Ruby S. Bouhassira¹, Alan A. Stocker², Geoffrey K. Aguirre¹; ¹Department of Neurology, University of Pennsylvania, ²Department of Psychology, University of Pennsylvania

Temporal sensitivity functions recorded from ganglion cells and post-retinal sites differ for achromatic and chromatic modulations, and at high and low light levels. Visual experience statistics, however, suggest that the 1/f^2 distribution of wide-field temporal variation has the same form for achromatic and chromatic modulations (JH van Hateren et al 2002 J Neurosci; data not available across light levels). Efficient coding models therefore predict similar encoding fidelity across frequency for achromatic and chromatic modulations, despite differences in neural tuning. We tested this idea in a delayed flicker frequency estimation task, using luminance and L–M modulations against high and low photopic backgrounds. METHODS: On each of 1200 trials five participants were shown a 2s “reference” flicker (30° wide, uniform-field, bimodal sinusoidal temporal flicker, 1–32Hz uniform log random sampling). After a 2s delay, the participant adjusted the frequency of a “test” flicker to match their memory of the reference (starting point randomly selected within ±7dB of the reference). Positive feedback was provided for responses within ±1.5dB of the reference. Blocks of trials within a session alternated between LMS (33% contrast) and L–M (7.5%) modulations, and the six sessions alternated between a high (3400cd/m^2) and low (2cd/m^2) luminance background. RESULTS: Variance in estimation responses was a constant proportion of the reference frequency (except for low luminance L–M >20Hz), and did not differ with light level or post-receptoral direction (1.6dB). Responses were biased across frequency, with a negative slope and an intercept at ~2 Hz (i.e., rapid flicker was reported as slower); this bias did not differ with stimulus condition. These results are consistent with a Bayesian observer model constrained by efficient coding, reflecting a common 1/f^2 temporal frequency prior across our four stimulus conditions. Our results demonstrate a distinction between neural tuning functions and encoding for temporal variation.

Acknowledgements: R01EY036255