Figure 6.

Spiking is more linear with imaging and predicts its shape better than does gamma LFP. A₁–A₃, Plots of spiking (A₁), imaging (A₂), and gamma LFP (A₃) as functions of stimulus contrast, for the same experiment as in Figure 2. Data points (circles) indicate normalized mean responses ± SEM (vertical lines; N = 31 trials/contrast) here and in B₁ and B₂. Red traces depict the optimally fitted hyperbolic sigmoid response function R(C) of contrast C, of the following form:

(Albrecht and Hamilton, 1982; Cardoso et al., 2012). Note the similarity of the contrast response for imaging and spikes, captured by the similarity of their fit parameters, compared with gamma LFP [the fitted R_MAX values, n, and C₅₀ (the semisaturation constant) are listed in each panel with the discrepant values for gamma LFP highlighted in red]. B₁, B₂, Same data, with imaging plotted against spiking (B₁) and gamma LFP (B₂). Linear regression R² values and equations are indicated in each panel. C₁, C₂, Population scatter plots for imaging versus spiking (C₁) and gamma LFP (C₂), combining normalized responses as in B₁ and B₂, respectively, across all experiments [N = 261 points for Monkey S (blue circles); N = 109 points for Monkey T (green circles)]. Gray lines show linear regressions for each individual experiment (N = 65). D₁, D₂, Shape comparison between measured imaging responses and those predicted from spikes (D₁) and gamma LFP (D₂) for the same experiment as in A. Traces are shown with mean subtracted and normalized by SD to make their shapes easier to compare. Spike predictions matched the measured imaging shape better than gamma LFP predictions, as quantified by the mean Pearson's r values (increasing order of stimulus contrast: r = 0.88, 0.93, 0.94, 0.98, 0.97, 0.94 for spikes vs imaging; r = 0.52, 0.71, 0.88, 0.95, 0.96, 0.97 for gamma LFP vs imaging), giving mean values of 0.94 and. 0.83, respectively, for spiking and gamma LFP.