Involvement of superior colliculus in complex figure detection of mice

  1. J Leonie Cazemier
  2. Robin Haak
  3. TK Loan Tran
  4. Ann TY Hsu
  5. Medina Husic
  6. Brandon D Peri
  7. Lisa Kirchberger
  8. Matthew W Self
  9. Pieter Roelfsema
  10. J Alexander Heimel

  • Curated by eLife

    eLife logo

    eLife assessment

    The authors present a valuable work suggesting that the superficial, retinorecipient layers of the mouse superior colliculus (SC) may participate in figure-ground segregation and object recognition. These data are based largely on optogenetic perturbations of SC but the strength of evidence is currently incomplete: although the effects are statistically significant, there are significant technical limitations that are not adequately addressed via controls.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Object detection is an essential function of the visual system. Although the visual cortex plays an important role in object detection, the superior colliculus can support detection when the visual cortex is ablated or silenced. Moreover, it has been shown that superficial layers of mouse SC (sSC) encode visual features of complex objects, and that this code is not inherited from the primary visual cortex. This suggests that mouse sSC may provide a significant contribution to complex object vision. Here, we use optogenetics to show that mouse sSC is involved in figure detection based on differences in figure contrast, orientation, and phase. Additionally, our neural recordings show that in mouse sSC, image elements that belong to a figure elicit stronger activity than those same elements when they are part of the background. The discriminability of this neural code is higher for correct trials than for incorrect trials. Our results provide new insight into the behavioral relevance of the visual processing that takes place in sSC.

Article activity feed

  1. Version published to 10.7554/elife.83708 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Pulic Review):

    The authors aimed to understand whether the superficial, retinorecipient layers of the mouse superior colliculus (sSC) participate in figure-ground segregation and object recognition. To address this question, they use a combination of optogenetic perturbations of sSC and recordings. These data are consistent with SC being causally involved in object recognition. This would be useful information for the field and likely to be cited.

    Thank you for your positive evaluation.

    However, I have several concerns regarding their conclusions.

    A significant limitation of this study is methodological. The major novelty is the effect of optogenetic silencing, because the recordings are largely correlative, but the optogenetic silencing approach lacks appropriate controls for the effects of the optogenetic excitation light. The authors acknowledge that the optogenetic light is a potential confound, but attempt to address this by shielding the fiber to eliminate light leak and strobing a blue led in the arena. The former does not account for the effects of excitation light scattering intracerebrally--during optogenetic experiments, intracerebral scattering causes the eyes to light up--and for the latter, there is no way to compare the intensity or qualia of the externally strobed LED and the intracerebral light. The proper control would be a cohort of mice lacking channelrhodopsin expression in sSC. Regardless, it is essential to acknowledge this potential confound.

    This is a good point. We have added discussion of this in lines 90-95. The proposed experiment was done in Kirchberger et al. (Sci Adv 2021, Suppl Figure 3). In mice without expression of channelrhodopsin trained on the same task as in our study, blue laser light in the cortex did not affect accuracy. Although the exact location of these fibers is different from ours, the distance from the fiber to the eye is very similar. Furthermore, in answer to this comment, we have done a new set of experiments with 4 wild type mice, in which we recorded neural activity in the sSC while delivering optogenetic light stimulation. The procedure was similar to our previous experimental animals except that they did not receive a virus injection. In these mice, we did not see any response in the superior colliculus to the laser light, but we noticed a 5% reduction in response to the visual stimuli (new Figure 1—figure supplement 3). This small reduction could be a small reduction of contrast of the visual stimulus due to the laser light hitting the retina, but given that we did not see any response to the laser alone, it is more likely to come from the known inhibiting effects of light on neural activity (e.g. through heat, see Owen et al. Nat Neurosci 2019). Because our aim was to silence sSC, this particular effect is not a strong confound for our study.

    Relatedly, as the authors note, there are GABAergic projection neurons in sSC that may be driving these effects via gain of function. This is a significant concern that has limited the widespread adoption of this approach in sSC despite its popularity in studies in cortex. Indeed, one recently published study of behavioral functions of deep SC found that activating inhibitory neurons actually caused paradoxical behavioral effects consistent with gain of function in the targeted hemisphere, due to the effects of long-range inhibitory projections on the other SC hemisphere. Given the presence of inhibitory projections in sSC, it would be preferable to use an orthogonal method for silencing and at least to thoroughly acknowledge these concerns and cite these recent studies.

    This is a valid point. When we started our study, we had some experience with inhibitory opsin (archaerhodopsin and halorhodopsin) and were not confident that we could widely inhibit the sSC reversibly, repeatedly and consistently for an extended period. Other labs have now shown this is feasible with improved inhibitory opsins, so this would now be our preferred option too. The method of silencing sSC by inhibition of GABAergic neurons, however, is still the most common optogenetic method to silence sSC for an extended period (e.g. Hu et al. Neuron 2019, Brenner et al. Neuron 2023) .

    We thank the reviewer pointing us to recently published paradoxical behavioral effects. These effects, that we found in Essig et al. (Comm. Biol. 2021) are very interesting, but are not really a concern for the interpretation of our results, partially because as the reviewer pointed out, the GABAergic neurons activated there were in the deep and intermediate layers of the SC, below the sSC that we targeted. The paradoxical effects in that manuscript were attributed to direct inhibition of the contralateral superior colliculus. In our case, we activated the inhibitory neurons bilaterally, and this interhemispheric GABAergic connectivity, if it extends to sSC, only strengthened the bilateral silencing of the sSC. However, we have now discussed the possibility of our transfection of these deeper GABAergic neurons (lines 272-278). The more general point that activating GABAergic neurons in the sSC may also cause inhibition in other structures is indeed a concern. GABAergic neurons in the sSC project to the PBG and the LGN (in particular the vLGN) (Gale & Murphy, 2014; Whyland et al., 2019; Li et al., 2023). Although the primary effect of our manipulation is silencing of the superior colliculus, including the GABAergic neurons (see our answer further below), we cannot exclude the possibility that activating these extracollicular GABAergic projections has an effect. We have edited our discussion of this and updated the references (lines 268-272). However, our measurements in anesthetized (previous submission) and in awake mice (new Figure 1—figure supplement 2) show that apart from a short period directly after the onset of the laser, also almost all putative GABAergic neurons are reduced in their response (see also our answer to the next comment).

    A minor point is that although activation of GABAergic neurons in sSC is expected to cause inhibition of neighboring neurons, I would expect channelrhodopsin-expressing GABAergic cells to show an increase in firing during optogenetic excitation. However, it seems that none of the cells plotted (assuming each point in Supplementary Fig 4D is a cell, which the legend does not specify) had such an increase. Do these extracellular recordings not detect inhibitory neurons well?

    This is indeed an intriguing observation. The data in the original figure (Supp Fig 1D) was spiking data from 15 recording sites and not from sorted units. This was mentioned in panel C, but not in the caption. For the purpose of the amount of silencing, there was no need to sort single units. Still, it is surprising to see the reduction on almost all channels. The data of Supp Fig 1D came from experiments in anesthetized mice. Prompted by a question from another reviewer, we have now redone these experiments in head-fixed awake mice. The new Figure 1—figure supplement 2 shows these results, for single- and multi-unit clusters. In response to a short laser pulse (50 ms), we find that many units significantly increase their firing rate (Figure 1—figure supplement 2A-B). However, almost all activated then reduce there firing rate and again, we see an overall reduction of responses to visual stimuli. Only one unit fires significantly more when the laser is on during the period of visual stimulation compared to when the laser is off, and the overall firing rate is strongly reduced (Figure 1—figure supplement 2C-E). It appears that optogenetically activating the inhibitory neurons in the sSC for a longer period also reduces the activity of these neurons. The effect that we are seeing might be similar to the paradoxical effects that may occur in visual cortex, where additional excitation of inhibitory neurons leads also leads to their reduced activity due to network dynamics (see e.g. Sadeh & Clopath, Nat Neurosci Rev 2021). However, the effect may also be due to a few inhibitory neurons having a strong inhibitory effect on other inhibitory neurons. This is an interesting point worthy of more investigation, but it falls out to scope of this manuscript.

    Finally, the relationship between these stimuli and objects is not entirely clear. The authors acknowledge this but it would be worthwhile to devote more attention to this point. In effect, as the authors note, the gray screen and sinuisoidal grating do not have any sharp edges on the screen, whereas each of the behaviorally relevant stimuli will create a sharp, step-like edge on the screen. Whether edge detection is truly object detection or simply a variant of more general visual detection is unclear.

    Indeed, the task can be solved by detection of texture edges, and it is not necessary to integrate the edge components into an object to successfully perform the task. A linear decoder fed with simple cell-like inputs is able to do the orientation task (Luongo et al., 2023). The same network failed to learn the phase task, but also the image of a phase-defined figure contains features that are not present in the background image, and could be solved by learning only local features. Even the texture-defined figures used in Kirchberger et al. (2021) and in earlier monkey studies (Lamme, 1995) which do not contain any sharp stimulus edges can be detected without integrating the local edges into objects and segregation the figure from the background. Several monkey studies show that late neuronal responses in V1 are enhanced for neurons with receptive fields on what we, humans, perceive as the figure. This effect has also been seen in mouse V1, even in the case where there are no local features distinguishing the figure from the background (Fig 7. in Kirchberger et al. 2021). Interfering with activity in V1 in this late phase reduces the ability to detect the figure in human (by TMS) and mouse (by optogenetics). This is suggestive that this figure-ground modulation is used in solving the task, but not a proof. To understand if mice solve the tasks by detecting a figure or by detecting specific features, we can look at generalization. Mice were previously shown to generalize to some degree for size, position and spatial phase of the figure grating patch (Schnabel et al., 2018), suggesting that the mice did not train to detect specific features at specific locations. Rats trained on a similar task had difficulty generalizing from a luminance-defined object to an orientation-defined object (De Keyser et al., 2015), as do mice (Khastkhodaei et al., 2016), but once the rats were acquainted with one set of oriented figures, they immediately generalized to other texture-orientations above chance. On a slightly different figure-detection task mice also showed generalization for different orientations once the initial task was learned (Luongo et al. 2023). This suggests that at least some generalization to object detection occurs in this task. We have added these observation to the discussion (line 301-305).

    Reviewer #2 (Public Review):

    The goal of this study is to show that the superficial superior colliculus (sSC) of mouse signals figure-ground differences defined by contrast, orientation, and phase, and that these signals are necessary for the animal to detect such figure-ground differences. By inhibiting sSC while the animals perform a figure-ground detection task, the study shows that detection performance decreases when sSC activity is suppressed during the onset of the visual stimulus. The study then intends to show that sSC neurons exhibit surround suppression based on orientation differences, and that surround suppression is stronger when the animal detects the correct location of the figure on the background.

    The major strength of this study is the use of a behavioural paradigm to test detection performance of figure-ground stimuli while manipulating neural activity in the sSC during different times after stimulus onset. This paradigm would show whether activity in the sSC is relevant for performing the task. Secondly, the study collected data to confirm previous findings: sSC neurons exhibit orientation specific surround suppression. Additionally, it is impressive that the authors were able to train mice to generalize their task performance across different stimulus categories (figure-ground differences in orientation and phase). This should be highlighted as it may inform future studies.

    Thank you for your positive evaluation. We have extended our discussion on the generalization in object detection tasks in mice.

    The study has, however, methodological and analytical weaknesses so that the stated conclusions are not supported by the presented results.

    1. Optogenetic inhibition is not limited to sSC (even expression may not be limited) About 30% of inhibitory neurons in the sSC project to other areas, e.g. ventral LGN, parabigeminal nucleus and pretectum (Whyland et al, 2019, see ref in manuscript). This means that these areas receive direct inhibition when inhibitory sSC neurons are optogenetically stimulated. This fact is mentioned in the discussion but the consequences and implications for the results are ignored. This is a major flaw of the optogenetic experiments of this study. Additionally, no evidence is given that opsin expression was limited to the superficial layers (except for one histological slice), which the authors acknowledge in line 285. Deeper layers may have other inhibitory neurons with long-range projections.

    The finding that sSC neurons show no figure-ground modulation for phase while the optogenetic manipulation has behavioural effects may be an indication for other areas being affected by the optogenetic manipulation.

    This is a valid point, also raised by reviewer 1. Although the primary effect of activating the GABAergic neurons in the sSC is a strong reduction of activity in the sSC (see also new figure S1), we cannot rule out that we also activate GABAergic neurons below the sSC and that there are some effects of activating GABAergic connections to the LGN and PBG. We have extended our discussion of this point in lines 269-277. However, as shown in new Figure 1—figure supplement 2, the effect of optogenetically activating Gad2-positive neurons appears to lead to a counter-intuitive reduction of their activity. This effect has previously been observed in cortex.

    1. Could other behavioural variables explain the results?

    a) Are there any task events other than the visual stimuli that the mice could use to make their decisions? The authors state the use of a custom made lick spout but it is not clear how this spout works, i.e. how do mechanics of the spout deliver water to the right versus the left output and could the mouse perceive these mechanics?

    We believe there were no task events besides the visual stimuli that the mice could use to make their decisions. The lick spout was Y-shaped (see Figure 1B) to facilitate the two-alternative forced choice task. Each side of the lick spout was connected to a separate water tube. The water flow in each tube was controlled using a valve. Also, each side of the lick spout was connected to its own lick detector wire. The two valves and the two detector wires were connected to an Arduino which was controlled by our MATLAB task script. The task script was coded such that, when the lick of the mouse had been on the correct side, the valve controlling the water flow on the correct side would briefly open to deliver the water reward. To summarize, the water would only flow after the mouse had licked and if the first lick had been on the correct side. Hence, the water reward did not produce additional cues. We have edited the description of the lick spout in the Methods section to make the functioning of the lick spout more clear (lines 511-513).

    b) Could the different neural responses to figure versus ground shown in Fig 2I-J and Fig 3B be explained by behaviours varying between the trial types, e.g. by early lick movements (which are conceivable even if the spout is not present), eye movements or changes in pupil-linked arousal? A behavioural difference seems even more likely to occur between hit and error/miss trials (Fig 4). If these behaviours were not measured, the possibility of behavioural modulation should be discussed.

    In the awake behaving electrophysiology experiments, the lick spout was not present until 500 ms after stimulus onset, so the mouse could not lick the spout. We did not record whisking or other face and jaw movements, hence we cannot say for sure whether the mice performed early ‘licks’ in the absence of the lick spout. We did, however, add a supplementary figure showing the licking behavior of the mice in the optogenetic interference experiments (see Figure 1—figure supplement 5). In this experiment, the lick spout was present at all times so all early licks would be recorded. Any licks before 200 ms after stimulus onset were disregarded as this would be too early for the decision to include knowledge about the stimulus. Figure 1—figure supplement 5B shows that the mice indeed only performed very few early licks as they probably knew this would not yield reward. The mice that performed the awake electrophysiology experiments were trained on the same task as these mice before introducing the lick spout delay of 500 ms. So although we cannot rule out early licks during electrophysiology, we think early licks would be an unlikely explanation for the neural response differences.

    We have added a new supplementary figure (Figure 2—figure supplement 2) showing data for eye movements and pupil dilation during the tasks. We had excluded all trials where the mice performed eye movements between 0-450 ms after stimulus onset, and indeed we saw no eye movements during the peak of the visual response (0-250 ms). Furthermore, the pupil dilation of the mice also did not change in this period.

    All in all, we view it as unlikely that the differences in neural activity in sSc were caused by either licking, eye movements or pupil-linked arousal.

    1. What is the behavioural strategy of the animals? Only licks beyond 200 ms after stimulus onset determine the choice of the animal because "mice made early random licks" from 0 to 200 ms. To better understand the behavioural strategies of the animals we need to see their behavioural data, i.e. left and right licks aligned to stimulus onset. It would be particularly interesting to see how number and latency of licks changes during optogenetic manipulation.

    Based on these suggestions, we investigated the licking behavior of the mice during the optogenetic experiments in more detail. Our new Figure 1—figure supplement 5 taught us several things:

    1. The fully trained mice hardly perform any early licks; they seem to understand that early licks cannot yield reward.

    2. The mice typically only lick one side of the lick spout during one trial. In correct trials the fluid reward is given directly after a correct lick, which causes the mouse to lick the correct side of the spout even more. However, even if the first lick is incorrect (bottom rows), the mouse generally does not lick the other (correct) side afterward. They seem to know that correct licks after an incorrect lick do not yield reward.

    3. The maximum licking rates were not significantly affected by laser onset.

    4. The latency of the first lick (reaction time) was not significantly affected by laser onset. (Please also see our response to question 2b).

    1. Data relating to misses should be included in analyses to provide a complete picture of behaviour and neural responses

    a) In the optogenetic manipulations, an increase in misses seems to dominate the decreased accuracy (please, explain when a response was counted as a miss). A separate analysis of miss trials may be more robust than of error trials and also offers a different interpretation of the data, namely that the mouse did not see the stimulus rather than perceiving the figure on the opposite side. However, if the mice reduced their lick rate in general during optogenetic stimulation, this begs the question whether their motor performance was affected by optogenetic manipulation. Can this possibility be excluded?

    Trials were counted as follows: A trial was counted as a hit when the first lick after 200 ms after stimulus onset was on the correct side. A trial was counted as an error, when the first lick after 200 ms after stimulus onset was on the incorrect side. A trial was counted as a miss, when the mouse did not lick in the window between 200 and 2000 ms after stimulus onset. We have clarified this in the methods section (line 517-526).

    Our previous text may not have been sufficiently clear but the decrease in accuracy during optogenetic trials is not dominated by an increase in missed trials. As we have now indicated explicitly in its caption, in figure 1, missed trials are excluded from the analysis. Hence, the significant effects shown in figure 1 are not driven by an increase in missed trials but rather by an increase in erroneous licks. When comparing figure 1 vs figure S3, where the missed trials are added to the analysis as if they were error trials, we can see an overall downward shift of the performances. Indeed, mice miss more trials when the laser is on. The increase in number of missed trials is lower than the increase in number of wrong choices. Furthermore, the range between the performances at early laser onset and late laser onset is still very similar. This indicates that the mice on average do not have higher miss rates when laser onset is early.

    Finally, nor maximum licking rate, nor the reaction time is affected by the laser onset (see the new figure S2)

    Related to Fig 4, it would be equally interesting to see how FGM changes during misses. Do the changes support the observations for error trials?

    We are not convinced that the neural data from missed trials can be interpreted in a simple way. Mice may have various reasons to miss a trial: they may be tired or not paying attention, they may not have seen the stimulus well, they may not feel thirsty enough, they might be distracted by some sensory input that humans might not be aware of, etc. This is why we specifically opted to not use a go-no/go task but instead opted to use a 2-alternative forced choice task.

    1. Statistical tests do not support the conclusions, are missing or inadequate

    a) In Fig 1E, accuracy is significantly affected at only 1-2 time points in each task, specifically either the 1st and 3rd or the 2nd time point. How do the authors interpret these results? If inhibition starting at the 2nd time point has no significant effects, why would it be significant when inhibition starts later (at the 3rd time)? Furthermore, given that all other starting points of laser stimulation have no significant effects, there is no reason to trust the latency of inhibition effects based on mostly insignificant data points. This analysis in its current form should be removed, including a comparison of latencies between tasks, which was not tested for significance. It may be more meaningful to analyse accuracy for each animal separately. This may reduce variability.

    We can understand that the reviewer may have concerns regarding the post-hoc analysis of Fig 1E, but we feel these concerns stem from a misinterpretation of our goal with this analysis. In Figure 1E, we use a 1-way repeated-measures ANOVA. By using this test, we ask whether the performance of the animals is affected by the laser onset. More specifically “does the performance increase or decrease with increasing laser onset?” The test is significant, so indeed the performance goes up as laser onset goes up. This indicates that the performance of the mice is affected by the inhibition of sSC. For the sake of completeness we had included the post-hoc tests for each latency in the statistics table. Indeed, some individual latencies are not significantly different to the no-laser condition. However, this does not invalidate the conclusion of the main test: a repeated measures ANOVA can only be performed on data with 3 or more groups, so the conclusion of the repeated measures ANOVA could not have been drawn from simply those laser onset(s) that is/are significantly different from the no-laser condition. The main effect of higher performance with higher latencies is significant, even if some individual comparisons are non-significant. The difference in significance of the post-hoc tests does not indicate a significant difference between the groups, but insufficient power to do six individual tests.

    We have changed the wording in the reporting of the statistics of Figure 1E to hopefully more precisely indicate the conclusions we drew from the statistics. We do not draw conclusions from the post hoc tests. We have considered removing them from the statistics table 1, but believe that some readers might be interested. We can remove them if the reviewer believes that would be better.

    b) Analyses regarding the difference in neural response to figure and ground (Fig 2I-J, Fig 3B, Fig 4B, Fig 5C) would be more convincing and informative if the differences were analysed on the level of single neurons in response to the same orientation within their RF (or at the location where the figure is presented, for edge-RF neurons). A histogram of these differences would show how many neurons are affected and how large the effect is in single neurons.

    We fully appreciate this idea, but the way we set up the behavioural task does not quite allow for this type of statistical analysis. This is because we tested all three of the tasks during single sessions (contrast/orientation/phase), and on top of that, we varied the orientations of the stimuli (0/90deg), as well as the phase of the gratings (60 different phases). This all was done with the idea that it would prevent the mice from memorizing the individual stimuli of the task. This also had the effect that only very few trials per session contained the exact same stimulus type, figure-ground condition, orientation and phase. For example, if a mouse would perform around 120 trials in a session. 25% of those were contrast-stimulus-trials, 37.5% of those were orientation-stimulus-trials and 37,5% were phase trials. If we look into 120*0.375 = 45 orientation-stimulus-trials, half of those were figure trials, half were ground trials: 22 trials each. If we split these trials up by their individual orientations, we are left with only about 11 trials per condition to analyse for figure-ground effects, each of which would probably have a different grating phase. Given the firing rate variations that the individual neurons show in awake mice, this amount of trials would not provide enough statistical power to test the significance of modulation in single neurons.

    Although we feel the study design would not allow analysis of individual neurons in response to the same orientation within their RF, we did perform an aggregated analysis on orientation selectivity. For this analysis, we included all the trials where the RF of the recorded neurons was on the background-half of the screen. We then computed the responses of each neuron to the trials where the background orientation was 0 and 90, respectively. This analysis showed that most neurons had no preference for either of the two tested orientations of the other. Only 4 out of 64 (6%) neurons showed a significant preference. We therefore believe that splitting the data by orientation preference would not be very informative.

    c) All statistical tests performed across neurons should account for dependencies due to simultaneous recordings (dependency on session) and due to recordings in the same animal (dependency on animal). This can be done in most cases by using linear mixed-effects models.

    We agree with the reviewer and have changed the analysis for figure 2I, 3B and 3E to an LME analysis (see also Table 1).

    d) There was no significant difference between model weights (Fig 3D), so the statement in line 210 (RF-edge neurons had higher weights) should be removed.

    In answer to previous we question changed the analysis for what is now Figure 3E to an LME. This shows that relative weights were significantly higher for the orientation compared to the phase task. We have adapted our conclusion accordingly (line 214-218).

    e) Fig 4B compares FGM during correct and error trials. This comparison has to be performed with the same set of neurons in correct and error trials (not the case for orientation). Again, the most compelling and informative comparison would be on the level of single neurons: response difference between figure and ground (same visual features at figure position) during hits versus errors.

    As described above, we feel the study design does not allow analysis on the level of individual neurons. The analysis in 4B was actually performed using the same set of neurons, we have removed the typo.

    f) There is no evidence that FGM for phase was different between hit and error trials as stated in line 234.

    Indeed, we had phrased this incorrectly. Since we recorded all task during single recording sessions, we have data for each task for most neurons. We were therefore able to pool the results from the different tasks, and the main d-prime difference between hit vs. error was significant. Post-hoc tests showed that this is mainly driven by the difference in the orientation task. We have edited the wording to be more accurate (line 239-242).

    g) It is not clear why and how the mixed linear effects model was used pooling data across tasks (Fig 4C and Fig 5D). Different neurons were recorded for each task, so the sample points (neurons) are not affected by both task effects (orientation and phase). Each task should be analysed separately.

    Since we recorded all three task versions during single behavioral sessions, we have data for multiple tasks from each neuron. This is why the linear mixed effects model pools the data across the tasks. We have added a note in the main text for clarity (line 238-242)

    h) Bonferroni correction in Fig 1E should correct multiple comparisons across time points, not across tasks (see Table 1).

    The multiple time points all belong to the same one-way repeated measures ANOVA, so there’s no need to correct the post-hoc analysis. We did run the ANOVA for three tasks, which is why we corrected the p-values of each task. We think that this is best way, but can also present uncorrected p-values if needed.

    i) What is the reason to perform some tests one-tailed, others two-tailed?

    Following the reviewer comments, we changed some analyses to LME models. The remaining tests that require definition of the tails are all two-tailed.

    1. The results relating to "multisensory neurons" are ambiguous regarding their interpretation (if significant at all) and seem unrelated to the goal of the study. It is particularly likely that behaviours like licking or other movements cause the response differences between figure and ground.

    We agree with the reviewer that finding these neurons was not the aim of the study. We did not include enough type of tests in our paradigm to fully determine the properties of these neurons. Furthermore, we note that we have recorded too few of these neurons to draw strong conclusions. The data shown in new Figure 2—figure supplement 1H suggest that the responses of these neurons or not as strongly time-locked to the first lick as they are to the trial onset. We presented the behavior of these neurons in our manuscript, because, whatever their exact behavior, they are clearly distinct from the visually responsive cells that show a short latency response to the visual stimulus (Figure 2—figure supplement 1). We still feel that it is useful for the reader to know there are cells in the sSC that show such a distinct behavior, but we have moved the figure and the accompanying text to a figure supplement to avoid distraction from the main message of the manuscript.

    1. What depth were neurons recorded from (Fig 3 and 4)?

    The depths of the recorded visually responsive neurons is now shown in Figure 2—figure supplement 1E.

    Reviewer #3 (Public Review):

    The authors used optogenetic manipulations and electrophysiology recordings to study a causal role and the coding of superficial part of the mouse Superior Colliculus (SCs) during figure detection tasks.

    Authors previously reported that figure-ground perception relies on V1 activity (Kirchberger et al. 2021) and pointed out that silencing of V1 reduced the accuracy of the mice but still the performance was above the chance level. Therefore, visual information necessary in this task, could be processed via alternative pathways. In this study, authors investigated specifically SCs and used similar approach and analysis as in Kirchberger et al. 2021. Optogenetic silencing of the activity of visual neurons in SCs impaired the accuracy in all 3 versions of the figure detection task: contrast, orientation, and phase. Electrophysiology recordings revealed that SCs neurons are figure-ground modulated, but only by contrast- and orientation-based figures. They show SCs visually responsive neurons reflect behavioral performance in orientation-based figure task. The authors conclusion is that SCs is involved in figure detection task.

    Overall, this study provides evidence that mouse SCs is involved in a figure detection task, and codes for task-related events. Authors heroically compared results between 3 different versions of the figure-based detection task. The logic of the study flows through the manuscript and authors prepared a detailed description of methods.

    Thank you for your positive comments.

    However, my main concern is with 1) the amount of data used to make the key arguments, and 2) the interpretation of results. The key findings of this study (figure-ground modulations in SCs) could be a result of the visual cortical feedback in SCs during the task, or pupil diameter changes. Unfortunately, the authors did not rule out these possibilities.

    Still, this study can be relevant to a general neuroscience audience, and results could be more convincing if the authors could clarify:

    1. Optogenetic inactivation

    a) The impact of laser stimulation on neural activity is not satisfactory (Supplementary Figure 1). The method seems to be insufficient to fully salience neurons. Electrophysiology control recordings of inactivation are performed in anesthetized mice, which is not a fair estimation of the effect in awake state. Therefore, it rises a major question how effective the inactivation is during the task?

    We have conducted new control experiments for the impact of laser stimulation on neural activity, now in awake animals (see Figure 1—figure supplement 2). The reviewer was right to ask for these experiments. We had not expected much difference in the effect of silencing in the awake and anesthetized state. To minimize the animal discomfort, we had therefore done these control experiments in terminal experiments under anesthesia. However, these new set of experiments showed that the impact of laser stimulation was much stronger in awake mice than anesthetized mice. We see an average spike rate reduction of 90% when the laser is on. Although it is not full silencing, we think this reduction is sufficient to draw some conclusions on the role of sSC in the behavioral tasks.

    b) Could authors provide more details if laser stimulation has an effect only on visual, or all sampled units? How many of units were recorded, and how many show positive and negative laser modulation?

    We defined visually responsive units as units that have an evoked rate of at least 2 spikes/s. In the new figure 1—figure supplement 2D from the new set of control experiments, we plotted, for every unit, the mean rate in laser ON and OFF trials - also including the non-visually responsive units. It is evident that the spiking activity of most units – including those that were not classified as ‘visual’ – is reduced in the laser ON compared to OFF trials. We observed 1 unit that showed strong positive laser modulation over the entire duration (figure 1—figure supplement 1D). Many units were activated by shorter laser pulses directly after laser onset (Figure 1—figure supplement 2A-B), but these also reduced in activity as the stimulation continued.

    c) How local the inactivation effect is? Where was the silicon probe placed in relation to AAV expression and optical fiber position?

    The AAV was injected at 0.3 mm anterior and 0.5 mm lateral to the lambda cranial landmark. With this injection location we aimed to focus the expression at low/nasal receptive fields, in front of the mouse, because that is where the visual stimulation would take place. From there, the expression did spread laterally across sSC (see Figure 1C). The silicon probe was placed roughly in the same location as the viral injection. The optical fiber was positioned such that the tip would shine on the surface of the sSC at a slight angle, from a lateral distance of ~200 µm from the silicon probe. We have edited the methods section to make this more clear (line 583-585). This procedure allowed us to record only relatively local effects of the inactivation. Although we did not record neural activity across the entirety of sSC, we did record from multiple electrode penetrations per mouse, each time slightly varying the recording location with up to ~300µm and ~500µm in the anterior and lateral directions, respectively. In these variations of recording location the optogenetic effect was always present (see new Figure 1—figure supplement 2G). Moreover, the suppressive effect of optogenetic stimulation of GAD2+ neurons was observed across the entire depth of the sSC (new Figure 1—figure supplement 2H).

    1. Number of sessions and units

    a) The inactivation effect on behavior (Figure 1E) during phase-task has a significantly larger effect at 66ms after stimulus onset. How can authors explain this? Could this result be biased by one animal/session, or low number of trials for this condition? There is no information about number of trials, or sessions from individual animals. Adding a single example of animal's performance, and sessions for individual mice could clarify results in Figure 1.

    The criterium for each mouse to be included in the analysis for one of the tasks was to have 100 trials where optogenetics were used (aggregated across the latencies). So at minimum, we would have about 100 trials/6 latencies = 17 trials per latency per mouse. For most mice though, the number of trials per latency was closer to about 40. We have added more information about this to the methods section (lines 567-570). Despite these inclusion criteria, the 66 ms effect is present for multiple mice (we have now added data visualizations for the individual mice in Figure 1—figure supplement 4). To address the reviewer’s concerns, we can only speculate as to why this happens. It might be random variation. A more speculative conclusion would be that perhaps this 66ms laser onset is particularly disturbing to the visual processing and/or decision-making of the mouse. But we feel that we do not have enough evidence to conclude this.

    b) Figure 2H shows an example of neuron with an effect in the figure detection task based on phase difference, but Figure 2I/J (population response) shows there is no effect. Overall, the conclusion is that SCs neurons are not modulated by a phase-defined object. It seems that number of mice and hence units are smaller in phase-detection task comparing to two other tasks. How many of single units are modulated in each version of the task? How big is the FGM effect on single neuron response (could authors provide values in spikes/s)? One task is dropped from analysis which it is one of the main points of the paper: to compare responses across different versions of the figure detection task in SCs. But Figures 3-5 only focuses on two tasks, because there is not enough of data for figure-based contrast task.

    We have updated Figure 2H to show spikes/s of the example single neuron response. For the population responses, we explicitly normalized the individual neurons because they all have different baseline and peak firing rates. This normalization was important for the decoding, so we decided to print the data such that the data from Figures 2I and 3B went into the decoding as printed. If we look at the non-normalized values, the maximum amplitude of the average FGM effect is 22.3, 5.9 and 2.9 sp/s respectively for the three tasks (for neurons with RF on stimulus center).

    We have furthermore updated the FGM analysis such that the clustered statistic is now based on linear mixed effects statistics instead of T-test statistics. The results based on this new analysis are largely the same (see statistics table T1). We checked the significance of individual neurons in the time window where the grouped LME analysis was significant. For the phase task (n.s. in grouped analysis), we used the significant window from the orientation task. For this analysis, we want to stress that the number of trials for each version of the task for each individual neurons is quite limited as we recorded all three of the tasks during each recording session. Individually, 7/23 neurons were significant for the contrast task, 1/49 were significant for the orientation task, 0/32 were significant for the phase task (after Bonferroni-holm correction).

    To address the final part of this comment on dropping the contrast task: we indeed have recorded too few data points to draw conclusions on decoding (Fig. 3) and discriminability (Fig. 4) for the contrast task. However, we do not see the contrast detection task as the main point of the paper. As earlier work had already shown involvement of the sSC in visually-evoked behaviours based on objects that are clearly isolated from the background, the main focus in this work is to show involvement of sSC in complex object detection, where the visual contrast and luminance is the same across object and background.

    1. Figure-ground modulation in SCs

    a) How is neural activity correlated with pupil size, movement (eg. whisking, or face), or jaw movement (preparation to lick)? Can activity of FGM neurons in SCs be explained by these behavioral variables?

    We did not record whisking or other face and jaw movements. We did record the eye of the mice, so have included a new Figure 2—figure supplement 2 which shows eye position and pupil dilation during the task. For the analysis in the originally submitted paper, trials with substantial eye movement (Z-score of eye speed > 2.5) between 0 and 450 ms had already been removed from the analysis. This way, we could exclude effects of eye movements (but not pupil dilation) on the visual responses in sSC. The additional figures and analyses have been done using the same inclusion criteria. Indeed, in the included trials mice did not move their eyes during the peak of the visual response (0-250 ms). The pupil dilation also did not change in this period.

    b) Could authors describe in more detail how they measure a pupil position and diameter, by showing raw data, pupil size aligned to task events?

    We have added a new Figure 2—figure supplement 2 to show the pupil position and diameter aligned to task onset.

    c) How does pupil diameter change between tasks? Small pupil changes can affect responses of visual neurons, and this could be an explanation of FGM effect in SCs. Can authors rule out this possibility, by for example showing pupil size and changes in position at stimulus onset in different tasks?

    Our new Figure 2—figure supplement 2B shows that pupil dilation changes and differences in pupil dilation between figure/ground trials do occur, but only after ~300 ms, so after the peak of the visual response and after the FGM is present in sSC.

    d) Authors in discussion mentioned that the modulation of V1 could be transferred to SCs through the direct projection. Moreover, animals perform above chance in both inactivation experiments (V1 and SC), which could be also an effect of geniculate projections to HVAs (eg. Sincich et al. 2004). Could authors discuss different possibilities?

    The direct geniculate projection to HVAs is an interesting possibility that we had not considered yet. The dLGN in the mouse projects (apart from V1) mostly to the medial HVAs (Bienkowski et al. 2018). The lateral extrastriate regions receive only very sparse input from the dLGN. The medial HVAs, however, could be silenced without drop in performance in a simple visual detection task (Goldback et al., 2020). Therefore, it does not seem likely that this geniculate to HVAs projections would be important in the figure detection task.

    1. Interpretation of multisensory neurons is not clear. In Figure 5B, there is an example of neuron with two peaks of response. Authors speculate about the activity (pre-motor) but there is lack of clear measurement showing "multisensory" response of these neurons. Could these responses be related to the movement of the lick spout towards the mouth of the mouse (500 ms after the presentation of the stimulus)? Moreover, the number of "multisensory" units is very low (5 units, and 8 units).

    We have not done definitive test to show what these putative multisensory neurons exactly respond to. Because of their response was after the appearance of the lick and time locking to the trial start, rather than to the licking response, we think that is likely that these neurons responded to the appearance of the spout. There might have been visual, auditory, vibrational or touch clues to which these neurons respond. We believe it is interesting for the reader to know that there is class of neurons in the sSC that did not show a visual stimulus but was time locked to the trial. This was the reason that we had included this figure in the manuscript. However, given the reviewers comments we have decided to move the figure and accompanying text to a figure supplement (Figure 2—figure supplement 1) in order to not distract from the main message of the manuscript.

  3. eLife

    eLife assessment

    The authors present a valuable work suggesting that the superficial, retinorecipient layers of the mouse superior colliculus (SC) may participate in figure-ground segregation and object recognition. These data are based largely on optogenetic perturbations of SC but the strength of evidence is currently incomplete: although the effects are statistically significant, there are significant technical limitations that are not adequately addressed via controls.

  4. eLife

    Reviewer #1 (Public Review):

    The authors aimed to understand whether the superficial, retinorecipient layers of the mouse superior colliculus (sSC) participate in figure-ground segregation and object recognition. To address this question, they use a combination of optogenetic perturbations of sSC and recordings. These data are consistent with SC being causally involved in object recognition. This would be useful information for the field and likely to be cited. However, I have several concerns regarding their conclusions.

    A significant limitation of this study is methodological. The major novelty is the effect of optogenetic silencing, because the recordings are largely correlative, but the optogenetic silencing approach lacks appropriate controls for the effects of the optogenetic excitation light. The authors acknowledge that the optogenetic light is a potential confound, but attempt to address this by shielding the fiber to eliminate light leak and strobing a blue led in the arena. The former does not account for the effects of excitation light scattering intracerebrally--during optogenetic experiments, intracerebral scattering causes the eyes to light up--and for the latter, there is no way to compare the intensity or qualia of the externally strobed LED and the intracerebral light. The proper control would be a cohort of mice lacking channelrhodopsin expression in sSC. Regardless, it is essential to acknowledge this potential confound.

    Relatedly, as the authors note, there are GABAergic projection neurons in sSC that may be driving these effects via gain of function. This is a significant concern that has limited the widespread adoption of this approach in sSC despite its popularity in studies in cortex. Indeed, one recently published study of behavioral functions of deep SC found that activating inhibitory neurons actually caused paradoxical behavioral effects consistent with gain of function in the targeted hemisphere, due to the effects of long-range inhibitory projections on the other SC hemisphere. Given the presence of inhibitory projections in sSC, it would be preferable to use an orthogonal method for silencing and at least to thoroughly acknowledge these concerns and cite these recent studies.

    A minor point is that although activation of GABAergic neurons in sSC is expected to cause inhibition of neighboring neurons, I would expect channelrhodopsin-expressing GABAergic cells to show an increase in firing during optogenetic excitation. However, it seems that none of the cells plotted (assuming each point in Supplementary Fig 4D is a cell, which the legend does not specify) had such an increase. Do these extracellular recordings not detect inhibitory neurons well?

    Finally, the relationship between these stimuli and objects is not entirely clear. The authors acknowledge this but it would be worthwhile to devote more attention to this point. In effect, as the authors note, the gray screen and sinuisoidal grating do not have any sharp edges on the screen, whereas each of the behaviorally relevant stimuli will create a sharp, step-like edge on the screen. Whether edge detection is truly object detection or simply a variant of more general visual detection is unclear.

  5. eLife

    Reviewer #2 (Public Review):

    The goal of this study is to show that the superficial superior colliculus (sSC) of mouse signals figure-ground differences defined by contrast, orientation, and phase, and that these signals are necessary for the animal to detect such figure-ground differences. By inhibiting sSC while the animals perform a figure-ground detection task, the study shows that detection performance decreases when sSC activity is suppressed during the onset of the visual stimulus. The study then intends to show that sSC neurons exhibit surround suppression based on orientation differences, and that surround suppression is stronger when the animal detects the correct location of the figure on the background.

    The major strength of this study is the use of a behavioural paradigm to test detection performance of figure-ground stimuli while manipulating neural activity in the sSC during different times after stimulus onset. This paradigm would show whether activity in the sSC is relevant for performing the task. Secondly, the study collected data to confirm previous findings: sSC neurons exhibit orientation specific surround suppression. Additionally, it is impressive that the authors were able to train mice to generalize their task performance across different stimulus categories (figure-ground differences in orientation and phase). This should be highlighted as it may inform future studies.

    The study has, however, methodological and analytical weaknesses so that the stated conclusions are not supported by the presented results.

    1. Optogenetic inhibition is not limited to sSC (even expression may not be limited)
      About 30% of inhibitory neurons in the sSC project to other areas, e.g. ventral LGN, parabigeminal nucleus and pretectum (Whyland et al, 2019, see ref in manuscript). This means that these areas receive direct inhibition when inhibitory sSC neurons are optogenetically stimulated. This fact is mentioned in the discussion but the consequences and implications for the results are ignored. This is a major flaw of the optogenetic experiments of this study. Additionally, no evidence is given that opsin expression was limited to the superficial layers (except for one histological slice), which the authors acknowledge in line 285. Deeper layers may have other inhibitory neurons with long-range projections.
      The finding that sSC neurons show no figure-ground modulation for phase while the optogenetic manipulation has behavioural effects may be an indication for other areas being affected by the optogenetic manipulation.

    2. Could other behavioural variables explain the results?
      a) Are there any task events other than the visual stimuli that the mice could use to make their decisions? The authors state the use of a custom made lick spout but it is not clear how this spout works, i.e. how do mechanics of the spout deliver water to the right versus the left output and could the mouse perceive these mechanics?
      b) Could the different neural responses to figure versus ground shown in Fig 2I-J and Fig 3B be explained by behaviours varying between the trial types, e.g. by early lick movements (which are conceivable even if the spout is not present), eye movements or changes in pupil-linked arousal? A behavioural difference seems even more likely to occur between hit and error/miss trials (Fig 4). If these behaviours were not measured, the possibility of behavioural modulation should be discussed.

    3. What is the behavioural strategy of the animals?
      Only licks beyond 200 ms after stimulus onset determine the choice of the animal because "mice made early random licks" from 0 to 200 ms. To better understand the behavioural strategies of the animals we need to see their behavioural data, i.e. left and right licks aligned to stimulus onset. It would be particularly interesting to see how number and latency of licks changes during optogenetic manipulation.

    4. Data relating to misses should be included in analyses to provide a complete picture of behaviour and neural responses
      a) In the optogenetic manipulations, an increase in misses seems to dominate the decreased accuracy (please, explain when a response was counted as a miss). A separate analysis of miss trials may be more robust than of error trials and also offers a different interpretation of the data, namely that the mouse did not see the stimulus rather than perceiving the figure on the opposite side. However, if the mice reduced their lick rate in general during optogenetic stimulation, this begs the question whether their motor performance was affected by optogenetic manipulation. Can this possibility be excluded?
      b) Related to Fig 4, it would be equally interesting to see how FGM changes during misses. Do the changes support the observations for error trials?

    5. Statistical tests do not support the conclusions, are missing or inadequate
      a) In Fig 1E, accuracy is significantly affected at only 1-2 time points in each task, specifically either the 1st and 3rd or the 2nd time point. How do the authors interpret these results? If inhibition starting at the 2nd time point has no significant effects, why would it be significant when inhibition starts later (at the 3rd time)? Furthermore, given that all other starting points of laser stimulation have no significant effects, there is no reason to trust the latency of inhibition effects based on mostly insignificant data points. This analysis in its current form should be removed, including a comparison of latencies between tasks, which was not tested for significance. It may be more meaningful to analyse accuracy for each animal separately. This may reduce variability.
      b) Analyses regarding the difference in neural response to figure and ground (Fig 2I-J, Fig 3B, Fig 4B, Fig 5C) would be more convincing and informative if the differences were analysed on the level of single neurons in response to the same orientation within their RF (or at the location where the figure is presented, for edge-RF neurons). A histogram of these differences would show how many neurons are affected and how large the effect is in single neurons.
      c) All statistical tests performed across neurons should account for dependencies due to simultaneous recordings (dependency on session) and due to recordings in the same animal (dependency on animal). This can be done in most cases by using linear mixed-effects models.
      d) There was no significant difference between model weights (Fig 3D), so the statement in line 210 (RF-edge neurons had higher weights) should be removed.
      e) Fig 4B compares FGM during correct and error trials. This comparison has to be performed with the same set of neurons in correct and error trials (not the case for orientation). Again, the most compelling and informative comparison would be on the level of single neurons: response difference between figure and ground (same visual features at figure position) during hits versus errors.
      f) There is no evidence that FGM for phase was different between hit and error trials as stated in line 234.
      g) It is not clear why and how the mixed linear effects model was used pooling data across tasks (Fig 4C and Fig 5D). Different neurons were recorded for each task, so the sample points (neurons) are not affected by both task effects (orientation and phase). Each task should be analysed separately.
      h) Bonferroni correction in Fig 1E should correct multiple comparisons across time points, not across tasks (see Table 1).
      i) What is the reason to perform some tests one-tailed, others two-tailed?

    6. The results relating to "multisensory neurons" are ambiguous regarding their interpretation (if significant at all) and seem unrelated to the goal of the study. It is particularly likely that behaviours like licking or other movements cause the response differences between figure and ground.

    7. What depth were neurons recorded from (Fig 3 and 4)?

  6. eLife

    Reviewer #3 (Public Review):

    The authors used optogenetic manipulations and electrophysiology recordings to study a causal role and the coding of superficial part of the mouse Superior Colliculus (SCs) during figure detection tasks. Authors previously reported that figure-ground perception relies on V1 activity (Kirchberger et al. 2021) and pointed out that silencing of V1 reduced the accuracy of the mice but still the performance was above the chance level. Therefore, visual information necessary in this task, could be processed via alternative pathways. In this study, authors investigated specifically SCs and used similar approach and analysis as in Kirchberger et al. 2021. Optogenetic silencing of the activity of visual neurons in SCs impaired the accuracy in all 3 versions of the figure detection task: contrast, orientation, and phase. Electrophysiology recordings revealed that SCs neurons are figure-ground modulated, but only by contrast- and orientation-based figures. They show SCs visually responsive neurons reflect behavioral performance in orientation-based figure task. The authors conclusion is that SCs is involved in figure detection task.

    Overall, this study provides evidence that mouse SCs is involved in a figure detection task, and codes for task-related events. Authors heroically compared results between 3 different versions of the figure-based detection task. The logic of the study flows through the manuscript and authors prepared a detailed description of methods. However, my main concern is with 1) the amount of data used to make the key arguments, and 2) the interpretation of results. The key findings of this study (figure-ground modulations in SCs) could be a result of the visual cortical feedback in SCs during the task, or pupil diameter changes. Unfortunately, the authors did not rule out these possibilities.

    Still, this study can be relevant to a general neuroscience audience, and results could be more convincing if the authors could clarify:

    1. Optogenetic inactivation
      - The impact of laser stimulation on neural activity is not satisfactory (Supplementary Figure 1). The method seems to be insufficient to fully salience neurons. Electrophysiology control recordings of inactivation are performed in anesthetized mice, which is not a fair estimation of the effect in awake state. Therefore, it rises a major question how effective the inactivation is during the task?
      - Could authors provide more details if laser stimulation has an effect only on visual, or all sampled units? How many of units were recorded, and how many show positive and negative laser modulation? How local the inactivation effect is? Where was the silicon probe placed in relation to AAV expression and optical fiber position?

    2. Number of sessions and units
      - The inactivation effect on behavior (Figure 1E) during phase-task has a significantly larger effect at 66ms after stimulus onset. How can authors explain this? Could this result be biased by one animal/session, or low number of trials for this condition? There is no information about number of trials, or sessions from individual animals. Adding a single example of animal's performance, and sessions for individual mice could clarify results in Figure 1.

    - Figure 2H shows an example of neuron with an effect in the figure detection task based on phase difference, but Figure 2I/J (population response) shows there is no effect. Overall, the conclusion is that SCs neurons are not modulated by a phase-defined object. It seems that number of mice and hence units are smaller in phase-detection task comparing to two other tasks. How many of single units are modulated in each version of the task? How big is the FGM effect on single neuron response (could authors provide values in spikes/s)?

    - One task is dropped from analysis which it is one of the main points of the paper: to compare responses across different versions of the figure detection task in SCs. But Figures 3-5 only focuses on two tasks, because there is not enough of data for figure-based contrast task.

    1. Figure-ground modulation in SCs
      - How is neural activity correlated with pupil size, movement (eg. whisking, or face), or jaw movement (preparation to lick)? Can activity of FGM neurons in SCs be explained by these behavioral variables?
      - Could authors describe in more detail how they measure a pupil position and diameter, by showing raw data, pupil size aligned to task events?
      - How does pupil diameter change between tasks? Small pupil changes can affect responses of visual neurons, and this could be an explanation of FGM effect in SCs. Can authors rule out this possibility, by for example showing pupil size and changes in position at stimulus onset in different tasks?
      - Authors in discussion mentioned that the modulation of V1 could be transferred to SCs through the direct projection. Moreover, animals perform above chance in both inactivation experiments (V1 and SC), which could be also an effect of geniculate projections to HVAs (eg. Sincich et al. 2004). Could authors discuss different possibilities?

    2. Interpretation of multisensory neurons is not clear. In Figure 5B, there is an example of neuron with two peaks of response. Authors speculate about the activity (pre-motor) but there is lack of clear measurement showing "multisensory" response of these neurons. Could these responses be related to the movement of the lick spout towards the mouth of the mouse (500 ms after the presentation of the stimulus)? Moreover, the number of "multisensory" units is very low (5 units, and 8 units).

  7. Version published to 10.1101/2022.09.25.509365v1 on bioRxiv