fMRI analysis

1st level analyses were performed in SPM12 using linear regression at each voxel, using generalized least squares with a global approximate AR(1) autocorrelation model. The following contrasts were modelled: happy vs neutral, angry vs neutral, happy vs angry, happy versus implicit baseline (IB), i.e. blank screen, fixation cross etc., angry vs IB, neutral vs IB, happy and angry vs IB, and happy, neutral, and angry vs IB, contrasts including an implicit baseline were ad hoc analyses to investigate possible effects related to viewing of faces in general.

2nd level analyses were performed in SPM12 using linear regression at each voxel, using generalized least squares with a global repeated measures correlation model. To confirm the presence of expected main effects from the experimental paradigm, we first performed whole-brain analyses.

The main analyses focused on regions of interest (ROI:s) specified in the study preregistration. Amygdalae were defined using the Jülich atlas⁵⁹. Fusiform face area ROI:s were defined based on Henson and Mouchlinaitis⁶⁰ as spheres with 6 mm diameter centered on top coordinates. All ROI:s can be visualised and downloaded from https://neurovault.org/collections/RLWUZRQN/. Mean contrast estimates were determined for each region of interest and contrast of interest [happy vs neutral, angry vs neutral, happy vs angry, happy versus implicit baseline (IB), i.e. blank screen, fixation cross etc., angry vs IB, neutral vs IB, happy and angry vs IB, and happy, neutral, and angry vs IB] and were analysed using mixed-effects models in R, see below.

As secondary analyses, we investigated whole-brain effects of sleep deprivation and age group. We conducted a paired t test comparing the full sleep and sleep deprived conditions across both age groups and a two sample t tests to compare younger and older subjects.

Connectivity was investigated through a psycho-physiological interaction (PPI) approach. Seeds were based in bilateral amygdalae on the top coordinates from the contrast happy, neutral, and angry > baseline in the full sleep condition, with a 6 mm sphere around. Time courses for the BOLD signal were extracted from these seeds and entered into PPI models with the contrast happy, neutral, and angry > baseline and the PPI interaction term at first level. At second level one sample t tests were performed to investigate any change in connectivity related to viewing faces for the two sleep conditions separately. No regions displayed a significant increase in connectivity from amygdala in response to seeing faces and therefore no further analyses of connectivity were performed, since the paradigm was not judged sensitive enough to capture changes in connectivity.