Fig 1.
Examples of experiment stimuli and design.
(A) Examples of a beach scene that was jointly contrast normalized across each spatial frequency condition: full spectrum (FS), low spatial frequencies (LSF) and high spatial frequencies (HSF). The four scene categories were beaches, highways, forests, and cities. (B) Examples of the noise stimuli at each of the four spatial frequencies included in the experiment. (C) Sample experiment sequence. Each of the eight runs consisted of 16 blocks of images (blue boxes; one per stimulus condition). Each block contained a sequence of eight different images of that condition.
Fig 2.
Multivariate decoding results.
(A) Decoding of scene categories (4-way classification) is as accurate for HSF as for FS scenes. LSF scenes, on the other hand, cannot be decoded above chance. (B) Dividing the PPA along the A-P axis, reveals similar decoding accuracy for HSF and LSF scenes in posterior PPA, but significantly higher accuracy for HSF than LSF in anterior PPA. (C) Decoding of high versus low spatial frequencies from scene stimuli succeeded in PPA, OPA, and LOC but not the RSC. Decoding of spatial frequency from filtered noise stimuli succeeded only in RSC. (D) Decoding of spatial frequency was more accurate from the posterior than the anterior subdivision of the PPA. This difference was statistically significant for noise images but not for scenes. Error bars represent standard errors of the mean. *p < 0.05, **p < 0.01, ***p < 0.001; N = 10.
Fig 3.
Percent signal change in visual ROIs.
(A) Univariate activation of high-level visual ROIs by spatial frequency-filtered scenes and noise stimuli. (B) HSF scenes activate pPPA more strongly than LSF scenes. This difference does not persist in aPPA. Error bars are standard errors of the mean. *p < 0.05.