It is an important task to recognize an object when it is in motion. Humans and nonhuman primates have an excellent ability in detection of motion boundaries (MBs) (Regan, 1986; Regan, 1989). In nonhuman primates, it has been shown that neurons selective for the orientations of MBs are located in the ventral visual pathway (Marcar et al., 2000; Mysore et al., 2006; Chen et al., 2016), but not in the dorsal pathway (Marcar et al., 1995). Particularly, it has been shown that area V2 exhibits significant MB responses. Compared with V1, V2 has more MB orientation neurons (Marcar et al., 2000) and a significant map for MB orientation (Chen et al., 2016). However, other visual areas, including V3 and V4, also have strong responses to MB (Zeki et al., 2003; Mysore et al., 2006). It is unclear whether all these areas contribute to the eventual perception of MBs. This question is particularly important for V2 since it is the lowest area in the visual processing hierarchy that possesses MB sensitivity.

V2 is the largest extrastriate visual area in primates. It receives inputs from V1 and contains a full retinotopic map (Gattass et al., 1981). V2 has different CO stripes (thick, pale, thin), in which neurons exhibit different functional properties and project to different downstream areas (Shipp and Zeki, 1985; DeYoe and Van Essen, 1985; Munk et al., 1995). V2 performs higher-level analysis on visual information in multiple dimensions, for example, shape, color, binocular disparity, and motion. Much of the analysis contributes to figure-ground segregation (e.g., von der Heydt et al., 1984; Zhou et al., 2000).

In the classical view, visual motion information is processed in the dorsal visual pathway. However, mounting evidences have shown that many other brain areas participate in visual motion processing (Orban et al., 2003). In macaque monkeys, there are a large number of direction-selective (DS) neurons in V2. Nevertheless, their functional contributions to visual perception remain unclear. These DS neurons cluster and form functional maps in V2 (Lu et al., 2010; An et al., 2012). Cooling studies have found that these neurons do not contribute to the motion detection in the dorsal pathway (Ponce et al., 2008; Ponce et al., 2011). Compared with middle temporal (MT) neurons, V2 DS neurons have smaller receptive fields (RFs), stronger surround modulation, and higher sensitivity to motion contrast (Hu et al., 2018). Thus, these neurons are suitable for figure-ground segregation and/or MB detection (Chen et al., 2016; Hu et al., 2018). These findings suggest that V2 may calculate the MB orientation using a local DS-to-MB circuit. Testing this hypothesis will also help us to understand the motion processing in V2 and other extra-dorsal areas.

In this study, we trained two monkeys to perform an orientation-discrimination task. From electrode arrays implanted over areas V1 and V2, we recorded neural activity and examined (1) their selectivity to the orientation of MB, (2) their correlation with monkeys’ behavioral choice, and (3) neuronal couplings between DS neurons and MB neurons. We found that many neurons in V2 exhibited a robust orientation selectivity to MBs. The responses of V2 neurons to MBs also had a clear relationship with the animals’ behavioral choice. As a comparison, these features were much weaker or absent in area V1. Finally, cross-correlogram and timing analysis also showed a potential functional link between DS and MB neurons.

Two macaque monkeys (monkey S and monkey W) were trained to do orientation-discrimination tasks. In each of their four hemispheres, a surgery was performed, during which intrinsic signal optical imaging was performed and a 32-channel array was implanted. The array was placed to cover as much V2 direction-preferred domains as possible (Figure 1A–D). The depths of the electrode tips were ~600 μm. For each array, recordings were performed daily and lasted for 1.5–2 months. In 128 channels of the four arrays, 96 channels were located in area V2 and 32 were in V1. Recordings were performed on multiple days. We used three alternative methods in cell selection (with different levels of strictness) and constructed three cell datasets: (1) ‘all cell’ dataset (n = 723). For daily recordings, a spike sorting was first performed for each channel, and signal-to-noise ratios were calculated for the resulting clusters (see Materials and methods). The best signal-to-noise ratio cluster for a channel was selected if its signal-to-noise ratio was larger than 3.5. With this method, we obtained 723 units from 85 V2 channels. (2) ‘Unique-unit’ dataset (n = 287). Based on the ‘all cell’ dataset, we excluded potential duplicated units (i.e., had similar waveforms and tunings) recorded from the same electrodes on different days so that the remaining units were ‘unique’ ones. This means that the neurons in this dataset were either from different electrodes or from the same electrode but had different waveform or tunings. (3) ‘One-cell-per-channel’ dataset (n = 85). In this dataset, only one neuron was chosen for each channel based on the ‘all cell’ dataset. Detailed cell identification procedures can be found in Materials and methods.

Figure 1 with 2 supplements see all

Download asset Open asset

In the above cell identification, the majority of the neurons were single units (SUs, see Materials and methods). The percentages of multiunits (MUs) were 29.3%, 12.5%, and 27% in ‘all cell’, ‘unique-unit’, and ‘one-cell-per-channel’ datasets, respectively. We also performed data analysis with only SUs in each dataset and found that the inclusion of MUs does not change the main conclusions (not shown).

We compared the results obtained from these three datasets and found that the main results were the same. In order to exhibit more details of the data, the results presented below are based on the ‘all cell’ dataset. The results from the other two datasets are presented in Figure 5—figure supplements 2 and 3 for comparison.

In daily recordings, we first performed basic RF tests when the monkey was doing a simple fixation, including cell RF mapping, orientation tests with sine-wave gratings, motion-direction tests with random dots (RDs), and MB orientation tests. The stimuli were 4° circular patches, covering the whole RF region of the array without being optimized for particular cells.

V2 neurons selective to MB orientation

As shown in Figure 1F and Video 1, MB stimuli were composed with moving RDs. In the circular stimulus, RDs moved in opposite directions in the two sides of the virtual midline. The dots were the same on the two sides except for their moving directions. Thus, the boundary was mostly induced by motion. Similar to previous work (Marcar et al., 2000; Chen et al., 2016), two sets of MB stimuli were used in this study, in which the axes of the dot motion were either 45°-angle with the MB (MB1 stimuli) or 135°-angle with the MB (MB2 stimuli).

Video 1

Download asset

Neurons were considered MB orientation selective (MB neurons) if their response functions were well fitted (R² >0.7), preferred orientations to the two MB stimulus sets were consistent (difference <30°), and both orientation selectivity indices (OSIs) for MB stimuli were larger than 0.5. According to these criteria, 10.9% (70/642) of V2 neurons were MB neurons (Supplementary file 1). This proportion is consistent with previous recordings in anesthetized monkey V2 (10.6%, 13/123, Marcar et al., 2000). Importantly, these MB neurons’ preferred MB orientations were very similar to their preferred orientations for grating stimuli (Figure 1H). Thus, these MB neurons exhibited cue invariance for their orientation selectivity. Figure 1I shows that, when tested with grating stimuli, MB neurons exhibited stronger orientation selectivity than those non-MB neurons (mean OSI: 0.85 ± 0.20 vs. 0.80 ± 0.17, p=0.0013, Wilcoxon test). Their RF sizes, however, do not differ with the non-MB neurons’ (2.38 ± 0.70° vs. 2.48 ± 0.87°, p=0.65, Wilcoxon test). Same for their response time delays (described in ‘Response time courses’).

These results were obtained in awake monkeys performing a fixation task and were very similar to those obtained in anesthetized monkey V2 (Marcar et al., 2000). Both studies showed that V2 has around 10% neurons that exhibit strong and robust orientation selectivity to both luminance and MBs, and their preferred orientation to these two stimuli is the same. These findings indicate that V2 has the capability to detect MBs and integrate this information with other boundary cues.

Correlation between V2 activity and monkey behavior

To study whether V2 neurons contribute to MB perception, monkeys were trained to do an orientation-discrimination task. The task is illustrated in Figure 2A: after a sample MB stimulus was presented for 750 ms, two target dots were presented. The monkeys were required to indicate whether the MB line was tilted to the left or right of vertical by saccading to the left or right target. The difficulty of the task was adjusted by using different levels of motion coherence or dot brightness (in one chamber). After training, monkeys can perform well in MB orientation discrimination and achieved at least 95% correct rate for 100% coherence stimuli. Figure 2B shows the psychometric function (fitted with a Weibull function) for monkey S. The coherence thresholds, as determined at the 75% correct rate from the fitting function, were 36.7% for monkey S (Figure 2B) and 67.2% for monkey W (Figure 2—figure supplement 1A).

Figure 2 with 1 supplement see all

Download asset Open asset

Based on the initial MB tests, we selected a MB cell for the following ‘MB-orientation-discrimination test’. The stimuli were similar to the one used in previous MB tests, and its position was centered at the MB neuron’s RF. Only two MB orientations were used either at the neuron's preferred orientation or orthogonal to that orientation. The stimuli were presented at different levels of difficulties (coherence or brightness). Recordings were made while the monkeys were performing an orientation-discrimination task. The responses of an example neuron are shown in Figure 2C. While the coherence of the moving dots increased, the differences of responses to preferred and null MB orientations also increased. Figure 2D shows the receiver operating characteristic (ROC) curves for four out of seven tested coherence levels for this neuron. Each curve indicates the difference of responses to preferred and null orientations. Figure 2E shows the example neuron’s neurometric function, calculated based on the area sizes under the ROC curves. This curve is also well fitted by a Weibull function, and the threshold for this neuron is 40.1%. In our sampled neurons, this neuron was very sensitive to MB orientation. Its neurometric function and coherence threshold were both very similar to the monkey’s behavior measurements.

Fifteen MB neurons were tested with coherence-level stimuli (Figure 2F). Their neurometric curves were generally shallower than the behavioral ones, and the neuronal thresholds were higher than the behavioral ones (Figure 2G). The average ratio of neuronal and behavioral thresholds was 2.78 (Figure 2G, inset plot). These results indicate that most neurons had a poorer sensitivity to MB orientation compared with the animals’ behavioral performance. However, some most sensitive neurons had thresholds close to the behavioral ones. This indicates that at least some neurons had sufficient information to support the animals’ performance in the MB orientation-discrimination tasks.

We further calculated the choice probability (CP) for these MB neurons (Figure 2H), which reflect the biases of neurons when their responses were grouped according to the monkeys’ choices (Britten et al., 1996). In these neurons, 12 had a CP significantly larger than the chance level of 0.5 (bootstrap test, p<0.05). The average CP, 0.59, was also larger than 0.5 (t-test, p=1.4 × 10⁻⁵). This indicates that during the MB orientation-discrimination tasks MB neurons in V2 showed significant choice-related activity. In 17 neurons recorded in one array, we also tested neural-behavioral relevance with different levels of dot brightness. The overall CP was lower in neurons tested with dot brightness. This might be due to the relative easiness for the brightness task. Nevertheless, in both tasks, the average CPs were larger than 0.5 and portions of individual neurons had CPs values larger than 0.5 (Figure 2—figure supplement 1B–E). The CP distribution for the pooled neurons (n = 32, Supplementary file 2) is shown in Figure 2H inset, which also exhibit the same trend.

Thus, V2 not only contained neurons sensitive to MB orientation, the highest sensitive ones also exhibited orientation-discrimination performances that are close to the animal’s behavioral performances. Their random response fluctuations correlated with the fluctuation animals made in the MB orientation-discrimination tasks. All these results indicate that V2 MB neurons likely contribute to the MB discrimination task.

Comparison of MB responses in V1 and V2

In our sampled V2 population, 63% of the neurons were not tuned for MB orientation (either one or two MB responses could not be well fitted by the fitting function, i.e., R² <0.7), 26% either had a low OSI (<0.5) or their two preferred MB orientations were different (>30°), and the remaining 10.9% neurons were classified as MB neurons (Figure 3A, top). In our array recordings, 32 channels in three arrays were located in area V1 and from which 93 V1 neurons were recorded (Figure 1—figure supplement 2, Figure 3—figure supplement 1). Compared with V2, a higher percentage (82%) of V1 neurons were not tuned to MB orientation; 16% had a low OSI or their two preferred orientation differed; only 2% (two neurons) were classified as MB neurons (Figure 3A, bottom). V1 and V2 were different in their cell distributions in the three groups (chi-squared test, χ²=14.2, p=8.25 × 10⁻⁴). In addition, the mean orientation index (OSI) for MB stimuli was larger for V2 neurons than for V1 neurons (calculated from the neurons whose responses could be well fitted, i.e., R²≥0.7). Thus, V2 exhibited significantly higher MB detection capability than V1 at the population level.

Figure 3 with 2 supplements see all

Download asset Open asset

In Figure 3A, we compared two OSIs for those well-fitted neurons, including both MB and non-MB neurons. V1 had an overall lower OSI value than that in V2 (V1: 0.36 ± 0.17; V2: 0.44 ± 0.19, p=0.033, Wilcoxon test). In both V1 and V2 neurons, their two OSIs were strongly correlated (V1: r = 0.70, p=0.0016; V2: r = 0.72, p=3.9 × 10⁻³⁹). In addition, the OSI values of the MB and non-MB neurons were continuously distributed and did not form two clusters.

We further analyzed CPs for the whole V2 population and compared with those in V1. In this analysis, we also included both MB neurons and non-MB neurons. Most neurons recorded in an array had overlapped RFs (Figure 1—figure supplement 1B). In offline analysis, we identified non-MB neurons that had their RFs and preferred/null orientations matched the MB stimuli tested on that day and calculated their CP values in a similar way as for the MB neurons (Figure 2H). Figure 3B top panel shows the CP distribution for all V2 neurons suitable for CP analysis (n = 140), including 108 non-MB neurons and 32 MB neurons. V2 CPs had a unimodal distribution and shifted toward the right side. Its mean CP (0.53) was higher than 0.5 (t-test, p=1.9 × 10⁻¹²), but lower than the mean CP for MB neurons (0.56, Figure 2H). There were 49 V2 neurons that had a CP either larger or smaller than 0.5 (bootstrap test, p<0.05), among which 15 were MB neurons (all had CPs larger than 0.5). For the neurons recorded in V1, 26 were suitable for CP analysis, including 2 MB neurons (Figure 3C, bottom). The mean CP (0.51) was not different from 0.5 (t-test, p=0.11), and six neurons had a CP larger than 0.5 (bootstrap test, p<0.05). In addition, non-MB V2 neurons (n = 108) had a mean CP (0.53) larger than 0.5 (t-test, p<0.001), but not for non-MB V1 neurons (n = 24, CP = 0.51, p=0.21, t-test) (Figure 3—figure supplement 2B, D).

In summary, V2 not only had stronger orientation tuning to MB stimuli and more MB neurons than V1, their behavioral relevance was also stronger than the V1’s.

Correlations between MB and DS neurons

In our array recordings, we also tested neurons’ direction selectivity with moving RDs. We identified 88 V2 DS neurons that had direction-selective index (DSI) larger than 0.5. In our hypothesis, V2 MB neurons receive their motion information from V2 DS neurons. In order to test this hypothesis, we examined the correlations between the MB and DS neurons for different types of visual stimuli.

The stimuli we used was a 4° square RD patch (Figure 4—figure supplement 1A and Video 2) that was divided into two parts by a virtual line (MB). The RD in one part moved in the DS neuron’s preferred direction, and the RD in another part moved in its opposite direction (null direction). The square patch was rotated so that the MB matched the MB neuron’s preferred/null orientations. We tested seven MB locations. The monkey was performing an orientation-discrimination task. We found that when the MB stimuli was presented cross-correlograms (CCGs) show an elevated correlation at 0 time lag (Figure 4A, data pooled for seven MB locations). Importantly, the correlation was much higher for the preferred MB orientation than the null one. Note that the moving directions of the RD were the same for these two conditions. We also analyzed DS–MB pairs (n = 5) in which DS neurons were not optimally stimulated. The CCGs for preferred MB and null MB do not differ (t-test, p=0.26). All CCGs we used were shuffle-corrected CCGs so that stimulus-related effects were removed.

Figure 4 with 1 supplement see all

Download asset Open asset

Video 2

Download asset

We tested a luminance-line control stimulus (real line) (Figure 4—figure supplement 1A and Video 3), in which a white line replaced the MB, and in both sides the RD moved in two opposite directions (transparent motion). The correlation in CCG was reduced in these conditions and did not differ for the preferred and null orientations.

Video 3

Download asset

We further tested another version of control stimulus called temporal boundary (TB, Chen et al., 2016; Figure 4—figure supplement 1A and Video 4). TB stimuli were similar to the MB except that RDs in the two sides of the TB moved in the same direction. The dots still disappeared and appeared at the virtual boundaries. In examined neurons, none exhibited tuning to the TB orientation. The animals also made random choices to these stimuli in the orientation-discrimination tasks. In CCG, the MB–DS pairs exhibited larger modulation amplitudes to the TB stimuli (note the scale difference), but did not peak at any particular positions (Figure 4C). No differences were found for preferred and null orientations. The cortical distances of paired neurons in Figure 4A–C had a range of 0.93–1.1 mm.

Video 4

Download asset

We did the same analysis on pairs of non-MB and DS neurons. To make fair comparisons, we randomly selected equal numbers of non-MB–DS pairs as the MB–DS pairs (Figure 4A–C, right panel). Among all three stimulus conditions (MB, real line, TB), only under the MB condition that the non-MB–DS pairs exhibited small peaks at time zero, but did not differ between two orientations. To quantify the CCGs, we calculated the area sizes under the CCG curves from −40 ms to 40 ms. As shown in Figure 4D, only the MB–DS pairs under the MB condition showed significant differences between the preferred and null orientations (t-test, p=1.9 × 10⁻⁴).

In summary, results from response correlation analysis support the hypothesis that functional connections between V2 DS and MB neurons do exist. Such connections operate in specific stimulus conditions, but peak location (at time zero) is inconsistent with a simple DS to MB contribution model. These phenomena can also be due to bidirectional interactions between DS and MB neurons or their common inputs.

Response time courses

To further investigate the mechanisms underlying V2 MB orientation selectivity, we measured the time courses of V2 MB neurons when they responded to the MB and real-line stimuli. Responses PSTHs (Figure 5A, B) were calculated from the data used in the correlation analysis above. In seven MB positions (Figure 4—figure supplement 1A), only the conditions that the MB was in the RF center (0°) were used in this analysis. With an analysis method similar to the one used in Chen et al., 2017, we calculated a visual response latency (i.e., the time a neuron started to respond to the visual stimulus) and an orientation-selectivity latency (i.e., the time responses started to differ for preferred and null orientations). The orientation-selectivity latency for the MB stimuli (85 ms, Figure 5A) was 50 ms later than that for the real-line stimuli (35 ms, Figure 5B), while the visual response latencies for these two stimuli were similar (both were 37 ms, Figure 5A, B). This indicates that, compared with the luminance stimuli, additional time and circuits are required to calculate orientation from the MB stimuli. In addition, MB neurons had similar visual response latencies as those in non-MB neurons when they were responding to the MB stimuli (Figure 5—figure supplement 1A, B). This is different from the previous findings obtained in anesthetized monkeys (Marcar et al., 2000), where MB neurons responded slower to the MB stimuli than that of the non-MB neurons.

Figure 5 with 4 supplements see all

Download asset Open asset

In order to see whether the response time of DS neurons falls into a reasonable range so that their contribution to MB neurons is valid, we analyzed the response time courses of V2 DS neurons (n = 9). Considering that V2 DS neurons have strong surround suppression (Hu et al., 2018), we tested three types of RD stimuli to stimulate their RF centers and surrounds: (1) a center-only RD patch in which RDs moved in the neurons’ preferred directions; (2) a large RD patch covered both RF center and surround in which RDs moved in the neurons’ preferred directions; and (3) a center patch moved in the preferred directions and a surround patch (annulus) moved in the opposite (null) directions. Latencies were calculated with a population latency method (see Materials and methods). The visual response latency for DS neurons was 41 ms, similar for all three stimuli. The surround suppression time started at 51 ms, when responses to stimuli 1 and 2 started to differ. The surround modulation time started at 109 ms, when responses to stimuli 2 and 3 started to differ. These time points, together with those from MB neurons, are summarized in Figure 5D. Since we had only tested a limited number of DS neurons, and there are also large variations among DS neurons, these values may not be very precise. Nevertheless, what we can see is that the emergence of MB orientation selectivity (85 ms) roughly falls in the time range (51–109 ms) of surround effects operated in the DS neurons. This indicates that DS neurons have the potential to contribute to the MB orientation selectivity in MB neurons. In addition to the population latency method, we also calculated latencies for individual neurons, thus the variance can be evaluated (Figure 5—figure supplement 1C, D). The temporal relationship between MB and DS responses calculated from these two methods was consistent.

We studied MB neurons in awake monkeys performing an orientation-discrimination task. There were 10.9% V2 neurons that exhibited robust, cue-invariant orientation selectivity to MB and luminance stimuli. V2 neurons also exhibited a correlation with animals’ MB orientation-discrimination behavior. Compared with V2, V1 had fewer MB neurons and a much weaker correlation with animals’ MB orientation-discrimination behavior. Evidence from temporal correlation between MB and DS neurons, as well as their response latencies, supports the model in which MB neurons receive input from DS neurons during MB orientation detection.

These results are largely consistent with previous findings on anesthetized monkeys (Marcar et al., 2000), which indicate that V2 is important for MB detection. With array recordings on awake behaving monkey, we provided novel evidence showing that (1) V2 MB neurons may contribute to MB perception and (2) V2 MB detection may use motion information within V2.

Data and codes are available in the Mendeley dataset: https://doi.org/10.17632/fjy37kc8pd.3.

1. 1. Ma H
   2. Li P
   3. Hu J
   4. Cai X
   5. Song Q
   6. Lu HD

   (2021) Mendeley Data

   Processing of motion-boundary orientation in macaque V2.

   https://doi.org/10.17632/fjy37kc8pd.3

1. 1. An X
   2. Gong H
   3. Qian L
   4. Wang X
   5. Pan Y
   6. Zhang X
   7. Yang Y
   8. Wang W

   (2012) Distinct functional organizations for processing different motion signals in V1, V2, and V4 of macaque

   Journal of Neuroscience 32:13363–13379.

   https://doi.org/10.1523/JNEUROSCI.1900-12.2012

   • PubMed
   • Google Scholar
2. 1. Anzai A
   2. Peng X
   3. Van Essen DC

   (2007) Neurons in monkey visual area V2 encode combinations of orientations

   Nature Neuroscience 10:1313–1321.

   https://doi.org/10.1038/nn1975

   • PubMed
   • Google Scholar
3. 1. Bosking WH
   2. Zhang Y
   3. Schofield B
   4. Fitzpatrick D

   (1997) Orientation selectivity and the arrangement of horizontal connections in tree shrew striate cortex

   The Journal of Neuroscience 17:2112–2127.

   https://doi.org/10.1523/JNEUROSCI.17-06-02112.1997

   • PubMed
   • Google Scholar
4. 1. Bredfeldt CE
   2. Cumming BG

   (2006) A simple account of cyclopean edge responses in macaque v2

   Journal of Neuroscience 26:7581–7596.

   https://doi.org/10.1523/JNEUROSCI.5308-05.2006

   • PubMed
   • Google Scholar
5. 1. Britten KH
   2. Newsome WT
   3. Shadlen MN
   4. Celebrini S
   5. Movshon JA

   (1996) A relationship between behavioral choice and the visual responses of neurons in macaque MT

   Visual Neuroscience 13:87–100.

   https://doi.org/10.1017/S095252380000715X

   • Google Scholar
6. 1. Chen M
   2. Yan Y
   3. Gong X
   4. Gilbert CD
   5. Liang H
   6. Li W

   (2014) Incremental integration of global contours through interplay between visual cortical Areas

   Neuron 82:682–694.

   https://doi.org/10.1016/j.neuron.2014.03.023

   • PubMed
   • Google Scholar
7. 1. Chen M
   2. Li P
   3. Zhu S
   4. Han C
   5. Xu H
   6. Fang Y
   7. Hu J
   8. Roe AW
   9. Lu HD

   (2016) An orientation map for motion boundaries in macaque V2

   Cerebral Cortex 26:279–287.

   https://doi.org/10.1093/cercor/bhu235

   • PubMed
   • Google Scholar
8. 1. Chen R
   2. Wang F
   3. Liang H
   4. Li W

   (2017) Synergistic processing of visual contours across cortical layers in V1 and V2

   Neuron 96:1388–1402.

   https://doi.org/10.1016/j.neuron.2017.11.004

   • PubMed
   • Google Scholar
9. 1. DeYoe EA
   2. Van Essen DC

   (1985) Segregation of efferent connections and receptive field properties in visual area V2 of the macaque

   Nature 317:58–61.

   https://doi.org/10.1038/317058a0

   • PubMed
   • Google Scholar
10. 1. El-Shamayleh Y
    2. Movshon JA

    (2011) Neuronal responses to texture-defined form in macaque visual area V2

    Journal of Neuroscience 31:8543–8555.

    https://doi.org/10.1523/JNEUROSCI.5974-10.2011

    • PubMed
    • Google Scholar
11. 1. Gattass R
    2. Gross CG
    3. Sandell JH

    (1981) Visual topography of V2 in the macaque

    The Journal of Comparative Neurology 201:519–539.

    https://doi.org/10.1002/cne.902010405

    • PubMed
    • Google Scholar
12. 1. Hegdé J
    2. Van Essen DC

    (2007) A comparative study of shape representation in macaque visual Areas v2 and v4

    Cerebral Cortex 17:1100–1116.

    https://doi.org/10.1093/cercor/bhl020

    • PubMed
    • Google Scholar
13. 1. Hu J
    2. Ma H
    3. Zhu S
    4. Li P
    5. Xu H
    6. Fang Y
    7. Chen M
    8. Han C
    9. Fang C
    10. Cai X
    11. Yan K
    12. Lu HD

    (2018) Visual motion processing in macaque V2

    Cell Reports 25:157–167.

    https://doi.org/10.1016/j.celrep.2018.09.014

    • PubMed
    • Google Scholar
14. 1. Hubel DH
    2. Livingstone MS

    (1987) Segregation of form, color, and Stereopsis in primate area 18

    The Journal of Neuroscience 7:3378–3415.

    https://doi.org/10.1523/JNEUROSCI.07-11-03378.1987

    • PubMed
    • Google Scholar
15. 1. Ito M
    2. Komatsu H

    (2004) Representation of angles embedded within contour stimuli in area V2 of macaque monkeys

    Journal of Neuroscience 24:3313–3324.

    https://doi.org/10.1523/JNEUROSCI.4364-03.2004

    • PubMed
    • Google Scholar
16. 1. Jazayeri M
    2. Movshon JA

    (2006) Optimal representation of sensory information by neural populations

    Nature Neuroscience 9:690–696.

    https://doi.org/10.1038/nn1691

    • PubMed
    • Google Scholar
17. 1. Kang I
    2. Maunsell JH

    (2012) Potential confounds in estimating trial-to-trial correlations between neuronal response and behavior using choice probabilities

    Journal of Neurophysiology 108:3403–3415.

    https://doi.org/10.1152/jn.00471.2012

    • PubMed
    • Google Scholar
18. 1. Kobatake E
    2. Tanaka K

    (1994) Neuronal selectivities to complex object features in the ventral visual pathway of the macaque cerebral cortex

    Journal of Neurophysiology 71:856–867.

    https://doi.org/10.1152/jn.1994.71.3.856

    • PubMed
    • Google Scholar
19. 1. Kohn A
    2. Smith MA

    (2005) Stimulus dependence of neuronal correlation in primary visual cortex of the macaque

    Journal of Neuroscience 25:3661–3673.

    https://doi.org/10.1523/JNEUROSCI.5106-04.2005

    • PubMed
    • Google Scholar
20. 1. Lamme VA

    (1995) The neurophysiology of figure-ground segregation in primary visual cortex

    The Journal of Neuroscience 15:1605–1615.

    https://doi.org/10.1523/JNEUROSCI.15-02-01605.1995

    • PubMed
    • Google Scholar
21. 1. Leventhal AG
    2. Wang Y
    3. Schmolesky MT
    4. Zhou Y

    (1998) Neural correlates of boundary perception

    Visual Neuroscience 15:1107–1118.

    https://doi.org/10.1017/S0952523898156110

    • PubMed
    • Google Scholar
22. 1. Levitt JB
    2. Kiper DC
    3. Movshon JA

    (1994) Receptive fields and functional architecture of macaque V2

    Journal of Neurophysiology 71:2517–2542.

    https://doi.org/10.1152/jn.1994.71.6.2517

    • PubMed
    • Google Scholar
23. 1. Li P
    2. Zhu S
    3. Chen M
    4. Han C
    5. Xu H
    6. Hu J
    7. Fang Y
    8. Lu HD

    (2013) A motion direction preference map in monkey V4

    Neuron 78:376–388.

    https://doi.org/10.1016/j.neuron.2013.02.024

    • PubMed
    • Google Scholar
24. 1. Liu LD
    2. Pack CC

    (2017) The contribution of area MT to visual motion perception depends on training

    Neuron 95:436–446.

    https://doi.org/10.1016/j.neuron.2017.06.024

    • PubMed
    • Google Scholar
25. 1. Lu HD
    2. Chen G
    3. Tanigawa H
    4. Roe AW

    (2010) A motion direction map in macaque V2

    Neuron 68:1002–1013.

    https://doi.org/10.1016/j.neuron.2010.11.020

    • PubMed
    • Google Scholar
26. 1. Marcar VL
    2. Xiao DK
    3. Raiguel SE
    4. Maes H
    5. Orban GA

    (1995) Processing of kinetically defined boundaries in the cortical motion area MT of the macaque monkey

    Journal of Neurophysiology 74:1258–1270.

    https://doi.org/10.1152/jn.1995.74.3.1258

    • PubMed
    • Google Scholar
27. 1. Marcar VL
    2. Raiguel SE
    3. Xiao D
    4. Orban GA

    (2000) Processing of kinetically defined boundaries in Areas V1 and V2 of the macaque monkey

    Journal of Neurophysiology 84:2786–2798.

    https://doi.org/10.1152/jn.2000.84.6.2786

    • PubMed
    • Google Scholar
28. Book

    1. Mardia KV

    (1972)

    Statistics of Directional Data

    London: Academic Press.

    • Google Scholar
29. 1. Merigan WH
    2. Nealey TA
    3. Maunsell JH

    (1993) Visual effects of lesions of cortical area V2 in macaques

    The Journal of Neuroscience 13:3180–3191.

    https://doi.org/10.1523/JNEUROSCI.13-07-03180.1993

    • PubMed
    • Google Scholar
30. 1. Munk MH
    2. Nowak LG
    3. Girard P
    4. Chounlamountri N
    5. Bullier J

    (1995) Visual latencies in cytochrome oxidase bands of macaque area V2

    PNAS 92:988–992.

    https://doi.org/10.1073/pnas.92.4.988

    • PubMed
    • Google Scholar
31. 1. Mysore SG
    2. Vogels R
    3. Raiguel SE
    4. Orban GA

    (2006) Processing of kinetic boundaries in macaque V4

    Journal of Neurophysiology 95:1864–1880.

    https://doi.org/10.1152/jn.00627.2005

    • PubMed
    • Google Scholar
32. 1. Orban GA
    2. Fize D
    3. Peuskens H
    4. Denys K
    5. Nelissen K
    6. Sunaert S
    7. Todd J
    8. Vanduffel W

    (2003) Similarities and differences in motion processing between the human and macaque brain: evidence from fMRI

    Neuropsychologia 41:1757–1768.

    https://doi.org/10.1016/S0028-3932(03)00177-5

    • PubMed
    • Google Scholar
33. 1. Peterhans E
    2. Heider B
    3. Baumann R

    (2005) Neurons in monkey visual cortex detect lines defined by coherent motion of dots

    European Journal of Neuroscience 21:1091–1100.

    https://doi.org/10.1111/j.1460-9568.2005.03919.x

    • Google Scholar
34. 1. Peterhans E
    2. Heydt R

    (1993) Functional organization of area V2 in the alert macaque

    European Journal of Neuroscience 5:509–524.

    https://doi.org/10.1111/j.1460-9568.1993.tb00517.x

    • Google Scholar
35. 1. Ponce CR
    2. Lomber SG
    3. Born RT

    (2008) Integrating motion and depth via parallel pathways

    Nature Neuroscience 11:216–223.

    https://doi.org/10.1038/nn2039

    • PubMed
    • Google Scholar
36. 1. Ponce CR
    2. Hunter JN
    3. Pack CC
    4. Lomber SG
    5. Born RT

    (2011) Contributions of indirect pathways to visual response properties in macaque middle temporal area MT

    Journal of Neuroscience 31:3894–3903.

    https://doi.org/10.1523/JNEUROSCI.5362-10.2011

    • PubMed
    • Google Scholar
37. 1. Ponce CR
    2. Hartmann TS
    3. Livingstone MS

    (2017) End-Stopping predicts curvature tuning along the ventral stream

    The Journal of Neuroscience 37:648–659.

    https://doi.org/10.1523/JNEUROSCI.2507-16.2016

    • PubMed
    • Google Scholar
38. 1. Regan D

    (1986) Form from motion parallax and form from luminance contrast: vernier discrimination

    Spatial Vision 1:305–318.

    https://doi.org/10.1163/156856886X00106

    • PubMed
    • Google Scholar
39. 1. Regan D

    (1989) Orientation discrimination for objects defined by relative motion and objects defined by luminance contrast

    Vision Research 29:1389–1400.

    https://doi.org/10.1016/0042-6989(89)90194-6

    • PubMed
    • Google Scholar
40. 1. Shipp S
    2. Zeki S

    (1985) Segregation of pathways leading from area V2 to Areas V4 and V5 of macaque monkey visual cortex

    Nature 315:322–324.

    https://doi.org/10.1038/315322a0

    • PubMed
    • Google Scholar
41. 1. Shipp S
    2. Zeki S

    (1989) The organization of connections between Areas V5 and V2 in macaque monkey visual cortex

    European Journal of Neuroscience 1:333–354.

    https://doi.org/10.1111/j.1460-9568.1989.tb00799.x

    • Google Scholar
42. 1. Shipp S
    2. Zeki S

    (2002) The functional organization of area V2, II: the impact of stripes on visual topography

    Visual Neuroscience 19:211–231.

    https://doi.org/10.1017/S0952523802191176

    • PubMed
    • Google Scholar
43. 1. Shoham S
    2. Fellows MR
    3. Normann RA

    (2003) Robust, automatic spike sorting using mixtures of multivariate t-distributions

    Journal of Neuroscience Methods 127:111–122.

    https://doi.org/10.1016/S0165-0270(03)00120-1

    • PubMed
    • Google Scholar
44. 1. Smith MA
    2. Kohn A

    (2008) Spatial and temporal scales of neuronal correlation in primary visual cortex

    Journal of Neuroscience 28:12591–12603.

    https://doi.org/10.1523/JNEUROSCI.2929-08.2008

    • PubMed
    • Google Scholar
45. 1. Thomas OM
    2. Cumming BG
    3. Parker AJ

    (2002) A specialization for relative disparity in V2

    Nature Neuroscience 5:472–478.

    https://doi.org/10.1038/nn837

    • PubMed
    • Google Scholar
46. 1. Vaina LM
    2. Soloviev S
    3. Bienfang DC
    4. Cowey A

    (2000) A lesion of cortical area V2 selectively impairs the perception of the direction of first-order visual motion

    NeuroReport 11:1039–1044.

    https://doi.org/10.1097/00001756-200004070-00028

    • PubMed
    • Google Scholar
47. 1. von der Heydt R
    2. Peterhans E
    3. Baumgartner G

    (1984) Illusory contours and cortical neuron responses

    Science 224:1260–1262.

    https://doi.org/10.1126/science.6539501

    • PubMed
    • Google Scholar
48. 1. von der Heydt R
    2. Zhou H
    3. Friedman HS

    (2000) Representation of stereoscopic edges in monkey visual cortex

    Vision Research 40:1955–1967.

    https://doi.org/10.1016/S0042-6989(00)00044-4

    • PubMed
    • Google Scholar
49. 1. Von der Heydt R
    2. Peterhans E

    (1989)

    Mechanisms of contour perception lines of pattern discontinuity

    Journal of Neuroscience 9:1731–1748.

    • Google Scholar
50. 1. Wichmann FA
    2. Hill NJ

    (2001) The psychometric function: I. fitting, sampling, and goodness of fit

    Perception & Psychophysics 63:1293–1313.

    https://doi.org/10.3758/BF03194544

    • PubMed
    • Google Scholar
51. 1. Yin J
    2. Gong H
    3. An X
    4. Chen Z
    5. Lu Y
    6. Andolina IM
    7. McLoughlin N
    8. Wang W

    (2015) Wei wang breaking cover: neural responses to slow and fast camouflage-breaking motion

    Proceedings. Biological Sciences 282:20151182.

    https://doi.org/10.1098/rspb.2015.1182

    • PubMed
    • Google Scholar
52. 1. Zeki S
    2. Perry RJ
    3. Bartels A

    (2003) The processing of kinetic contours in the brain

    Cerebral Cortex 13:189–202.

    https://doi.org/10.1093/cercor/13.2.189

    • PubMed
    • Google Scholar
53. 1. Zhou H
    2. Friedman HS
    3. von der Heydt R

    (2000) Coding of border ownership in monkey visual cortex

    The Journal of Neuroscience 20:6594–6611.

    https://doi.org/10.1523/JNEUROSCI.20-17-06594.2000

    • PubMed
    • Google Scholar

Acceptance summary:

The study by Lu and colleagues is an in-depth investigation into the selectivity of V2 neurons in the primate for motion-boundaries (MB), the underlying functional circuitry and the potential perceptual contribution of these neurons. Previous studies in anesthetized monkeys demonstrated that V2 contains neurons selective for the orientation of MB, and the authors previously showed that these neurons form a map of MB orientation. In this study, they confirm these previous findings in awake animals and further demonstrate that the responses of MB neurons correlate with the animals' behavioral performance in a MB orientation-discrimination task. The activity of MB neurons also correlates with that of direction selective V2 neurons. These findings are important for our understanding of the functional circuitry underlying shape perception in primates.

Decision letter after peer review:

Thank you for submitting your article "Processing of motion-boundary orientation in macaque V2" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Timothy Behrens as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Alessandra Angelucci (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional revisions and data are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

The study "Processing of motion-boundary orientation in macaque V2" by Lu and colleagues is an in-depth investigation into the selectivity of V2 neurons in the primate for motion-boundaries (MB), the underlying functional circuitry and the potential perceptual contribution of these neurons. Previous studies in anesthetized monkeys demonstrated that V2 contains neurons selective for the orientation of MB, and the authors previously showed that these neurons form a map of MB orientation. In this study, the authors confirm these previous findings in awake animals and further demonstrate that the responses of MB neurons correlate with the animal's behavioral performance in a MB orientation-discrimination task. Moreover, the authors demonstrate that the responses of direction selective (DS) V2 neurons are correlated in time with those of MB neurons and that the visual response latencies of DS cells are consistent with a model in which MB neurons integrate motion signals from DS neurons in V2 itself to extract information about the orientation of the MB. The results are novel and important, the study appears overall well executed.

However the presentation of the manuscript needs improvement by addition of more experimental details, explanations of rationales for some of the analyses, and substantial proof reading to clarify the language. Despite a large numbers (n) of neurons from array recordings, interpretation is in parts difficult because of the low n for neurons showing MB selectivity. To understand the robustness of MB results better, more detail is required in the main manuscript about the distribution of results for MB neurons across array electrodes, across the two animals as well as the relationship of MB results to stimulus size and position with regards to the neuronal selectivity. Finally, the interpretation of the cross-correlation and latency studies is somewhat weak and needs additional analysis and justification.

Detailed comments are provided below.

Essential revisions:

1. To understand the robustness of the results obtained for MB selective neurons better, more detail is required in the main manuscript about the distribution of MB neuron results across array electrodes, across the two animals and about the relationship of stimulus size and position to the neuronal selectivity:

– Figure 1: to what extent would the results be influenced by surround inhibition? Show RF positions and sizes of recorded neurons on the same plot as the stimulus position and size used.

– p6, line 140: 10.9% (70/642) of V2 neurons were MB neurons. From how many individual electrode points did they come in each of the monkeys and hemispheres?

– p8, line 207-226 and Figure 2. CP and neurometric measurements

Given the low n, how many unique electrode points in each of the two monkeys do the data stem from?

– p9: V1/V2 comparison first paragraph ("significantly higher percentage MD detection capability"). Given the small sample size in V1, what is the statistics on differences in distribution – is this really significant given the small sample size ? Also, where "could not be fitted" or "is well fitted" is mentioned, please provide the specific fit, goodness-of-fit measure and statistical criterion in the text.

– p9: V1/V2 comparison. Figure 3. Please show graphically V1 RF positions and size in relation to stimulus position and size. Were the same stimulus parameters used for V1 as for V2. If yes, could surround suppression account for the poorer results in V1 (MB tuning distribution, lower OSI, lower CP)?

– p11. Figure 4: Could the coverage of the DS RF by the preferred motion direction component of the MB stimulus explain some of the interneuronal correlation? For example, between pref and null orientation, when the angle of the dividing line changes, the RF parts that are exposed to preferred and null motion change around the line. Please mention in the main manuscript the range of cortical distances over which pairs of neurons have been simultaneously recorded.

– p12, lines 320- 328: Were TBs resented at pref or null motion direction or both. Were monkeys rewarded for correct choices only as for MBs?

– p14, line 388: please give n of neurons for time course data. Also state for delay times whether this is the mean (across how many neurons?) and what the standard deviation is.

– The authors should discuss how these results might relate to the distribution across different types of stripes (thick, thin, interstripe) across their recording sites in V2.

2. The authors used optical imaging to position the recording arrays to MB-selective domains in V2. How was the exact position of the recording channels relative to the imaging maps of V1 and V2 recovered? Using imaging as the only guidance is relatively imprecise. Could additional criteria to discern between V1 and V2 cells, such as receptive field (RF) size and retinotopy be used?

The authors state that spike sorting was used to isolate single units (SUs) from multiunits (MUA). On p. 4, lines 91-95 they state: "…we found that neurons recorded from the same channel over different days usually had different waveforms and/or tuning-properties, thus were different cells. We compared the results obtained either using the whole dataset ("all-cells", n=723) or the "one-cell-per-channel" dataset (n=85). However, neither approach is accurate, as the first approach will count some of the same neurons twice, and the second approach averages across different, rather than the same, cells. The authors should spike sort the cells, then select SUs based on the spike waveform and tuning across days. That should be their n.

3. Figure 2F: Please describe in the figure legend what the black curve and red dots are. Is this the mean of all neuronal neurometric functions? Also please add the psycometric function on this plot for reference.

4. Lines 223-225: "In 17 neurons recorded in one array, we also tested neural-behavioral relevance with different levels of dot brightness. The results were similar as those from coherence stimuli (Figure S2B-E)". In fact, the results are not so similar. It seems much fewer neurons had CPs that were significantly larger than chance. How do the authors interpret this result?

5. Lines 267-276. Wouldn't a better comparison be between the CP distribution of MB neurons vs non-MB neurons, instead of vs. all V2 neurons? The same apply to the comparison with V1, although We do realize that only 2 cells in their sample passed the test for MB neuron classification.

6. Lines 303-313: a good control here would be to look at the correlation with DS cells whose preference does not match the motion of the random dots generating the MB.

7. Analysis of time correlations. It is unclear to us why the authors interpret CCGs between the responses of DS and MB neurons peaking at zero, as supporting their model of MB responses resulting from integration of DS responses. Doesn't such a model predict that the responses of DS cells should rather precede those of MB cells?

8. Analysis of response time course. The authors find that the latency of MB selectivity falls within the range of latencies of surround suppression in DS neurons, and from this finding conclude that this suggests that V2 DS neurons contribute to the generation of MB responses in V2 MB neurons. The rationale for why surround suppression in DS neurons contributes to MB is unclear to us. MB responses occur within the RF of the V2 MB cells, so how does surround suppression play a role in their generation? Perhaps, we are failing to understand something here, but this needs to be better explained. Also, given the hypothesis here is that DS neurons contribute to MB responses, shouldn't the authors look at the onset latency of direction selectivity in DS neurons? Our rationale is that inputs regarding the direction of RDs need to be fed to the MB neurons in order to extract the MB orientation. The authors need to provide a better explanation of what model they have in mind for how DS cells contribute to the extraction of MB in MB neurons.

9. Lines 440-442. A simple model to support the statement that MB detection results from population coding would help here. At a minimum the authors should provide a sense of how this could be achieved. Also, why couldn't MB detection result from the activity of the fewer neurons that show neurometric functions and thresholds similar to the psychometric functions?

10. Was eye movements correction applied to the data analysis? how? Please specify.

11. Line 825: what algorithm/method was used for spike sorting? Please describe in the Methods.

12. Figure S1 legend line 988-989: the definition of MB neurons provided here does not match that provided in the results, according to which MB neurons were considered those that respond to MB AND are orientation selective. So group E in panel E does not fit this definition.

13. In the methods, the authors often refer to their previous publications for method and analysis details. However, at brief description should be provided in this manuscript. For example, on line 867: how were OSI and DSI defined?

14. Monkey task (page 6, line 162 and methods line 778): to me, the expression that the monkeys did discriminated between an acute or obtuse angle does not mean much in this contexts I am unclear which angle they refer to. If the authors meant, the monkeys discriminated whether the MB line was tilted e.g. left or right of vertical (or another axis), they need to say so more explicitly. The specific tasks the monkey had to carry out need a better description and figures to understand what stimuli configurations would lead to which choices.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Processing of motion-boundary orientation in macaque V2" for further consideration by eLife. Your revised article has been reviewed by 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Timothy Behrens as the Senior Editor.

We found that you have thoroughly revised the manuscript and performed additional analyses according to the Reviewer's comments, which is appreciated. The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1. We are somewhat concerned about the lack of significant difference in the receptive field (RF) size of V1 and V2 neurons. The reported 2.5 deg mean RF size for V1 cells is certainly inconsistent with what has been reported in the literature for parafoveal V1 neurons (more around 0.8-1 deg). This discrepancy to the published literature should be discussed in the paper – alongside the potential reasons for this, specifically:

a. The most likely cause for this is eye movements (in fact the fixation window is reported to be 1.2deg, which is rather coarse for V1). The question here is in what way this imprecision in the mapping of V1 RFs may have affected the results, e.g. the reported differences between V1 and V2 cells. You should at the very least add a discussion of this issue.

b. The illustrated placement of the arrays looks like that for 3 out of 4 arrays very few electrodes would be safely in V1, which could also explain the lack of a difference in RF size. You should specify in the methods of the paper how exactly they determined functionally whether each electrode was in V1 or V2.

c. The V1 RF measurements shown in the authors' response in comparison with V2 should also be included in the supplementary figure (could be added to the supplement to Figure 3).

2. Authors' Reply point 7). Here, you acknowledge that the zero lag in the CCG between DS and MB neuron responses is inconsistent with your hypothesis of DS inputs contributing to the MB responses. Yet, in the Results section of the paper you do not say so, rather you still maintain that the results are consistent with your hypothesis but concede that alternative/additional mechanisms may also be at play. If you indeed agree this result is inconsistent with your hypothesis, you should state so. Alternatively, you should clarify in the manuscript why you think this result may still be consistent with your hypothesis.

3. Authors' Reply point 9. The two alternative model schemes do a good job at clarifying why and you think that surround suppression may play a role in the generation of MB responses and how. We recommend to include this figure in the discussion of the paper.

4. Authors' Reply point 11. Please include in the manuscript your explanation provided here of why the 4 neurons in group E were included in the MB population analysis. This is ok, but this needs to be clear in the manuscript (methods and figure legend need to be consistent).

5. line 1041-1051: ".… sorted response clusters were usually multi-unit responses,.…" "Neurons with inter-spike-intervals larger than 2.5 ms were identified as SUs." The paragraph reads as if most data presented in the paper were multi-unit, but were re-classified as SU based on spike timing only, which by itself is not enough. The main results (line 106) however state that SUs were identified on unique spike wave form and asserts that most cells are SU. We presume this means that the SUs were clearly separable from the remaining MU or noise.

There need to be clear, consistent statements in both Results and Methods, as the reader needs to know (i) whether they are dealing with SU or MU results and (ii) if both were combined whether specific results about basic MB tuning differ for clear SUs und MUs.

6. The reviewers would be grateful if you could please always point to the pages in the revised manuscript where the revisions were made. It really saves reviewers a lot of time.

Essential revisions:

1. To understand the robustness of the results obtained for MB selective neurons better, more detail is required in the main manuscript about the distribution of MB neuron results across array electrodes, across the two animals and about the relationship of stimulus size and position to the neuronal selectivity:

– Figure 1: to what extent would the results be influenced by surround inhibition? Show RF positions and sizes of recorded neurons on the same plot as the stimulus position and size used.

Figure 3—figure supplement 1 shows the spatial relationship between V2 RFs and the stimuli (A-D). The dotted circles represent the 4° stimulus patch. Since the stimuli were not adjusted for individual neurons, their size was chosen to slightly larger than the V2 RF (2.47±0.85°), Therefore, there should be some surround inhibition for these neurons. However, this effect should be similar for the MB and non-MB neurons, since the sizes of their RFs were similar (C, same as in Figure 1J), and their RF centers were located in similar distances to the stimulus center (D). In addition, the major stimuli we used were oppositely moving random dots, which were different from those homogenous stimuli that usually cause strong surround inhibition.

Similarly, V1 RFs are also plotted in Figure 3—figure supplement 1E-H, and related to the answers to another question below.

– p6, line 140: 10.9% (70/642) of V2 neurons were MB neurons. From how many individual electrode points did they come in each of the monkeys and hemispheres?

The 70 MB neurons were from 26 individual electrodes. In the revised manuscript, we added a table showing the detailed information (Supplementary File 1).

– p8, line 207-226 and Figure 2. CP and neurometric measurements

Given the low n, how many unique electrode points in each of the two monkeys do the data stem from?

In the revised manuscript, we included a table (Supplementary File 2) showing the number of unique electrodes from each hemisphere that neurons’ CPs were calculated.

– p9: V1/V2 comparison first paragraph ("significantly higher percentage MD detection capability"). Given the small sample size in V1, what is the statistics on differences in distribution – is this really significant given the small sample size ? Also, where "could not be fitted" or "is well fitted" is mentioned, please provide the specific fit, goodness-of-fit measure and statistical criterion in the text.

In the revised manuscript, we added statistics to support the above conclusions. Author response table 1 shows the neuron numbers being compared (same as in Figure 3A). V1 and V2 were different in their cell distributions in the 3 groups (Chi-squared test, χ²=14.2，p=8.25×10^-4). In addition, the mean orientation index (OSI) for MB stimuli was larger for V2 neurons than for V1 neurons (calculated from the neurons in the first two rows whose responses could be well fitted, i.e. R²≥0.7). We also added fitting criterion into the main text: The fitting criteria for MB was the same as for gratings, i.e. with von Mises function, R²<0.7 was considered not fitted; R²≥0.7 was well fitted.

– p9: V1/V2 comparison. Figure 3. Please show graphically V1 RF positions and size in relation to stimulus position and size. Were the same stimulus parameters used for V1 as for V2. If yes, could surround suppression account for the poorer results in V1 (MB tuning distribution, lower OSI, lower CP)?

The same stimuli were used for V1 and V2 neurons in our experiments. In the figure for question 1 above, we also plotted V1 RF positions and sizes. V1 and V2 neurons we recorded had similar RF sizes (V1: 2.50±1.18°; V2: 2.47±0.85°; p=0.77, Wilcoxon test). The distances between the RF centers and stimulus center were slightly larger for V1 neurons (V1: 0.47±0.41°; V2: 0.33±0.24°; p=2.97×10^-4, Wilcoxon test), but the difference was relatively small (0.14°). We also compared the firing rate of V1 and V2 neurons in two arrays having sufficient numbers of V1 neurons, and observed no differences (Monkey S, right hemisphere: V2 13.95±8.60 spike/s; V1: 12.63±6.84 spike/s, p=0.45, t-test; Monkey W，right hemisphere: V2: 15.55±6.34 spike/s; V1: 15.36±13.60 spike/s, p=0.93, t-test). Thus, we believe that the surround inhibition caused by the common stimuli was similar for V1 and V2, and was not the cause for the poorer results in V1.

– p11. Figure 4: Could the coverage of the DS RF by the preferred motion direction component of the MB stimulus explain some of the interneuronal correlation? For example, between pref and null orientation, when the angle of the dividing line changes, the RF parts that are exposed to preferred and null motion change around the line. Please mention in the main manuscript the range of cortical distances over which pairs of neurons have been simultaneously recorded.

The reviewer is correct that a DS neuron might have different response levels to the two MB orientations (preferred and null) due to the change of preferred motion in its RF. However, since this factor was random for the two MB orientations, it would not cause a biased enhancement of the MB-DS correlation for a particular MB orientation. To confirm this, we identified and removed DS neurons (n=4) that had large RF coverage differences in preferred and null conditions and found that the auCCG was still significantly larger for the preferred condition (preMB: 0.10±0.016 vs. nullMB: 0.05±0.011, p=3.11X10^-4, t-test). The cortical distances of paired neurons in Figure 4A-C had a range of 0.93~1.1 mm. We have added this information into the main text.

– p12, lines 320- 328: Were TBs resented at pref or null motion direction or both. Were monkeys rewarded for correct choices only as for MBs?

Yes, the TB stimuli were presented at both preferred and null directions. The monkeys were rewarded for correct choices. However, the correct rate for TB was around 50%.

– p14, line 388: please give n of neurons for time course data. Also state for delay times whether this is the mean (across how many neurons?) and what the standard deviation is.

We have added n=9 for the number of DS neurons. We calculated delay times with two methods, the one described in the main text was based on population time courses (no standard deviation), and another based on individual neurons’ time courses (Figure 5—figure supplement 1D). We’ve now described this in the main text.

– The authors should discuss how these results might relate to the distribution across different types of stripes (thick, thin, interstripe) across their recording sites in V2.

We have added discussion about the stripe types in the manuscript (line 531-542). Since CO histology is unavailable for these monkeys. The stripe types were estimated based on optical imaging maps (ocular dominance, orientation, color, motion direction). To separate thick and pale stripes, we use a width ratio previously described (thick : pale=2 : 1.5, Shipp and Zeki 1989). Figure 1—figure supplement 2 shows the locations of the 4 arrays overlaid on the corresponding color vs. luminance maps. The white dotted lines represent V1/V2 borders. The thin stripes were identified based on the color/luminance patches in V2 (black and white patched in V2). Thick (between two red dashed lines) and pale (between green and red dashed lines) stripes were identified between the thin stripes and according to their width ratio. For all V2 channels, the total numbers of channels located in thin, thick and pale stripes were: 15, 44, and 37, respectively. Figure 1—figure supplement 2M lists neuron numbers according to their stripe and tuning types. Note that the tuning types were not mutually exclusive (e.g. a MB neuron could also be an orientation neuron). The percentage values were calculated for each stripe types. Although a statistical analysis was not available due to the overlapped tuning types, one can still draw some conclusions from the table. For example, DS neurons were more likely found in thick stripes, while MB neurons were more likely found in thick and pale stripes. These results were consistent with previous findings (Hubel and Livingstone, 1987; Levitt et al., 1994; Shipp and Zeki, 2002; Lu et al. 2010; Chen et al. 2016).

2. The authors used optical imaging to position the recording arrays to MB-selective domains in V2. How was the exact position of the recording channels relative to the imaging maps of V1 and V2 recovered? Using imaging as the only guidance is relatively imprecise. Could additional criteria to discern between V1 and V2 cells, such as receptive field (RF) size and retinotopy be used?

We estimated the electrode locations by comparing the blood vessel pictures taken before and after the array implants. We agree with the reviewer that this method is relatively imprecise (the maximum offset we estimate is within 200 um). According to the reviewers’ suggestion, we compared the RF sizes and retinotopy for the V1 and V2 neurons. However, no significant trends were observed, and thus did not help in determining the V1/V2 borders. In one array we also tested the monocular/binocular driven properties of the neurons. Probably due to the superficial layers the electrodes located (layer 2/3), neither did we observe monocularly-driven neurons even in electrodes obviously located in V1. A post-mortem histology would help but is not available at this time.

The authors state that spike sorting was used to isolate single units (SUs) from multiunits (MUA). On p. 4, lines 91-95 they state: "…we found that neurons recorded from the same channel over different days usually had different waveforms and/or tuning-properties, thus were different cells. We compared the results obtained either using the whole dataset ("all-cells", n=723) or the "one-cell-per-channel" dataset (n=85). However, neither approach is accurate, as the first approach will count some of the same neurons twice, and the second approach averages across different, rather than the same, cells. The authors should spike sort the cells, then select SUs based on the spike waveform and tuning across days. That should be their n.

Following the reviewers’ suggestions, we used a “unique-unit method”. In unique-unit method, we first spike sorted the neurons according to their waveforms, then identified single-units (SUs) that had unique waveforms or tuning properties. If for a channel no SU was found from all the recording days, we used multi-units (MUs) from different cell types as supplements. We obtained 287 unique-units (251 SUs and 36 MUs) in V2 with this method. We repeated the data analysis on this cell set and obtained similar results (Figure 5—figure supplement 3) as those obtained with the cell sets described in the original manuscript. Thus, we have 3 cell identification methods: “all-cells n=723”, “unique-units, n=287”, and “one-cell-per-channel, n=85”, in order of the selection strictness. The main conclusions obtained from these 3 datasets were the same. We added this into the manuscript.

Additional notes: In “one-cell-per-channel” method, we identified only one cell for each channel (instead of averaging all the cells in that channel), and thus this method is the strictest one among the 3 methods.

3. Figure 2F: Please describe in the figure legend what the black curve and red dots are. Is this the mean of all neuronal neurometric functions? Also please add the psycometric function on this plot for reference.

In Figure 2 legend we have added appropriate descriptions. The red dots are mean values of the neurometric functions and the black curve is the fitting curve. In Figure 2F, we also added the psychometric function obtained by averaging the behavioral performance during the 15 neurons’ recordings (green dots). Similarly, average psychometric function was added into Figure 2—figure supplement 1C for neurons tested with brightness stimuli.

4. Lines 223-225: "In 17 neurons recorded in one array, we also tested neural-behavioral relevance with different levels of dot brightness. The results were similar as those from coherence stimuli (Figure S2B-E)". In fact, the results are not so similar. It seems much fewer neurons had CPs that were significantly larger than chance. How do the authors interpret this result?

We agree that the overall CP was lower in neurons tested with dot brightness. This might be due to the relatively easiness for the brightness task. In 5-6 brightness levels, the monkey achieved more than 80% correct rate. In contrast, the same monkey achieved 80% correct rate in only 3-4 levels in the coherence tasks. An easier task requires less effort and may lead to lower CP values. Nevertheless, in both tasks, the average CPs were both larger than 0.5 and portions of neurons had CP values larger than 0.5. We revised the manuscript to reflect both the differences and similarity.

5. Lines 267-276. Wouldn't a better comparison be between the CP distribution of MB neurons vs non-MB neurons, instead of vs. all V2 neurons? The same apply to the comparison with V1, although We do realize that only 2 cells in their sample passed the test for MB neuron classification.

We added a figure (Figure 3—figure supplement 2) into the manuscript, in which CP distributions for MB and non-MB neurons are compared. In V2, although non-MB neurons had a lower mean CP (0.53) than the MB neurons’ (0.56), this value was still significantly larger than 0.5 (t-test, p<0.001). There were 31.5% (34/108) non-MB neurons had a CP larger than 0.5 (bootstrap test, p<0.05). In comparison, this proportion in MB neurons was 46.9% (15/32). In V1, non-MB neurons had a non-significant mean CP (0.51) and the portion of significant CP neurons was low (21%). These results were now added into the manuscript.

6. Lines 303-313: a good control here would be to look at the correlation with DS cells whose preference does not match the motion of the random dots generating the MB.

We took the reviewers’ suggestion and analyzed DS-MB pairs in which DS neurons were not optimally stimulated. The results are shown in Author response image 1. It shows that the CCGs for preferred MB and null MB do not differ (5 pairs, t-test, p=0.26), which is consistent with our hypothesis. In the revised manuscript, we have included this results.

Author response image 1

Download asset Open asset

7. Analysis of time correlations. It is unclear to us why the authors interpret CCGs between the responses of DS and MB neurons peaking at zero, as supporting their model of MB responses resulting from integration of DS responses. Doesn't such a model predict that the responses of DS cells should rather precede those of MB cells?

Indeed, according to our model, the DS neurons made contribution to the MB neurons, and CCG should peak at the right side of zero. However, the peak we observed was centered at zero (Figure 3A). In addition, similar peaks were also observed in control conditions (e.g. real line in Figure 3B). These phenomena were not predicted by our model, and possibly due to other extra connections. For example, both DS and MB neurons might receive certain common inputs, or these neurons had bi-directional interactions (Bastos and Schoffelen 2016). Similar phenomena were also observed in previous pair-recording studies (e.g. orientation neurons in V1 and V2 in Roe and Ts’o 2015). We revised the interpretation in the manuscript accordingly.

8. Analysis of response time course. The authors find that the latency of MB selectivity falls within the range of latencies of surround suppression in DS neurons, and from this finding conclude that this suggests that V2 DS neurons contribute to the generation of MB responses in V2 MB neurons. The rationale for why surround suppression in DS neurons contributes to MB is unclear to us. MB responses occur within the RF of the V2 MB cells, so how does surround suppression play a role in their generation? Perhaps, we are failing to understand something here, but this needs to be better explained. Also, given the hypothesis here is that DS neurons contribute to MB responses, shouldn't the authors look at the onset latency of direction selectivity in DS neurons? Our rationale is that inputs regarding the direction of RDs need to be fed to the MB neurons in order to extract the MB orientation. The authors need to provide a better explanation of what model they have in mind for how DS cells contribute to the extraction of MB in MB neurons.

We took the reviewers’ suggestion and considered two DS-MB models, in which surround suppression plays different roles (see Figure 5—figure supplement 4).

In model 1, MB detection mainly relies on the center-surround mechanism of the DS neurons, which detect the motion contrast at the MB. Our previous work has shown that, in this condition DS neurons are activated optimally (Figure 5 in Hu et al. 2018). Also, the precise location of the MB is detected at the same time. In the primate visual system, such a center-surround mechanism appears to be used as a general strategy in detecting changes of first order cues at edges (e.g. luminance, color, disparity etc.).

In model 2, MB detection is achieved by comparing the activation of two DS neurons, and surround suppression plays a less significant role in this condition. We have shown that DS neurons are sub-optimally activated in such conditions (Figure 5 in Hu et al. 2018), and a precise MB location is unavailable.

Thus, we mainly considered Model 1 in our manuscript. The onset time of direction selectivity for DS neuron was 49 ms. It was faster than the MB-selectivity time of MB neurons (85 ms) and thus was not against Model 2. In the revised manuscript we added relevant discussions.

9. Lines 440-442. A simple model to support the statement that MB detection results from population coding would help here. At a minimum the authors should provide a sense of how this could be achieved. Also, why couldn't MB detection result from the activity of the fewer neurons that show neurometric functions and thresholds similar to the psychometric functions?

In our results, the MB orientation index is a continuous distribution (Figure 3A). Similarly, the CP distribution does not show a bimodal distribution (Figure 3B). These results indicate that MB information is likely coded in a distributed fashion in V2. Since single neuron’s response is intrinsically noisy, a detection mechanism only relies on a few high-performance neurons may not be the best solution. Population coding strategy overcomes this by using all available useful information. For example, a weighted summation model (Jazayeri and Movshon 2006) can be used in MB detection task. In this model, a likelihood function (i.e. a probability density function) is calculated based on the weighted-summing of all activated neurons (optimally and sub-optimally). Based on the likelihood function, a series of behavioral tasks (e.g. detection, discrimination) can be achieved. We have added this discussion into the manuscript.

10. Was eye movements correction applied to the data analysis? how? Please specify.

We continuously monitored the eye movements during the awake experiments. The fixation window was a 1.2° square. If the eye position moved outside this window for 20 ms, the task was aborted and the data was discarded. Therefore, all the data presented in the manuscript was collected for eye movement within ±0.6°. During the awake experiments, the measured eye position usually had some baseline drift over the time due to the system factors. We monitored the eye position and corrected the drift manually when necessary.

We calculated the mean eye movement for each monkey during the fixation period: Monkey S: 0.05° (horizontal) and 0.08° (vertical); Monkey W: 0.05° (horizontal) and 0.1° (vertical). These values were relatively small. We did not find a correlation between the eye movement size and the main results (behavioral performance, CP, and auCCG).

11. Line 825: what algorithm/method was used for spike sorting? Please describe in the Methods.

We added the spike sorting procedures into the Method.

12. Figure S1 legend line 988-989: the definition of MB neurons provided here does not match that provided in the results, according to which MB neurons were considered those that respond to MB AND are orientation selective. So group E in panel E does not fit this definition.

Our definition of MB neuron was: Their MB responses can be fitted by a von Mises function, and exhibit orientation selectivity to MB stimuli (i.e. OSI_MB >0.5). In Figure 1—figure supplement 1E, the 4 neurons in group E were not orientation neurons (did not reach the criteria for orientation selectivity to grating stimuli). However, they exhibited MB orientation selectivity. Thus, these 4 neurons were considered MB neurons and included in the MB results.

13. In the methods, the authors often refer to their previous publications for method and analysis details. However, at brief description should be provided in this manuscript. For example, on line 867: how were OSI and DSI defined?

We now added descriptions in methods and do not relies on previous publications.

OSI=1- null-orientation response /preferred orientation response;

DSI=1- anti-preferred direction response /preferred direction response;

14. Monkey task (page 6, line 162 and methods line778): to me, the expression that the monkeys did discriminated between an acute or obtuse angle does not mean much in this contexts I am unclear which angle they refer to. If the authors meant, the monkeys discriminated whether the MB line was tilted e.g. left or right of vertical (or another axis), they need to say so more explicitly. The specific tasks the monkey had to carry out need a better description and figures to understand what stimuli configurations would lead to which choices.

We revised the text and Figure to reflect the details of animal tasks.

References:

Anzai A, Peng X, Van Essen DC. Neurons in monkey visual area V2 encode combinations of orientations. Nat Neurosci. 2007;10:1313-1321.

Bastos, A. M., and Schoffelen, J.-M. (2016). A Tutorial Review of Functional Connectivity Analysis Methods and Their Interpretational Pitfalls. Frontiers in Systems Neuroscience, 9:175, 1–23.

Bredfeldt, C. E., and Cumming, B. G. (2006). A simple account of cyclopean edge responses in macaque V2. Journal of Neuroscience, 26(29), 7581–7596.

Britten, K. H., Shadlen, M. N., Newsome, W. T., and Movshon, J. a. (1992). The analysis of visual motion: a comparison of neuronal and psychophysical performance. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 12(12), 4745–4765.

Chen, M., Li, P., Zhu, S., Han, C., Xu, H., Fang, Y., Hu, J., Roe, A.W., and Lu, H.D. (2016). An Orientation Map for Motion Boundaries in Macaque V2. Cereb. Cortex 26, 279–287.

DeYoe, E.A. and Van Essen, D.C. (1985). Segregation of efferent connections and receptive field properties in visual area 2 of the macaque. Nature 317, 58–61.

El-Shamayleh, Y., and Anthony Movshon, J. (2011). Neuronal responses to texture-defined form in macaque visual area V2. Journal of Neuroscience, 31(23), 8543–8555.

Gattass, R., Gross, C. G., and Sandell, J. H. (1981). Visual topography of V2 in the macaque. Journal of Comparative Neurology, 201(4), 519–539.

Hegdé, J., and Van Essen, D. C. (2007). A comparative study of shape representation in macaque visual areas V2 and V4. Cerebral Cortex, 17(5), 1100–1116.

Hu, J., Ma, H., Zhu, S.,, Li, P., Xu, H., Fang, Y., Chen, M., Han, C., Fang, C., Cai, X., Yan, K., Lu, H.D. (2018) Visual Motion Processing in Macaque V2. Cell Reports. 25, 157–167..

Hubel D. H., Livingstone M. S. (1987). Segregation of form, color and stereopsis in primate area 18. J. Neurosci. 7, 3378–3415

Ito, M., and Komatsu, H. (2004). Representation of Angles Embedded within Contour Stimuli in Area V2 of Macaque Monkeys. Journal of Neuroscience, 24(13), 3313–3324.

Jazayeri, M., and Movshon, J. A. (2006). Optimal representation of sensory information by neural populations. Nature Neuroscience, 9(5), 690–696.

Jiapeng Yin, Hongliang Gong, Xu An, Zheyuan Chen, Yiliang Lu, Ian M. Andolina, Niall McLoughlin, Wei Wang Breaking cover: neural responses to slow and fast camouflage-breaking motionProc Biol Sci. 2015 282: 20151182.

Kobatake, E., and Tanaka, K. (1994). Neuronal selectivities to complex object features in the ventral visual pathway of the macaque cerebral cortex. Journal of Neurophysiology, 71, 856–867.

Levitt, J. B., Kiper, D. C., and Movshon, J. A. (1994). Receptive fields and functional architecture of macaque V2. Journal of Neurophysiology, 71(6), 2517–2542.

Lu, H.D., Chen, G., Tanigawa, H., and Roe, A.W. (2010). A motion direction map in macaque V2. Neuron 68, 1002–1013.

M H Munk, L G Nowak, P Girard, N Chounlamountri, J Bullier Visual latencies in cytochrome oxidase bands of macaque area V2. Proc Natl Acad Sci U S A. 1995 92: 988-992.

Marcar, V.L., Raiguel, S.E., Xiao, D., and Orban, G.A. (2000). Processing of kinetically defined boundaries in areas V1 and V2 of the macaque monkey. J. Neurophysiol. 84, 2786–2798.

Peterhans, E., Heider, B., and Baumann, R. (2005). Neurons in monkey visual cortex detect lines defined by coherent motion of dots. European Journal of Neuroscience, 21(4), 1091–1100.

Peterhans, E., and von der Heydt, R. (1993). Functional Organization of Area V2 in the Alert Macaque. European Journal of Neuroscience, 5(5), 509–524.

Roe, A. W., and Ts’o, D. Y. (2015). Specificity of V1-V2 orientation networks in the primate visual cortex. Cortex, 72, 168–178.

Shipp, S. and Zeki, S. (1985). Segregation of pathways leading from area V2 to areas V4 and V5 of macaque monkey visual cortex. Nature 315, 322–325.

Shipp S, Zeki S (1989) The organization of connections between areas V5 and V2 of macaque monkey visual cortex. Eur J Neurosci 1:333–354.

Shipp, S., and Zeki, S. (2002). The functional organization of area V2, II: The impact of stripes on visual topography. Visual Neuroscience, 19(2), 211–231.

Sincich LC, Horton JC. The circuitry of V1 and V2: integration of color, form, and motion. Annual Review of Neuroscience. 2005;28:303–326.

Thomas OM, Cumming BG, Parker AJ. A specialization for relative disparity in V2. Nat Neurosci 5: 472–478, 2002.

Von der Heydt, R., Peterhans, E., and Baumgartner, G. (1984). Illusory contours and cortical neuron responses. Science, 224(4654), 1260–1262.

Von der Heydt, R., and Peterhans, E. (1989). Mechanisms of Contour Perception Lines of Pattern Discontinuity. Journal of Neuroscience, 9(5), 1731-1748.

Von der Heydt R, Zhou H, Friedman HS (2000) Representation of stereoscopic edges in monkey visual cortex. Vision Res 40:1955–1967.

Zhou, H., Friedman, H. S., and von der Heydt, R. (2000). Coding of Border Ownership in Monkey Visual Cortex. The Journal of Neuroscience, 20(17), 6594–6611.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

We found that you have thoroughly revised the manuscript and performed additional analyses according to the Reviewer's comments, which is appreciated. The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1. We are somewhat concerned about the lack of significant difference in the receptive field (RF) size of V1 and V2 neurons. The reported 2.5 deg mean RF size for V1 cells is certainly inconsistent with what has been reported in the literature for parafoveal V1 neurons (more around 0.8-1 deg). This discrepancy to the published literature should be discussed in the paper – alongside the potential reasons for this, specifically:

a. The most likely cause for this is eye movements (in fact the fixation window is reported to be 1.2deg, which is rather coarse for V1). The question here is in what way this imprecision in the mapping of V1 RFs may have affected the results, e.g. the reported differences between V1 and V2 cells. You should at the very least add a discussion of this issue.

We agree with the reviewers that the measured RF size for V1 (2.5 deg) was larger than those previously reported. Eye movements could be a cause, but it would equally increase the measurements of RF sizes for V2 neurons. In our recordings, the mean RF size for V2 neurons was 2.47 deg. Thus, we suspect that other factors (see below) might contribute more to the large V1 RF we measured. In the revised manuscript, whether the imprecision of V1 RF mapping affected the results was discussed (page 19 line 537-556).

b. The illustrated placement of the arrays looks like that for 3 out of 4 arrays very few electrodes would be safely in V1, which could also explain the lack of a difference in RF size. You should specify in the methods of the paper how exactly they determined functionally whether each electrode was in V1 or V2.

We added detailed description in Methods on how we determined that the electrodes were in V1 (page 34 line 1116-1123). We also added V1 ocular dominance maps into Figure 1—figure supplement 2 to show how the V1/V2 borders were determined.

Although these maps show clear V1/V2 borders, the actual transition from V1 neurons to V2 neurons was unlikely that sharp, especially for neurons in superficial layers. There should be a narrow “transition zone” in which V1 and V2 neurons were actually mixed or inseparable. Plus, there were always precision limits in our map-array alignments. Thus, it is possible that some of our V1 neurons were actually V2 neurons, or the other way around. We analyzed a subset of V1 cells (n=39) that were recorded from the electrodes further away from the V1/V2 border and found that their mean RF size was relatively smaller (2.05 deg).

The stimulus we used in measuring the RF size was a 0.8 deg grating patch, its size was also a little too large for V1 RF mapping. We added above information into discussion (page 19 line 537-556).

c. The V1 RF measurements shown in the authors' response in comparison with V2 should also be included in the supplementary figure (could be added to the supplement to Figure 3).

We have added the RF measurements into Figure 3—figure supplement 1.

2. Authors' Reply point 7). Here, you acknowledge that the zero lag in the CCG between DS and MB neuron responses is inconsistent with your hypothesis of DS inputs contributing to the MB responses. Yet, in the Results section of the paper you do not say so, rather you still maintain that the results are consistent with your hypothesis but concede that alternative/additional mechanisms may also be at play. If you indeed agree this result is inconsistent with your hypothesis, you should state so. Alternatively, you should clarify in the manuscript why you think this result may still be consistent with your hypothesis.

We revised the Results section to clearly indicate that the zero time lag is inconsistent with a simple DS-MB contribution model (page 15 line 407-410).

3. Authors' Reply point 8. The two alternative model schemes do a good job at clarifying why and you think that surround suppression may play a role in the generation of MB responses and how. We recommend to include this figure in the discussion of the paper.

We included this figure as Figure 5—figure supplement 4 and linked it to the relevant discussion (page 21 line 615-630).

4. Authors' Reply point 11. Please include in the manuscript your explanation provided here of why the 4 neurons in group E were included in the MB population analysis. This is ok, but this needs to be clear in the manuscript (methods and figure legend need to be consistent).

We have added the explanation into the Method section and the corresponding figure legend.

5. line 1041-1051: ".… sorted response clusters were usually multi-unit responses,.…" "Neurons with inter-spike-intervals larger than 2.5 ms were identified as SUs." The paragraph reads as if most data presented in the paper were multi-unit, but were re-classified as SU based on spike timing only, which by itself is not enough. The main results (line 106) however state that SUs were identified on unique spike wave form and asserts that most cells are SU. We presume this means that the SUs were clearly separable from the remaining MU or noise.

There need to be clear, consistent statements in both Results and Methods, as the reader needs to know (i) whether they are dealing with SU or MU results and (ii) if both were combined whether specific results about basic MB tuning differ for clear SUs und MUs.

We thank the reviewer for pointing out these unclear statements. In the revised manuscript, we have clarified the definition of SU (page 33 line 1073-1079), how much percentages of SUs in the 3 datasets, and whether inclusion of MUs affects the results (page 5 line 100-119 and page 33-34 line 1087-1100).

6. The reviewers would be grateful if you could please always point to the pages in the revised manuscript where the revisions were made. It really saves reviewers a lot of time.

We are very sorry for the inconvenience we made, we did so in this response letter.

Author details

1. Heng Ma

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Writing - original draft, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0322-278X

2. Pengcheng Li

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Formal analysis, Investigation

   Competing interests

   No competing interests declared

3. Jiaming Hu

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Software, Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5306-445X

4. Xingya Cai

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-7829-3833

5. Qianling Song

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9177-7429

6. Haidong D Lu

   State Key Laboratory of Cognitive Neuroscience and Learning, IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Supervision, Funding acquisition, Validation, Visualization, Writing - original draft, Project administration, Writing - review and editing

   For correspondence

   haidong@bnu.edu.cn

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1739-9508

• 786

  Page views

• 111

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.