Size tuning of neural response variability in laminar circuits of macaque primary visual cortex
- Department of Ophthalmology and Visual Science, Moran Eye Institute, University of Utah, United States
Figures
Figure 1
Size tuning of Fano-factor and mean firing rate in macaque V1: representative units.
(A) Representative supragranular (SG) layer unit. Left: MU spike rasters measured at two stimulus diameters, either a diameter equal to the receptive field (RF) diameter of the recorded MU (top), or a diameter of 26° (bottom). Middle: Peri-stimulus time histograms (PSTHs) of Fano-factor (red) and mean firing rate (black) computed in a 100-ms rectangular sliding window for the same two stimulus diameters. The shaded area represents the standard deviation (s.d.) of the bootstrapped Fano-factor distribution (for the Fano-factor curve) or the standard-error-of-the-mean (s.e.m., for the firing-rate curve). Inset: Zoomed-in Fano-factor curves for the smaller (darker red) and larger (lighter red) stimulus diameters between 50 and 350 ms after stimulus onset. Right: Fano-factor (red) and firing rate (black) averaged over 50–350 ms after stimulus onset and plotted against the stimulus diameter. Solid lines: Fits to the data. Dashed lines: Baseline Fano-factor (red) and firing rate (black), measured prior to stimulus onset. Error bars are s.d. of the bootstrapped Fano-factor distribution (red) or s.e.m. (black). (B) Representative granular (G) layer unit. (C) Representative infragranular (IG) layer unit. Conventions in (B, C) are as in (A).
Figure 2 with 3 supplements
Size tuning of Fano-factor and mean firing rate: MU population data.
(A) Average Fano-factor (red) and mean firing rate (black) as a function of stimulus diameter for the population of SG (left; n = 31), G (middle; n = 15), and IG (right; n = 36) layer MUs. Dashed lines: Average baseline Fano-factor (red) and firing rate (black); error bars: s.d. of the bootstrap distribution. Before averaging, stimulus diameter, Fano-factor, and firing rate of each multi-unit were normalized to the value at the receptive field (RF) diameter. In each panel, the black arrow points to the RF size, and the gray arrow to the stimulus size at minimum Fano-factor value. (B) Top: RF-normalized Fano-factor values averaged separately for six stimulus diameters: 0.1° or 0.2° (depending on which was the smallest diameter used for that penetration), a diameter equal to the RF diameter, a diameter twice the RF diameter (near surround), four times the RF diameter, eight times the RF diameter, and 26°. Error bars: s.d. of the bootstrap distribution. Bottom: Box plots of normalized Fano-factor for MUs in SG, G, and IG layers at the same six stimulus diameters as shown in the top panel. Dots: Fano-factor values for individual MUs; red horizontal lines: medians. The extent of the box marks the inter-quartile range of the data, and the whiskers mark the extent of the entire distribution, save for outliers that were outside 1.5 times the inter-quartile range. (C) Median percent change in Fano-factor relative to baseline induced by a stimulus matched in size to the RF diameter, for the different layers. Dots: Individual data points. Error bars (red): s.d. of the bootstrapped distribution. Black and white dots indicate data from different animals (case numbers indicated in panel D). (D) Median percent change in Fano-factor induced by a stimulus twice the RF diameter (near surround) relative to the Fano-factor value evoked by a stimulus matched to the RF diameter. Other conventions as in panel C. (E) Median percent change in Fano-factor induced by a stimulus with diameter equal to the surround diameter for individual cells (see Methods), relative to the Fano-factor value evoked by a stimulus matched to the RF diameter. Other conventions as in panel (C). (F) Median percent change in Fano-factor induced by a 26° diameter stimulus, relative to the Fano-factor value evoked by a stimulus matched to the RF diameter. Other conventions as in panel (C). (G) Median stimulus diameter at the smallest Fano-factor value (or max quenching), normalized to the RF diameter of each recorded MU, for different layers. Error bars: s.d. of the bootstrapped distributions. Dots: Individual MUs. (H) Median surround stimulus diameter at the highest Fano-factor value, normalized to the RF diameter. A value of 1 indicates that the highest Fano-factor value occurred at a stimulus size equal to the RF diameter. Error bars: s.d. of the bootstrapped distributions. Dots: Individual MUs. (I) Scatter plot of percent change in Fano-factor evoked by a stimulus with diameter twice the RF diameter relative to the Fano-factor evoked by a stimulus matched to the RF diameter vs. suppression index (see Methods). Color dots identify units in different layers, and the different symbols show data from two different animals, as indicated. Lines are regression lines fitted to the individual layer data. (J) Scatter plot of percent change in Fano-factor evoked by a stimulus with diameter equal to the per-neuron RF surround diameter relative to the Fano-factor evoked by a stimulus matched to the RF diameter vs. suppression index. Other conventions as in (I). (K) Percent change in Fano-factor evoked by a 26° diameter stimulus relative to the Fano-factor evoked by a stimulus matched to the RF diameter vs. suppression index. Other conventions as in (I).
Figure 2—figure supplement 1
Size tuning of Fano-factor across penetrations.
(A) Average size tuning curves for cells in different penetrations (indicated by different colors), for each layer group. Top: Non-normalized data. Bottom: Fano-factor and grating diameter are both normalized to their values at the receptive field (RF) diameter. Black dashed line marks Fano-factor value at the RF diameter. (B) Median stimulus diameter at the smallest Fano-factor value (max quenching), normalized to the RF diameter of each recorded MU, for different layers. Error bars: s.d. of the bootstrapped distributions. Same data as in Figure 2G, but here the dots represent individual MUs color coded by penetration number as indicated in the legend. (C) Median surround stimulus diameter at the highest Fano-factor value (max facilitation), normalized to the RF diameter. Error bars: s.d. of the bootstrapped distributions. Same data as in Figure 2H, but here the dots represent individual MUs color coded by penetration number.
Figure 2—figure supplement 2
Mean-matched Fano-factor analysis.
(A) SG layers. Left: The four left columns show population-averaged PSTHs of mean firing rate and bootstrapped standard error (top two panels) and of Fano-factor and bootstrapped standard error (bottom two panels) for two different stimulus diameters (from left to right: 0.1–0.2° vs. receptive field [RF] diameter; RF diameter vs. 2xRF diameter, RF diameter vs. 4xRF diameter, and RF diameter vs. 26°), after mean firing-rate matching. Right: The four right columns show population mean firing rate ±99% bootstrapped confidence interval (CI) (top) and mean Fano-factor ±99% bootstrapped CI (bottom) for the same two stimulus diameter comparisons, as indicated, after mean firing-rate matching. (B, C) Same as in (A) but for G and IG layer MUs, respectively.
Figure 2—figure supplement 3
Results for spike-sorted single units.
(A) Representative SG and IG SUs. Top and middle: Raster plots measured at two stimulus diameters, 0.8° (top) and 26° (middle). Bottom: Mean firing rate (black) and Fano-factor (red) as a function of stimulus diameter. Solid lines: Fits to the data. Dashed lines: Baseline Fano-factor (red) and firing rate (black), measured prior to stimulus onset. The error bars are s.e.m. for the firing rate and bootstrapped s.e. for the Fano-factor. (B) Left: Mean normalized (to the receptive field [RF] diameter) Fano-factor for several stimulus diameters, as indicated, shown separately for SG and IG layers. Error bars: s.d. of the bootstrapped distribution. Right: Box plots of normalized Fano-factor for SUs in SG and IG layers at the same six stimulus diameters as in the left panel. Dots: Fano-factor values for individual SUs; red horizontal lines: medians. (C) Median stimulus diameter at the smallest Fano-factor value (or max quenching), normalized to the RF diameter of the recorded SUs, for different layers. Error bars: s.d. of the bootstrapped distributions. Dots: Individual SUs color-coded by case and penetration number. (D) Median surround stimulus diameter at the highest Fano-factor value, normalized to the RF diameter. Error bars: s.d. of the bootstrapped distributions. Dots: Individual SUs.
Figure 3 with 1 supplement
Amplification of cortical response variability by small visual stimuli.
(A) Three example MUs showing stimulus-evoked increases in firing rate and Fano-factor relative to baseline for small stimuli. Top: Spike rasters. Middle: PSTH of firing rate. Dashed line: Baseline firing rate. Bottom: PSTH of Fano-factor. Dashed line: Baseline Fano-factor. (B) Population-averaged time course of Fano-factor (red) and firing rate (black) in SG (left), G (middle), and IG (right) layers computed at two stimulus diameters (top: 0.1°, bottom: 0.8°). Both the Fano-factor and firing rate were normalized to the pre-stimulus baseline of each unit before averaging. (C) Median stimulus diameter evoking the largest magnitude increase in Fano-factor, normalized to the receptive field (RF) diameter of the recorded MUs, for different layers. Error bars: s.d. of the bootstrapped distributions. Dots here and in (D, E) individual MU data. (D) Median difference in Fano-factor (Fano-factor at the stimulus diameter causing the largest change in Fano-factor minus the baseline Fano-factor) ± s.d. of the bootstrapped distributions at max quenching (gray) and max amplification (white) for different layers. (E) Mean baseline Fano-factor for amplifier (green) and quencher (yellow) MUs. Error bars: s.e.m.
Figure 3—figure supplement 1
Variability amplification is not due to eye movements.
(A) Raster plots in an example penetration (MK374-P2) in which a 0.6° stimulus (lasting 500 ms) caused variability amplification in some units (same penetration and condition as in the left column of Figure 3A). Each panel shows responses of 17 units across the electrode array in a single trial. Red: SG layer units (n = 7); black: G layer units (n = 3); cyan: IG layer units (n = 8). (B) PSTH of Fano-factor for 5 units showing variability amplification in this penetration and stimulus condition. The unit # above the panel corresponds to the unit # on the y-axis of the raster plots. Dashed line: Baseline Fano-factor.
Figure 4
Size tuning of shared variance across V1 layers: example penetration.
(A) Left: Raster plots showing the spike times of four simultaneously recorded SG MUs, across several trials, in response to 0.1° (top), 0.5° (middle), and 26° (bottom) diameter gratings. The responses of all four neurons in a single trial are shown between two consecutive horizontal lines. Horizontal lines separate different trials. Right: Network covariance matrices estimated with a single-factor Factor Analysis for each of the same three different stimulus diameters. The diagonal of the network covariance matrix holds the shared variance for each recorded MU. Bottom: Shared variance as a function of stimulus diameter. The red markers show mean ± s.e.m. of the shared variance computed for the SG MU population recorded in this example penetration (n = 4). The gray curves show the data for the four individual MUs. (B, C) same as in (A), but for G (n = 3) and IG (n = 7) layer MUs, respectively.
Figure 5 with 1 supplement
Size tuning of shared variance across V1 layers.
(A) Top: Size tuning of shared variance (red) and firing rate (black) averaged over the population of SG (left), G (middle), and IG (right) MUs. Solid lines: Fits to the data. Dashed lines: Baseline shared variance (red) and firing rate (black), measured prior to stimulus onset. Error bars are s.d. of the bootstrapped shared variance distribution (red) or s.e.m. (black). Bottom: Mean firing rate (black) and mean shared variance (red) as a function of normalized stimulus diameter, averaged over the population of recorded MUs separately for the different layers (SG, n = 31 units; G, n = 15; IG, n = 36). The stimulus diameter was normalized to the receptive field (RF) diameter. (B) Left: RF-normalized shared variance averaged over the population at six different stimulus diameters (as indicated); shared variance values at specific stimulus sizes were extracted from functions fitted to the size-tuning data (see Methods). Error bars: s.d. of the bootstrapped distribution. Right: Box plots of normalized shared variance for MUs in SG, G, and IG layers at the same six stimulus diameters as in the left panel. Dots: Shared variance values for individual MUs; red horizontal lines: medians. The extent of the box marks the quartiles of the data distribution, and the whiskers mark the width of the entire distribution, save for data points that are considered outliers because they are outside 1.5 times the inter-quartile range. (C) Median percent change in shared variance relative to baseline induced by a stimulus matched in size to the RF diameter for the different layers. Dots here and in (D–F): Individual data points. Error bars: standard deviation of the bootstrapped distribution of medians. Black and white dots indicate data from different animals. (D) Median percent change in shared variance relative to a stimulus matched in size to the RF induced by a stimulus matched in size to twice the RF diameter (i.e. near surround) for the different layers. Other conventions as in panel C. (E) Median percent change in shared variance induced by a stimulus matched in size to the per-neuron surround diameter, relative to shared variance induced by a stimulus with diameter equal to the RF diameter, for the different layers. (F) Median percent change in shared variance induced by a 26° diameter stimulus relative to the shared variance evoked by a stimulus matched to the RF diameter, for different layers. (G) Median stimulus diameter at the smallest shared variance value (i.e. at max quenching), normalized to the RF diameter of the recorded MUs, for different layers. Error bars: 68% bootstrapped confidence interval of the median (can be asymmetric but otherwise corresponds to bootstrapped standard error). Other conventions as in panel (C). (H) Median stimulus diameter at the largest shared variance value in the RF surround (i.e. at max facilitation), normalized to the RF diameter of the recorded MUs, for different layers. By construction of the analysis, the smallest possible value is 1. Conventions as in panel (G).
Figure 5—figure supplement 1
Mean-matched Factor Analysis.
(A) SG layers. Population mean ± s.e.m. firing rate (top) and mean ± s.e.m. shared variance (bottom) for different stimulus diameter comparisons, after mean firing-rate matching. RF: stimulus matching the RF diameter. (B, C) Same as in (A) but for G and IG layer units, respectively.
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Size tuning of neural response variability in laminar circuits of macaque primary visual cortex
eLife 15:e86334.
https://doi.org/10.7554/eLife.86334
