Color tuning of neurons in face patches of macaque inferior temporal cortex

Previous work has shown that under viewing conditions that break retinal mechanisms for color, one class of objects appears paradoxically colored: faces, and they look green. Interpreted within a Bayesian-observer framework, this observation makes the surprising prediction that face-selective neurons are sensitive to color and weakly biased for colors that elicit L>M cone activity (warm colors). We tested this hypothesis by measuring color-tuning responses of face-selective cells in alert macaque monkey, using fMRI-guided microelectrode recording of the middle and anterior face patches and carefully color-calibrated stimuli. The population of face-selective neurons showed significant color tuning when assessed using images that preserved the luminance contrast relationships of the original face photographs. A Fourier analysis of the color-tuning responses uncovered two components. The first harmonic was biased towards the L>M colors, consistent with the prediction. Interestingly, the second harmonic aligned with the S-cone cardinal axis, which may relate to the computation of animacy by IT cells.SignificanceThe results provide the first quantitative measurements of the color tuning properties of face-selective neurons. The results provide insight into the neural mechanisms that could support the role of color in face perception.

The multifocal pattern electroretinogram (mfPERG) and cone-isolating stimuli

The number of L cones in the retina normally exceeds that of the M cones. Because normal color vision does not depend on the ratio of L- and M-photoreceptors, their signals must undergo an alteration in gain before being analyzed in the cortex. Previous studies have shown that this gain must take place before the cortex, but after the bipolar/amacrine cell layer of the retina. The aim of this study was to obtain topographical information about L- and M-cone activity at the ganglion cell layer using multifocal pattern electroretinography (mfPERG). A standard (black and white) stimulus was used, as well as stimuli modulating only the long wavelength-sensitive (L) or only the middle wavelength-sensitive (M) cones. The L:M ratio was calculated from the amplitude of the L-cone isolating mfPERG to that of the M-cone isolating mfPERG of 10 trichromats. Both the positive and negative components of the waveform were analyzed. Additional recordings of single cone modulated mfERGs were obtained from nine of the 10 subjects. We also recorded from one protanope and one deuteranope. The L:M cone amplitude ratios for both deflections of the mfPERG in the trichromats were around unity (medians 1.18 and 1.16, respectively) for the central 8° of retina. In the peripheral retina between 12.8° and 26°, this ratio increased to 1.42 for the positive component, and 1.37 for the negative component. The median L:M cone amplitude ratios for the mfPERG were higher and ranged between 1.00–2.78 in the central 8° and 1.29–2.78 in the periphery. The results indicate that a major gain adjustment of the retinal signals takes place at the ganglion cell level, and that the ratio is higher at eccentric locations than in the central retinal area.