Modeling Face Recognition in the Predictive Coding Framework: A Combined Computational Modeling and Functional Imaging Study - Stage 2

Modeling Face Recognition in the Predictive Coding Framework: A Combined Computational Modeling and Functional Imaging Study by Zaragoza, Niehaus, et al. 2022 (under review)

This is the repository ONLY containing the fMRI raw data and associated code scripts of both the level 1 and level 2 SPM-analyses of the connected publication. Note that this code was run on Windows 10 using MATLAB. For additional links see below:

In-principle-accepted Stage 1 Pre-registration: https://doi.org/10.17605/OSF.IO/A8VU7

In-principle-accepted Stage 2 modeling scripts, behavioral data, and preprint: https://osf.io/tye24/

This is the fMRI Dataset for the FRAPPS study (bids format).

- study consists of two tasks: identity learning task ('FRAPPS') and face localizer task ('localizer')

- defacing of anatomical images was performed with pydeface 2.0.0!!!

- functional data is stored as 4D file including all scans (e.g. sub-01_task-frapps_bold.nii.gz) as well as without the 4 dummy scans (e.g. sub-01_task-frapps_bold-dummy-scans.nii). The version without dummy scans was used for all further analysis.

- Exclusion of participants: due to extensive motion (mriqc FD parameter > 0.3) two subjects (sub-16 and sub-26) were excluded. Additionally, we excluded sub-21 because of low performance in the FRAPPS task (< 60%). We also excluded sub-4, sub-18 and sub-28 from all fMRI analyses assessing the influence of the parametric modulators on brain activation.

- onsets and durations of both tasks are described in files called subID_spm_time.csv (for FRAPPS task) and subID-fp_no_Eyelink_onsets.mat (for the localizer task). The directory is subID/func/onsets_and_durations/. In this folder, there are also the multiple condition files for the first-level analysis with and without parametric modulations. They are called for example: sub-01_multi_cond_VIEW-INDxCONTEXT_pe_VI.mat

- Computational model parameters for the parametric modulation can be found in folder subID/func/model_parameters/

- Notes preprocessing FRAPPS: Slice timing, TA is not relevant and will not be used when slice timings are entered. TA was therefore set to 0, as reference slice TR/2 (1.53/2) was used as timing.

- Notes first level FRAPPS: slice time correction was performed, thus for the first level model the microtime resolution was changed to the number of slices divided by two because of multiband factor 2 (= 24), microtime onset needs to match the chosen reference slice so it was set to 12.

- Notes mean beta calculation FRAPPS: sub-4, sub-18, and sub-28 have extremely high (values around -1000) beta values for the history VI pmod. This is most probably caused by only a very slightly changing history VI parameter.

- Localizer coordinates group level (conjunction) (this is just as a reference, of note for OFA, FFA, and pSTS we used the group coordinates derived from the study Thome et al. 2022, Neuroimage to locate the individual maxima)

lOFA -42 -84 -10 t = 1.82
rOFA 48 -74 -6 t = 3.86

lFFA -42 -46 -18 t = 7.32
rFFA 42 -46 -16 t = 6.80

lpSTS -58 -54 8 t = 5.55
rpSTS 46 -54 14 t = 6.25

lPPA -22 -46 -8 t = 17.57
rPPA 24 -46 -10 t = 16.95

- individual coordinates in the localizer task were detected with an automatic algorithm starting at the group max and jumping to the nearest local max within the spmT_conjunction map thresholded at 0.001 unc. These coordinates were further refined (plausibility check) by visual inspection.