Tuning for Spatiotemporal Frequency and Speed in Directionally Selective Neurons of Macaque Striate Cortex

Speed-tuning and spatiotemporal response fields of a representative V1 simple cell (A–D) and complex cell (E–H). A, B, E, F, Plots of firing rate as a function of stimulus speed; different symbols show data obtained at three different spatial frequencies. A and E show responses at 32% contrast, and B and F show responses at 8% contrast. Curves were obtained by fitting Equations 2 and 3 to the data. C, D, G, H, Contour plots showing spatiotemporal response fields. The intensity at each combination of spatial and temporal frequency indicates the relative response, in which darker regions indicate larger responses. C and G show responses at 32% contrast, and D and H show responses at 8% contrast. The speed-tuning curves at the left were derived from the same data used to create the contour plots on the right. In the speed-tuning curves in A and B, squares, triangles, and circles show data for spatial frequencies of 0.5, 1, and 2 cycles/°. In G and H, squares, triangles and circles show data for spatial frequencies of 1.32, 2.64, and 5.28 cycles/°.

Quantitative comparison of speed-tuning index in V1 simple cells, V1 complex cells, and MT neurons. A, Histograms showing distribution of speed-tuning index for each cell type. Vertical dashed lines indicate values expected for separable responses to spatial and temporal frequency (speed-tuning index of 0) and speed-tuned responses (speed-tuning index of 1). Arrows on each distribution show the means. B, Scatter plot showing effect of contrast on speed-tuning index for V1 neurons. Gray and black symbols show data for simple and complex cells. C, Scatter plot showing effect of contrast on speed-tuning index for MT neurons. In B and C, the oblique line is the unity line.

Contrast gain modulation as a function of spatial and temporal frequency. Each graph is a contour plot showing a measure of response as a function of spatial and temporal frequency. The three columns show data for three complex cells, and the first two rows summarize spatiotemporal response fields for grating stimuli of 32 and 8% contrast. Response has been log transformed and normalized so that the peak response, indicated by the darkest shading, has a value of 1. The bottom row of contour plots show the difference between the high- and low-contrast responses, which is equivalent to contrast gain because the first two rows were log transformed. The horizontal and vertical dashed lines in the bottom row of plots were drawn at the spatial and temporal frequencies that provided the best response for stimuli of 8% contrast.

Quantitative summary of contrast gain as a function of spatial and temporal frequency for our full sample of 22 V1 complex cells. Each histogram shows the distribution of contrast gains for a particular combination of spatial (SF) and temporal (TF) frequencies. The top right and bottom left histograms show data for combinations of high temporal and spatial frequencies and low temporal and spatial frequencies. The top left and bottom right show contrast gains for low–high and high–low combinations of temporal and spatial frequency. The downward arrows indicate the geometric mean value of contrast gain in each spatiotemporal corner.

Responses of V1 complex cells to stimuli consisting of spatially superimposed dual gratings. A, Spatiotemporal response surface for the example complex cell used in this figure. Vertical lines indicate the spatial frequencies chosen for the two dual-grating stimuli. B, C, Firing rate as a function of speed for a representative neuron, for pairs of sine-wave gratings of low (B) or high (C) spatial frequencies (sf). The top row of speed-tuning curves show responses to single sine-wave gratings; each panel shows responses to two spatial frequencies. In the bottom row of speed-tuning curves, the open circles show the actual responses to dual gratings of the spatial frequencies used in the companion graphs in the left column, and the filled circles show the predicted firing based on simply adding the responses to the two gratings presented singly. Curves were obtained by fitting Equation 2 to the data. D, E, Population summaries for V1 (D) and MT (E). The scatter plots show the relationship between the predicted and actual preferred speeds for dual-grating stimuli from speed-tuning curves like those in B and C. Open and filled symbols show data for dual-grating stimuli comprising spatial frequencies above and below the preferred spatial frequency of the neuron under study.

Comparison of speed tuning for sine-wave gratings and dot textures in V1 and MT neurons. A, Response as a function of stimulus speed for a typical V1 neuron. The gray circles and curve show speed tuning for gratings, estimated by averaging responses across iso-speed lines in the spatiotemporal response fields. The black circles and curve show responses to dot textures moving at different speeds. Curves were obtained by fitting Equation 2 to the data. B, Scatter plot comparing tuning width for gratings and dots in V1 neurons; light and dark gray symbols plot data for simple and complex cells. C, Scatter plot comparing tuning width for dots and gratings in MT neurons. The same analysis methods were used to compute the tuning width for gratings for V1 and MT neurons. In B and C, the oblique dashed line is the unity line.

Comparison of speed-tuning preference in V1 and MT neurons. A, Histogram showing the distribution of preferred speeds in V1 neurons. The filled and open bars indicate summarized data from complex and simple cells, respectively. B, Histogram showing the distribution of preferred speeds in MT neurons.