Abstract
Variability is a fundamental feature of human brain activity that is particularly pronounced during development. However, developmental neuroimaging research has only recently begun to move beyond characterizing brain function exclusively in terms of magnitude of neural activation to incorporate estimates of variability. No prior neuroimaging study has done so in the domain of emotion regulation. We investigated how age and affective experiences relate to spatial and temporal variability in neural activity during emotion regulation. In the current study, 70 typically developing youth aged 8-17 years completed a cognitive reappraisal task of emotion regulation while undergoing functional magnetic resonance imaging. Estimates of spatial and temporal variability during regulation were calculated across a network of brain regions, defined a priori, and were then related to age and affective experiences. Results showed that increasing age was associated with reduced spatial and temporal variability in a set of frontoparietal regions (e.g., dorsomedial prefrontal cortex, superior parietal lobule) known to be involved in effortful emotion regulation. In addition, youth who reported less negative affect during regulation had less spatial variability in the ventrolateral prefrontal cortex, which has previously been linked to cognitive reappraisal. We interpret age-related reductions in spatial and temporal variability as implying neural specialization. These results suggest that the development of emotion regulation is undergirded by a process of neural specialization and open up a host of possibilities for incorporating neural variability into the study of emotion regulation development.
Footnotes
Figure 5B was inadvertently occluding Figure 5A. This revision fixes this issue.
1 Some kind of beta-series or timeseries is required for estimation of temporal variability, but most fMRI tasks can be marshalled to fit this prerequisite. Multiple spheres nested within multiple ROIs are required for the multilevel measurement models.
2 This could not be done for analyses in the main document, as spheres would have then substantially overlapped.