Visual Temporal Contrast Sensitivity in the Behaving Mouse Shares Fundamental Properties with Human Psychophysics

Values of d´ were stable during daily sessions. A, Hit vs FA pairs for four WT mice measured on different days clustered about the ROC curves implied by the pooled value of d´. Stimulus conditions (flicker frequency, contrast, illuminance at the cornea, and retinal illumination) for WT56, WT57, and WT59 were as follows: 12 Hz, 50%, 3000 ph/s/μm²; for WT69, 21 Hz, 40%, 8200 ph/s/μm². The values of d´ for each mouse were: WT57, 0.65; WT59, 1.18; WT56, 1.32; and WT69, 1.63, respectively. For comparison, ROC curves for d´ = 0 (main diagonal), 1, 2, and 3 are also shown (gray ROC curves, rising in order away from the main diagonal). The negative diagonal (dashed line) indicates unbiased behavior. Open symbols, Hit vs FA pairs every 100 trials; closed symbols, daily totals (400 trials). B, Values of d´ for the corresponding 100 trial blocks (estimated with Eq. 1; Materials and Methods). C, Changes in response bias expressed in terms of criterion location (Eq. 2; Materials and Methods). Note that for clarity purposes some of the symbols have been displaced laterally.

No significant long-term trends in d´. A, d´ values varied from day to day about a steady mean. d´ values of two C57BL/6J mice (WT) and two mice GNAT2 ^cpfl3 (G2) over the course of 10–12 consecutive days. The daily experimental sessions consisted of 100 trials in response to sinusoidal flicker stimulation. Individual contrast levels were adjusted during a pretrial run to elicit a d´ = 2 and were kept constant thereafter. Continuous lines represent linear regression fits (see text). The histograms at the right of each plot show the distribution of average d´ values estimated by sampling with replacement (n = 400 bootstrap samples) of the data in the respective plots. Flicker parameters were 21 Hz presented on a background illuminance level producing an estimated retinal irradiance of 8200 ph/s/μm². B, Corresponding values of CL. Continuous lines represent the zero (unbiased behavior) axis.

d´ is a bias-free measure of temporal contrast sensitivity in mice. A, ROCs implied by d´ can predict empirical data. Empirical ROC curves of three mice (two WT mice and one GNAT2 ^cpfl3 mouse labeled WTY, WTR, and G2Y, respectively) determined by varying the reward schedules. Mice were first tested with 100%:100% reward schedules, and successively with 100%:70% and 100%:50% reward schedules, or alternatively, with 70%:100% and 50%:100% reward schedules. Open symbols indicate daily pHit and pFA pair values, closed symbols represent the average ± SEM, n = 5-6 trials. The data are well fit by the model: d´ = z(pHit) − z(pFA), where z(pHit) and z(pFA) are the z score values (in SD units) of pHits and pFA. Model predictions are illustrated by the continuous ROCs with maximum likelihood predicted d´ values of 1.66 (WTY), 2.49 (WTR), and 1.73 (G2Y; Table 2). For reference, ROCs for d´ = 0, 1, 2, and 3 are indicated in gray. B, Same as in A, but plotted on z-coordinates. Note that in these coordinates the ROCs have unitary slopes, which is consistent with the notion that d´ sensitivity estimates do not depend on the location of the decision criteria (Macmillan and Creelman, 2005). C, CL derived from the empirical ROC curves decreases linearly with the logarithm of the reward schedule ratio. Continuous black lines indicate regression fits (R ² > 0.9) with Equation 5, and q values of 1.75 (WTY), 2.4 (WTR), and 1.4 (G2Y).

Linear psychometric functions and determination of empirical thresholds. A, Top, Representative set of PFs for a WT mouse measured on different days. Flicker stimulus parameters were 18 Hz flicker and a mean irradiance level of 3000 ph/s/μm². The contrast range sampled was 35–60% in 5% intervals. Linear regression parameters were cTH = 31%, slope = 0.068 d´/% contrast, R ² = 0.58, p < 0.001. Bottom, Scatter plot of the residuals. Consistent with linearity. we did not observe any bends, thinning, or thickening in the scatter of the residuals. B, Linear fits of the PFs in A. The respective cTh values were determined at the point where the extrapolated lines intersect the line defined by d´ = 1. In this dataset, threshold values ranged from 20% to 35% contrast. C, Family of PFs measured from a WT mouse in response to flickering lights at the indicated frequencies. The mean retinal irradiance was 8200 ph/s/μm². Each PF represents the cumulative average of two to three measurements. Linear regression (R ² > 0.70) was applied to determine the slope and empirical threshold where each PF intersects the line defined by d´ = 1. D, Slopes of PFs do not change with flicker frequency. The plot shows slopes of PFs vs frequency for five different mice at mean background light levels delivering 8200 ph/s/μm². The slope of the PFs remains constant and does not change with frequency (linear regression, p = 0.8). E, Slopes of PFs are constant over the mesopic range. The plot compares the mean values of the PF slopes measured at 8200 (black circles), 440 (red triangles), and 9.2 (green squares) ph/s/μm². Note that mice do not respond to flicker frequencies >24 Hz when exposed to 440 ph/s/μm², or >12 Hz when exposed to 9.4 ph/s/μm². Pairwise comparisons at 3, 6, and 12 Hz reveal no significant differences in mean slope values (two-way RM ANOVA, Holm–Sidak method, α = 0.05). Symbols represent the average ± SEM (n = 5). Note that for clarity purposes some of the symbols have been displaced laterally. F, cTh_SEM values estimated from cumulative averages of two (circles) or three (triangles) PFs using bootstrapping techniques and plotted as a function of cumulative R ². Box plot indicates distribution quartiles for cumulative R ² > 0.70 (indicated with gray arrow in plot).

Contrast thresholds decrease with mean irradiance. A–C, Empirical thresholds of WT mice measured as a function of flicker frequency determined at three different background illumination levels, dim (9.2 ph/s/μm² at the retina; A), intermediate (440 ph/s/μm²; B), and bright (8200 ph/s/μm²; C). The contrast thresholds for each mouse were determined from families of psychometric functions, as defined in Figure 5C. Note that for clarity purposes some of the symbols have been displaced laterally. D–F, Corresponding TCSFs estimated from the contrast threshold functions in A–C.