Specialized color modules in macaque extrastriate cortex

Abstract

Imaging studies are consistent with the existence of brain regions specialized for color, but electrophysiological studies have produced conflicting results. Here we address the neural basis for color, using targeted single-unit recording in alert macaque monkeys, guided by functional magnetic resonance imaging (fMRI) of the same subjects. Distributed within posterior inferior temporal cortex, a large region encompassing V4, PITd, and posterior TEO that some have proposed functions as a single visual complex, we found color-biased fMRI hotspots that we call "globs," each several millimeters wide. Almost all cells located in globs showed strong luminance-invariant color tuning and some shape selectivity. Cells in different globs represented distinct visual field locations, consistent with the coarse retinotopy of this brain region. Cells in "interglob" regions were not color tuned, but were more strongly shape selective. Neither population was direction selective. These results suggest that color perception is mediated by specialized neurons that are clustered within the extrastriate brain.

Figure 1.

FMRI of alert fixating macaque reveals hot-spots, or “globs”, of color-preferring brain activity in posterior inferior temporal cortex, the brain region anterior to V3, consisting of V4, PITd, and posterior TEO. Visual area boundaries were determined using responses to checkerboard stimuli restricted to wedges along the vertical and horizontal meridians (Supplementary Figure 1). Top, a computationally flattened map, the average of all data for this monkey; sulci are indicated by dark grey. Middle, coronal sections showing responses from two independent data sets (40 stimulus runs each; approximate anterior-posterior position indicated at right). Color-preferring regions were identified as those that responded more strongly to equiluminant colored stripes than to achromatic stripes. Labels 1-7 identify prominent globs, and facilitate a comparison between the raw slice data and the computationally manipulated flattened data. Significance depicted by color bar. LGN, lateral geniculate nucleus; s.t.s., superior-temporal sulcus; o.t.s., occipital-temporal sulcus; i.o.s., inferior-occipital sulcus; l.s., lunate sulcus. Bottom, traces show the time course of the fMRI response to achromatic stripes (grey columns); red/blue colored stripes of various red-to-blue luminance ratios (pink columns); and responses to uniform grey (white columns). Area MT shows stronger responses to achromatic stripes than to any colored stripes, and shows a minimum to colored stripes that are approximately equiluminant (color ratios 0 and 0.33). The globs show stronger responses to all colored stripes; and the inter-globs show similar magnitude responses to color and achromatic stripes. Responses were measured using a contrast agent, which results in a negative fMRI signal (traces have been flipped vertically, and de-trended). Scale bar = 1cm.

Figure 2.

FMRI of color-preferring brain activity for a second macaque. See Figure 1 for conventions. fMRI responses to the central 3° enclosed by the green contour. Globs are located both inside and outside the central 3°. Eccentricity maps were determined by measuring responses to gratings (0.29cycles/°) restricted to either the central 3° or the periphery (extending 28°, central 3° gray). These were interleaved with blank periods of neutral gray. Gratings were achromatic in half the blocks and equiluminant in half the blocks, and data from both were averaged.

Figure 3.

Magnetic resonance images showing the location of five (A-E) tungsten micro-electrode recordings, targeting color-preferring (glob) and non-color preferring (inter-glob) regions of alert macaque brain. Electrodes are black, highlighted by vertical white extension lines. Functional activity (response to equiluminant color > response to achromatic) is superimposed. The brain has been computationally sliced in the plane of the electrode: pseudo-frontal sections (left); pseudo-sagittal sections (right). Numbers relate to the globs identified in Figure 1; approximate A-P coordinates given in Figure 1. Supplementary Figure 4 shows electrodes targeting globs and inter-globs in a second animal. l.g., lunate gyrus; M, medial; D, dorsal; A, anterior; other conventions as for Figure 1. Scale bar is 1 cm.

Figure 4.

Glob and inter-glob cells show shape selectivity and lack direction selectivity; inter-glob cells are more strongly shape selective than glob cells. A, length-tuning curves and orientation/direction plots (insets) for three glob cells. B, length-tuning curves and orientation/direction plots (insets) for three inter-glob cells; Supplementary Figures 5&6 show more examples of both. C, quantification of orientation selectivity; D, direction selectivity; and E, bar-length selectivity (0, no selectivity; 1, maximal selectivity). Significant indices (>0.2) are shown in dark grey. Recording positions were confirmed using MRI (Figure 3). Polar plots were generated using bars of optimal length and color, drifted through the receptive field; plots are smoothed with a moving average (3 orientations wide) and normalized to the firing rate elicited by the optimally oriented bar. Direction tuning is indicated in the polar-plot insets by comparing the magnitude of each lobe of the response. Responses to twenty stimuli (10 bar orientations; both directions of motion) were measured. Length-tuning plots were generated using drifting bars of optimal orientation and color, and various lengths. Inter-glob cells were more orientation selective than glob cells (Kolmogorov-Smirnov test, p<10⁻²¹, maximum difference in cumulative fraction, D = 0.41), and had higher length-selectivity indices (KS test, p<10⁻⁵, D = 0.32). Background firing rate indicated by the open symbol in A&B. Standard errors shown.

Figure 5

Color-tuning of a typical glob cell (A), an inter-glob cell (B) and an MT cell (C). Left panels, post-stimulus time histograms to an optimally shaped bar of various colors. Responses were determined to white and black (top two rows in each histogram) and to three sets of 45 colored versions of the bar, one set darker (top section of each histogram), another equiluminant-with-the-background (middle section) and a third set, brighter (bottom section) than the neutral grey background. The colors within each set were assigned a number from 0 to 352 and were equiluminant with each other (numbers are only shown for the bright set of colors). The spectra given to the left of each histogram are schematic (Supplementary Table 1 and Supplementary Figure 8 give the C.I.E. coordinates of the stimuli). For ease of presentation, the responses to each color set have been compressed into 15 rows, each row showing the average response to three consecutive colors in the cycle. Stimulus onset aligned with 0 ms; stimulus duration (step at bottom): 200ms ON/200ms OFF; histogram bins, 1ms. Grey scale bar is average number of spikes per stimulus repeat per bin; the lack of activity immediately following stimulus onset, until about 70-90ms, indicates the response latency—the amount of time required for the signal to be pre-processed by the eyes, lateral geniculate nucleus, V1 etc. Middle panels, show the color-tuning to each of the stimulus sets in polar coordinates. Responses were averaged over 200ms, beginning after the visual latency of the cell. Units are spikes/stimulus repeat. Right panels, weighted-average color response; peak normalized to the maximum response to any color. Asterisk indicates the Rayleigh vector, a standard statistical measure of the asymmetry (i.e. hue tuning) of the polar plot. The polar plots were smoothed with a moving average spanning five colors.

Figure 6.

Single-cells located in globs, but not inter-globs, show strong hue tuning. A, responses of six glob cells: orientation-tuning curve and an icon for the optimal color/orientation (left), post-stimulus time histograms of the responses to a comprehensive set of colors (middle), and each cell’s weighted-average hue-tuning, as a polar plot within a hue circle (right). Orientation-tuning curve units: spikes/bar sweep. Standard Errors shown. See Figure 5 for other conventions. B, histograms and polar plots for six inter-glob cells. Asterisks in the polar plots indicate the Rayleigh vector.

Figure 7.

Quantification of the color tuning of the population of glob cells, inter-glob cells and MT cells. A, The y-axis is a measure of hue tuning (the Rayleigh vector length; 0 is no tuning, 1 is maximal tuning), determined from the weighted-average responses (see Figure 5); the x-axis shows the color-to-achromatic response ratio, a measure of color selectivity, determined as (R_color − R_achromatic)/( R_color + R_achromatic), where R_color is the maximum response to any color and R_achromatic is the stronger of the black or white response. The axes of the marginal distributions are number of cells per bin. Arrows indicate population means. B, The y-axis shows the Rayleigh vector length; the x-axis shows the average correlation coefficient (r²) of the hue tuning, between the three color sets, for each cell; this evaluates luminance-invariance of the color tuning (negative and 0 r² indicate no luminance-invariant hue tuning; 1, maximal luminance-invariance).

Figure 8

Visual-field location of the center of the receptive-fields of the single-units recorded within the different glob regions (numbers refer to globs in Figures 1, 2&3).