Disentangling perceptual and decision biases.
<p>(<b>A</b>) Both perceptual and decision-related biases can shift the psychometric function horizontally in 2AFC tasks. From a single psychometric curve, it is impossible to know whether the empirically measured shift, (<i>B</i>), was a decision bias, a perceptual bia...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | , , , , , |
| منشور في: |
2025
|
| الموضوعات: | |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| الملخص: | <p>(<b>A</b>) Both perceptual and decision-related biases can shift the psychometric function horizontally in 2AFC tasks. From a single psychometric curve, it is impossible to know whether the empirically measured shift, (<i>B</i>), was a decision bias, a perceptual bias, or a combination of the two. (B) Separating perceptual and decision biases with multiple stimulus conditions in the case of judging object motion during self-motion. Equations show how to compute the perceptual (<i>P</i><sub><i>R</i></sub> and <i>P</i><sub><i>L</i></sub>) and decision (<i>D</i>) biases from the empirically measured biases (<i>B</i><sub><i>L</i></sub>,<i>B</i><sub><i>R</i></sub> and <i>B</i><sub><i>N</i></sub>) in three stimulus conditions with leftward, rightward, and no self-motion, respectively. We assume that the decision-related bias (<i>D</i>) is constant across the three interleaved conditions. (<b>C</b>) The generative model of our Bayesian approach for estimating perceptual and decision biases. The animal’s choice in each trial, <i>C</i>, follows a Bernoulli distribution with denoting the probability of choosing rightward motion relative to the reference. The psychometric curve, , reflects the relationship between and object direction, , as captured by a cumulative Gaussian distribution. The sensory noise <i>S</i> and the empirical bias <i>B</i> influence the slope and shift of the psychometric curve, respectively. Empirical bias for each condition <i>B</i> is determined by the perceptual bias <i>P</i> and decision bias <i>D</i> variables. , denotes a gamma distribution while and denote normal distributions. Note that we also account for lapse rates, though they are omitted from this figure for simplicity. See Method section “<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3002764#sec015" target="_blank">Hierarchical Bayesian model</a>” for further details.</p> |
|---|