Recording receptive field properties across functional maps of cortex. (a) Optical imaging was used to measure the orientation maps in cats and monkeys. An example of a measurement in monkey is shown here. (b) A rapid sequence of gratings of pseudo-random orientation was used to stimulate all cells in the array. This yields a tuning curve for each electrode resulting from the stimulus triggered averages of the responses (solid dots). The tuning curve is then fit with a Gaussian (red curve) to yield an estimate of preferred orientation (θ₀) and tuning width (Δθ). (c) The location of the array is estimated by finding the placement that produces maximal agreement between the optical and electrophysiological estimates of preferred orientation (). The scatterplot shows the best correlation between these two variables in one of the experiments. The estimated location of the array is shown by the black dots in (a). (d) Detail of the orientation preference map, with the local homogeneity index computed for two pixels: one in an iso-orientation domain (where the index is 0.6) and one near a pinwheel center (where the index is 0.1). (e) A map of the local homogeneity index for the same patch of cortex. (f) Isolation of single units. In the scatter plot each dot represents a waveform that has crossed threshold for the example electrode. The x-axis and y-axis are the first and second principal coefficients of the waveforms. The blue cloud near the origin of the graph (black dot) is the noise and the red dots are spikes. The corresponding average waveforms for each cloud, +/− 1 standard deviation, are shown on top.

Sharpness of tuning as a function of map location. Examples of orientation tuning curves and their location within the orientation map are shown for a monkey (a) and a cat (b) All cells obtained in these experiments are shown. They are ranked in order of increasing values of the local homogeneity index (shown near each local orientation map). It appears evident that tuning width decreases with increasing values of the local homogeneity index.

Correlation between tuning width and local homogeneity index. Scatterplot of orientation tuning width and local homogeneity index in two monkeys (a) and two cats (b). In both panels, the solid and open symbols correspond to data from two different animals. Correlation between tuning width and local map homogeneity was statistically significant on all individual experiments (p < 0.01 in all cases). Typical orientation maps in cat and monkey illustrate the difference in spatial scale between two species (scale bar = 1mm). (c) A single relationship accounts for data in both species. Data for the two species are plotted together, showing that the points fall on a single line. (d) Marginal distributions of the local homogeneity index for the neurons recorded in the two monkeys (blue, n = 52) and in the two cats (red, n = 46). The mean local homogeneity index is higher in cats (0.64) than in monkeys (0.32), leading to significantly different distributions (rank sum, p < 10⁻¹¹). (e) There is also a pronounced difference between the two species in the orientation selectivity of single cells. The mean sigma parameter of a fitted Gaussian function is higher in monkeys (24.0 deg) than cats (14.6 deg), again leading to a significant difference between the two species (rank sum, p < 10⁻⁶).

Small errors in the localization of the electrodes would obscure the relationship between tuning width and local homogeneity index. For each of our experiments we ran 200 simulations by independently perturbing each of the estimated electrode locations with Gaussian noise and re-computed the correlation coefficient. (a) Correlation coefficient as a function of the standard deviation of the Gaussian noise. (b) The corresponding statistical significance (p-value). The solid dots at zero represent the values achieved by our data without any additional noise. For the monkey data (blue), an artificial displacement of ∼80 μm (blue arrow) is sufficient to obscure the relationship between tuning width and local homogeneity index, as statistical significance rises above the 0.01 level. For the cat data (red), the relationship is lost at a displacement of ∼115 μm (red arrow). Error bars = SEM.