An ideal observer model where task-irrelevant sounds compress rat visual perceptual space according to their average intensity accounts for rat perceptual choices.
<p><b><i>A</i></b>. Group average proportion of “High TF” choices (<i>n</i>=10 rats) as a function of the TF of the visual gratings, when the latter were paired with: 1) fixed amplitude sounds (green); 2) AM sounds with increasingly larger TFs (progressively...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | , , , |
| Published: |
2025
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | <p><b><i>A</i></b>. Group average proportion of “High TF” choices (<i>n</i>=10 rats) as a function of the TF of the visual gratings, when the latter were paired with: 1) fixed amplitude sounds (green); 2) AM sounds with increasingly larger TFs (progressively darker shades of gray, matching those in the stimulus matrix shown in <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1013608#pcbi.1013608.g001" target="_blank">Fig 1C</a>); and 3) no sounds (purple). The error bars denote s.e.m. over the group of 10 rats. The curves are group-level predictions of the proportions of “High TF” choices according to the Bayesian ideal observer model defined in <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1013608#pcbi.1013608.e005" target="_blank">Eq (1)</a>, with a single value of the sensitivity parameter <i>σ</i> and 5 different values of the scaling factor <i>γ</i>, one for each level of sound intensity (more precisely, each curve is the mean of the distribution of psychometric curves induced by the posterior distribution of the <i>σ</i> and <i>γ</i> parameters obtained for a given sound intensity level). For this model and the data as presented in the plot, (posterior mean ± st. dev.). The inset shows the difference between the expected log predictive density (ELPD) of the reference model with 1 <i>σ</i> and 5 <i>γ</i> values (i.e., the one yielding the psychometric curves) and variants of the model having a given number of free <i>σ</i> and <i>γ</i> parameters (as indicated on the ordinate axis). In all comparisons, the reference model was the one with lowest ΔELPD, i.e., with the highest predictive power. <b><i>B</i></b>. Relationship between the magnitude of the scaling factor <i>γ</i> in the ideal observer model and the intensity of the sounds that were paired with the visual gratings. Triangle, square and cross symbols refer to individual rats, while circles are group averages. Sound conditions are labeled according to the same color code as in <b><i>A</i></b> and <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1013608#pcbi.1013608.g001" target="_blank">Fig 1C</a>. Vertical error bars denote the std. dev. of the posterior distribution of the group-level <i>γ</i> parameters (see Methods). The horizontal error bars denote s.e.m. over the group of 10 rats.</p> |
|---|