A spatially invariant representation of spatial frequency. Classifiers were trained and tested to distinguish patterns of neural activity evoked by the presentation of a stimulus at different locations on the screen. In panel a, the circles represent possible stimulus presentation locations and one example configuration of train and test locations for classification. The dark green circle represents the location on which the classifier was trained, the dark purple circles indicate the near test locations, the medium purple (dashed) indicates the mid-test locations, and the light purple (dot dashed) indicates the far test locations. The red ring signifies current fixation. The gratings illustrate stimuli at low and high SF. b, The mean classification over time for fixation trials trained and tested at the same location (green line) and trained on one location and tested on near locations (dark purple line), mid locations (medium purple, dashed), and far locations (light purple, dot dashed). The plotted results reflect the average performance at corresponding training and testing timepoints. All decoding results are averaged over locations at the same eccentricity, across left and right fixation, and subsequently, across subjects. Shaded areas depict the standard error of the mean across subjects. Above-chance decoding onsets are plotted with 95% confidence intervals. The Bayes factors (BF) below the plots indicate the timepoints at which there was substantial evidence in favor of the alternative hypothesis, i.e., decoding above-chance (filled circles). Timepoints at which there was not enough evidence or substantial evidence for the null hypothesis, i.e., chance level, are indicated by open circles.