Receptive versus perceptive fields from the reverse-correlation viewpoint

Abstract

This brief review article brings together a series of related experiments in psychophysics and physiology that show striking similarities between measurements in human observers and in single neurons. We consider seven pairs of primary research articles, each pair consisting of one paper in physiology and one in psychophysics, and we highlight common features between receptive and perceptive fields obtained using reverse correlation. We conclude by discussing how to assess the validity of perceptive fields as predictors of human responses, and by deriving a novel expression for the maximum trial-by-trial predictability attainable by any model for any psychophysical task.

How inherently noisy is human sensory processing?

“…However, we can ask the following question: Given a certain statistical distribution for internal noise, what is the upper limit on trial-by-trial predictability? In previous work, we have shown that this question can be answered in general terms by establishing a range within which the upper limit for trial-by-trial predictability must lie (Neri, 2009;Neri & Levi, 2006); the exact value within this range depends on the details of the experiment, and it is not possible to determine it without restrictive assumptions, but we can state that it cannot lie outside the specified range in the absence of virtually any assumption at all (Neri & Levi, 2006). More specifically, suppose we run a double-pass experiment and find that a participant gives the same response to repeated presentations on fraction α of the trials (probability of agreement).…”

“…More specifically, suppose we run a double-pass experiment and find that a participant gives the same response to repeated presentations on fraction α of the trials (probability of agreement). We can then state that if we are able to construct the best possible model of the participant's decisional process, our ability to predict the participant's response may be as low as α or as high as {1 1 √ 1 1 n[α(n21) 2 1]}/n (for nAFC), depending on the statistical structure of the external stimuli and the internal noise source (Neri & Levi, 2006). The corresponding range for the probability of agreement we measured across our whole data set was 0.7-0.84: This is (on averlivered through headphones, whereas others came from experiments in which they were asked to discriminate local details of natural scenes displayed on a monitor (see Table 1).…”

How Motion Signals Are Integrated Across Frequencies: Study on Motion Perception and Ocular Following Responses Using Multiple-Slit Stimuli

“…A ceiling effect makes it difficult to estimate the suprathreshold frequency tuning map for motion perception using the proportion of correctly perceived directions for a suprathreshold sinusoidal grating. Therefore we generated stimuli that consist of multiple moving gratings of randomly selected frequencies and applied the reverse correlation technique (Ahumada 1996;Jones and Palmer 1987;Neri and Levi 2006;Ringach and Shapley 2004) to investigate how much each frequency contributed to the observed motion perception. The results showed that the measured tuning maps had peaks at low spatial frequency in both the OFR and motion perception.…”

Psychophysically measured task strategy for disparity discrimination is reflected in V2 neurons

“…Here we explore this possibility by combining measures of choice probability (in area V2) with an objective measure of the two macaque monkeys' strategy: image classification, also termed psychophysical reverse-correlation [10][11][12] . In our disparity discrimination task this method produces psychophysical kernels that quantify how disparities in the stimulus contribute to the subjects' decisions.…”