Entrainment to video displays in primary visual cortex of macaque and humans

Abstract

Cathode ray tubes (CRTs) display images refreshed at high frequency, and the temporal waveform of each pixel is a luminance impulse only a few milliseconds long. Although humans are perceptually oblivious to this flicker, we show in V1 in macaque monkeys and in humans that extracellularly recorded action potentials (spikes) and visual-evoked potentials (VEPs) align with the video impulses, particularly when high-contrast stimuli are viewed. Of 91 single units analyzed in macaque with a 60 Hz video refresh, 29 cells (32%) significantly locked their firing to a uniform luminance display, but their number increased to 75 (82%) when high-contrast stimuli were shown. Of 92 cells exposed to a 100 Hz refresh, 21 (23%) significantly phase locked to high-contrast stimuli. Phase locking occurred in both input and output layers of V1 for simple and complex cells, regardless of preferred temporal frequency. VEPs recorded in humans showed significant phase locking to the video refresh in all seven observers. Like the monkey neurons, human VEPs more typically phase locked to stimuli containing spatial contrast than to spatially uniform stimuli. Phase locking decreased when the refresh rate was increased. Thus in humans and macaques phase locking to the high strobe frequency of a CRT is enhanced by a salient spatial pattern, although the perceptual impact is uncertain. We note that a billion people worldwide manage to watch TV without obvious distortion of their visual perception despite extraordinary phase locking of their V1s to a 50 or 60 Hz signal.

Figure 8.

VEP amplitudes while an observer viewed either a contrast-reversing checker-board on a CRT (black lines) or a control stimulus (gray lines). The control throughout the VEP measurements consisted of the CRT displaying the mean luminance but covered by a piece of uniformly lit posterboard. A, Low-frequency portion of the VEP. Robust responses are visible at the 2 Hz reversal frequency and at the 4 Hz second harmonic. B, Higher in the spectrum, a large-amplitude line at the 57.47 Hz frame rate is clearly present in the checkerboard response, but not in the control. At the far right, the 60 Hz power line interference is visible in both traces. C, Raw VEP wrapped on the 17.4 msec period of the frame refresh. The four 20 sec repeats were averaged separately. The response to the checkerboard is clearly locked to a single phase in the cycle; the control stimulus is uniform across the cycle.

Figure 1.

Phase locking of spikes to a blank (uniform luminance) CRT.A, CRT run with a 60 Hz refresh rate. B, CRT run with a 135 Hz refresh rate. The peaky temporal waveform of the screen luminance (top row) causes the ordering of spikes into clusters locked to those impulses for the 60 Hz refresh, but not for the 135 Hz refresh (raster plots, second row; histograms, third row). Despite this temporal structuring the mean rate remains steady at ∼2-3 spikes/sec. The power spectral density histograms (bottom row, y-axis in arbitrary units) indicate a robust line present in the spectrum at the refresh rate when the refresh rate is 60 Hz but absent when the refresh rate is 135 Hz. When the refresh rate is 135 Hz, there is no 60 Hz line in the spectrum, indicating the absence of electrical power line artifacts; there is also no 135 Hz line, demonstrating the absence of electrical interference linked to the frame rate.

Figure 2.

Contrast dependence of 60 Hz phase locking in a complex cell. A, Spike rasters. Spike trains are shown wrapped on the 200 msec period of the 5 Hz drifting grating stimulus, yielding 40 cycles from two 4 sec repeats. The stratification of spikes into columns 16.7 msec apart sharpens with increasing spatial contrast in successive rows. B, Cumulative histograms of the same spikes as shown in A, averaged in 1 msec bins across the 40 cycles at each contrast. The power ratio, which is mean 60 Hz power divided by mean power in the 50-70 Hz band, is shown with each contrast. C, Mean spike rate (open circles) and the amplitude of the 60 Hz component of the spike train (filled circles) at each contrast. D, Amplitude spectra of the spike responses to the three highest contrasts (39,59,90%) are pooled for this graph. The dominance of the line at 60 Hz is obvious. Note in B and C that the power ratio increases from 16.4 at 39% contrast to 37.1 at 90% contrast even though the spike rates at these two contrasts are equal. Cell mt8 u025_005.

Figure 3.

Contrast dependence of 60 Hz phase locking in a simple cell. Data are organized as in Figure 2. Spike trains in A are shown wrapped on the 133 msec stimulus period (7.5 Hz) and show the 60 cycles from two 4 sec repeats. Phase locking to the blank screen (0% spatial contrast) occurs throughout the entire duration of the blank. However, with increasing contrast, the phase locking becomes increasingly restricted to the preferred half-cycle of the simple cell. In C, the open circles show the amplitude of the spike response at the stimulus first harmonic (7.5 Hz), not the DC component as in Figure 2. Lines in D are obvious at the 7.5 Hz stimulus frequency and at the 60 Hz video refresh rate as well as at intermodulation frequencies of 60-7.5 Hz, 60-15 Hz, and 60+7.5 Hz. A small stimulus second harmonic is also visible at 15 Hz. Cell mu1 u010_004.

Figure 4.

Relationship between firing rate and power ratio for 91 V1 neurons exposed to a 60 Hz CRT refresh rate. A, Blank screen. B, Optimal drifting grating. Power ratios greater than the dashed lines are expected by chance <1% of the time (see Materials and Methods). For both the blank screen and the drifting grating, a correlation between power ratio and mean firing rate is evident. However, note that significant phase locking to the video refresh occurs at low spike rates for many neurons and also that there are cells with high firing rates that are poorly entrained to the video refresh. Simple and complex cells show comparable power ratio distributions. In A, 13 cells, all with power ratio = 1 and mean firing rates <0.05 spikes/sec, are not shown. The circled points correspond to the example complex cell from Figure 2 and the simple cell from Figure 3.

Figure 5.

Laminar distribution is shown for 84 of the 91 V1 cells and their power ratios. Seven cells from Figure 4 could not be assigned a relative cortical depth confidently. A, Blank screen. B, Optimal drifting grating. For each cell within a given layer, the fractional cortical depth is shown on a map of standard layer thicknesses (Hawken et al., 1988). Power ratios greater than the dashed lines are expected by chance <1% of the time (see Materials and Methods). Phase locking to a blank screen is found in all layers, although the percentage of cells significantly entrained in layer 4Cα is twice as high as in any other layer. With a high-contrast drifting grating stimulus, the percentage of cells that is phase locked to the video raster is high in every layer. The circled points correspond to the example cells in Figures 2 and 3.

Figure 6.

Phase locking to a 100 Hz video refresh. Top, The spike rasters are shown wrapped on the 100 msec period of the stimulus, which consisted of three 2 sec repeats of a drifting grating at 96% contrast. Bottom, Spikes are averaged in a histogram with 1 msec bins. The power ratio (mean 100 Hz power divided by mean power in the 90-110 Hz band) for the response to this contrast is 18.1. Spikes from the four highest contrasts were combined for the population study, yielding a power ratio of 13.4 for this neuron. Cell mz1 u013_003.

Figure 7.

Laminar distribution of 92 V1 cells and power ratio for a 100 Hz video refresh. Format is the same as in Figure 5. Only one cell entrained its firing to a blank screen (data not shown), but 21 of the 92 cells (23%) significantly phase locked to the 100 Hz refresh when stimulated with optimal high-contrast drifting gratings. The percentage of cells significantly entrained in layer 4B is higher than in any other layer. The circled point marks the cell shown in Figure 6.

Figure 9.

Amplitude and phase of an observer's VEP response to all stimuli at a 57 Hz frame rate. All independent estimates of the Fourier components at the frame rate are shown in gray, and their vector averages are shown in black. The amplitude of the vector average is listed under each stimulus name. The control stimulus produced low-amplitude responses with randomly scattered phases, yielding a vector average near zero. All other stimuli generated responses that clustered in-phase. The checkerboards, which contained high spatial contrast, produced the largest amplitude responses, whereas the drifting grating and uniform luminance stimuli produced responses of intermediate amplitudes. Phase locking in the luminance and checkerboard stimuli was unaffected by whether they were stationary or modulated in time.

Figure 10.

Phase locking of all seven subjects to stimuli displayed at 57 Hz. Symbols mark the vector average amplitude of the 57 Hz line in the VEP spectrum; bars indicate the 99% confidence intervals around the average, as described in Materials and Methods. All subjects' responses to the control stimulus are statistically indistinguishable from zero, as expected of Gaussian randomnoise. Stationary and contrast-reversing checkerboards generated strong phase locking to the video refresh in all subjects, whereas responses to the unpatterned luminance stimuli were generally weaker.

Figure 11.

Amplitude and phase of the VEP response for the same subject shown in Figure 9 for frame rates of 72 Hz (A) and 90 Hz (B). Gray arrows are independent estimates of the Fourier components at the frame refresh rate; black arrows indicate their averages. The amplitude of the vector average is listed under each stimulus name. At 72 Hz the phase locking is apparent in the drifting grating, the checkerboard stimuli, and also in the modulated luminance stimulus. At 90 Hz, none of the stimuli exhibit phase locking.

Figure 12.

Phase of response to 60 Hz frame rate. For each cell studied with a 60 Hz video refresh, the spectral component at 60 Hz was vector averaged across all trials that were used previously in calculating the power ratio (for example, Figs. 4, 5), producing a single average response phase for each cell. The distribution of these phases is shown in the histogram, and a bias toward one half of the 60 Hz cycle is evident.