Frequency analysis showed that Y cell responses
to interference patterns contain power at the envelope TF and its second and third harmonics. For some of the presented stimuli, either one or two of the three component gratings also drifted at these frequencies. For instance, when the carrier is held static as it was in previous studies (Demb et al., 2001b and Rosenberg et al., 2010) two of the components drift at the envelope TF. This overlap of frequency content is a confounding factor that can be eliminated by only considering responses to interference patterns if the three component TFs are different than the envelope TF and its Everolimus research buy second and third harmonics. Across our data set, a total of 124 responses to interference patterns matching this criterion were recorded from 24 Y cells. To further examine if these responses were consistent with the output of a demodulating system, the PSTHs were fit with two models: (1) a “linear model”—the sum of three sinusoids at the component TFs and (2) a “demodulated model”—the sum of three sinusoids at the envelope TF and its second and third harmonics (see Experimental Procedures for details). The SF tuning and interference pattern parameter selections for selleck products a Y cell along with an example PSTH with linear and demodulated model fits are shown in Figures 4A and 4B. To compare the quality of the fits and
to classify the responses as either “linear” or “demodulated,” partial correlations were computed between the PSTHs and model fits and then converted into Z scores using
Fisher’s r-to-Z transformation (see Experimental Procedures). This transformation normalizes correlations so that their difference may be used as an index quantifying model performance (Smith et al., 2005). For each cell, the difference between the Z-scored demodulated fit (ZDem) and Z-scored linear fit (ZLin) was taken (ZDem − ZLin) such that a positive value indicates that the demodulated model outperformed the linear model and a negative value indicates that the linear model outperformed the demodulated model. On average, ZDem − ZLin = 9.41 ± 4.15 SD (N = 124 measurements from 24 Y cells), indicating that the demodulated model provided a significantly better description of Y cell responses than the linear model. Of the 124 Y cell measurements, 121 were classified as demodulated, 2 were Quetiapine unclassified, and 1 was classified as linear (Figures 4E and 4F). For comparison, the same analysis was performed on the responses of an A-layer X cell to interference patterns with component SFs within its passband (Figures 4C and 4D). Consistent with X cells performing a linear analysis of the visual scene (Enroth-Cugell and Robson, 1966, Hochstein and Shapley, 1976 and Victor et al., 1977), the X cell responded predominantly at the carrier TF and its responses were classified as linear (ZDem − ZLin = −5.55, average of two measurements).