  n  v  i  s  u  a  l  p  e  r  c  e  p  t  i  o  n  a  t  t  h  e  t  i  m  e  1  2  3  2  o  f  s  a  c  c  a  d  e  s  .  T  r  e  n  d  s  i  n  n  e  u  r  o  s  c  i  e  n  c  e  s  2  4  :  1  1  3  -1  2  1  .  1  2  3  3  R  u  b  i  n  N  ,  N  a  k  a  y  a  m  a  K  ,  S  h  a  p  l  e  y  R  (  1  9  9  6  )  E  n  h  a  n  c  e  d  p  e  r  c  e  p  t  i  o  n  o  f  i  l  l  u  s  o  r  y  c  o  n  t  o  u  r  s  i  n  t  h  e  1  2  3