Reassessing the Functional Significance of BOLD Variability

BOLD variability (SDBOLD) has emerged as a unique measure of the adaptive properties of neural systems that facilitate fast, stable responding, based on claims that SDBOLD is independent of mean BOLD signal (meanBOLD) and a powerful predictor of behavioural performance. We challenge these two claims. First, the apparent independence of SDBOLD and meanBOLD may reflect the presence of deactivations; we hypothesize that while SDBOLD may not be related to raw meanBOLD it will be linearly related to absolute meanBOLD. Second, the observed relationship between SDBOLD and performance may be an artifact of using fixed-length trials longer than response times. Such designs provide opportunities to toggle between on- and off-task states, and fast responders likely engage in more frequent state-switching, thereby artificially elevating SDBOLD. We hypothesize that SDBOLD will be higher and more strongly related to performance when using such fixed-length trials relative to self-paced trials that terminate upon a response. We test these two hypotheses in an fMRI study using blocks of fixed-length or self-paced trials. Results confirmed both hypotheses: (1) SDBOLD was robustly related with absolute meanBOLD; and (2) toggling between on- and off-task states during fixed-length trials reliably contributed to SDBOLD. Together, these findings suggest that a reappraisal of the functional significance of SDBOLD as a unique marker of cognitive performance is warranted. ### Competing Interest Statement The authors have declared no competing interest.