2.4. Image processing / general linear model analysis

All images were first preprocessed using statistical parametric mapping software (SPMv12,11, 12 https://www.fil.ion.ucl.ac.uk/spm) for use with both SPM and fMRIstat. Preprocessing included image realignment, which consisted of co‐registration of fMRI time series data from each session to the mean, and then smoothing with a Gaussian filter using the default SPM value (8 mm FWHM). A high‐pass filter with a cutoff frequency of 1/128 Hz was used in all analyses to mitigate low‐frequency signal components.

The localizer was analyzed first. The visual stimulus was modelled in a standard block design using the default canonical hemodynamic response function and default autoregressive model of SPM AR(1). The global maximum from the positive t‐contrast of the visual stimulus blocks was examined and the spatial location recorded. This location for each individual was used in all subsequent analysis. Any subjects without a global maximum within visual cortex for the localizer were not considered in subsequent analysis.

Analysis was performed both in SPM and fMRIstat13 (https://www.math.mcgill.ca/keith/fmristat/). In both cases, the 6 realignment parameters were used as effects of no interest and the smoothed realigned images were the input data. Separate GLM models were built for each of the 4 sessions (as the data had differing TRs) with 4 separate conditions (i.e., 1 regressor for each presentation frequency) using the timing of individual checkerboard presentations as an event relate design. These were convolved with the canonical hemodynamic response function using the default model within SPM and fMRIstat.