Contribution of color in natural images to recognition performance and neural activity in extrastriate and prefrontal cortex

Although objects can be identified based soley on the information provided by the spatial structure of an image, color adds another perceptual dimension which may facilitate object identification. However, the addition of color in form of colored noise may interfere with object identification. Here, we ask whether color in natural images is associated with changes in visual memory performance. Furthermore, we address whether behavioral changes are correlated with alterations in neural activity simultaneously recorded from extrastriate (V4) and prefrontal (PF) cortex. We used a procedure based on Fourier analysis to degrade colored and achromatic natural images with increasing amounts of achromatic noise. We also degraded achromatic images with chromatic noise. At a given degradation level, the difference between colored and achromatic images was thus provided only by color: the remaining image specific color or the colored noise. In a delayed matching to sample paradigm a sample stimulus (250ms) at various degradation levels was presented, followed by an undegraded probe stimulus (1s) after a delay period (1500ms). A lever press was required if the sample stimulus matched the probe. We have preliminary results from one monkey. The monkey’s recognition performance decreased as a function of noise level for all conditions. In addition, we found that the recognition performance was best for the ‘natural’ color condition and worst for the colored noise condition at the same degradation level (paired t-tests, p<0.01, N = 14). These results suggest that color, independent of spatial composition, confers either an advantage or impairment in object identification depending on whether the color is related to the object or not. Single neuron responses were recorded from a total of 84 neurons in V4 and 76 neurons in PF. We found stimulus selectivity for sample-evoked activity (31 V4, 27 PF) as well as during the delay period for both areas (11 V4, 36 PF; ANOVAs, p