| Paper: | PS-1A.1 |
| Session: | Poster Session 1A |
| Location: | Symphony/Overture |
| Session Time: | Thursday, September 6, 16:30 - 18:30 |
| Presentation Time: | Thursday, September 6, 16:30 - 18:30 |
| Presentation: |
Poster
|
| Publication: |
2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania |
| Paper Title: |
Potential cortical and computational biases in representational similarity analysis |
| Manuscript: |
Click here to view manuscript |
| DOI: |
https://doi.org/10.32470/CCN.2018.1127-0 |
| Authors: |
Daniel Leeds, David Shutov, Fordham University, United States |
| Abstract: |
Representational similarity analysis (RSA) has become a valuable and common tool in the understanding of cortical representations across diverse cognitive arenas. However, RSA typically employs assumptions that may bias model comparisons. Our present work identifies common statistics of cortical responses in object perception and finds that these responses may support inflated model comparison results with unusual resistance to noise. Similarly, we find differing constructions of permutation tests alter perceived significance of model-cortical matches. We employ an fMRI voxel searchlight method to compare local cortical responses to sixty objects, with 218 diverse candidate semantic groupings of the same objects. We find semantic properties with the highest cortical correlations are high skew distance matrices, while the lowest cortical correlations are often low skew. We also find additional restrictions on “randomized” permutations may be required for more accurate assessment of statistically significant matches in RSA. |
