Arterial spin labeling (ASL) sequences that incorporate multiple postlabeling delay (PLD) times allow estimation of when arterial blood signal arrives within a region of interest. Sequences that account for such variability may improve the reliability of ASL and therefore make the technique well suited for future clinical and experimental investigations of cerebral perfusion. This study assessed the within- and between-session reproducibility of an optimized pseudo-continuous ASL (pCASL) functional magnetic resonance imaging (FMRI) sequence that incorporates multiple postlabeling delays (multi-PLD pCASL). Healthy subjects underwent four identical scans separated by 30 minutes, 1 week, and 1 month using multi-PLD pCASL to image absolute perfusion (cerebral blood flow (CBF) and arterial arrival time (AAT)) during both rest and a visual-cued motor task. We show good test-retest reliability, with strong consistency across subjects and sessions during rest (inter-session within-subject coefficient of variation: gray matter (GM) CBF=6.44%; GM AAT=2.20%). We also report high sensitivity and reproducibility during the functional task, where we show robust task-related decreases in AAT corresponding with regions of increased CBF. Importantly, these results give insight into optimal PLD selection for future investigations using single-PLD ASL to image different brain regions, and highlight the necessity of multi-PLD ASL when imaging perfusion in the whole brain.