Total sleep deprivation (TSD) leads to neurobehavioral changes in experimental tasks of alertness, attention, learning, and motor responses. However, results from working memory (WM) studies are more equivocal. WM comprises multiple cognitive processes and the cerebral basis of this differential vulnerability is not known. The current experiment utilized tasks employing parametric manipulations within an event-related functional magnetic resonance imaging (fMRI) design to better understand the cerebral basis for the differential effects of TSD on WM components. Specifically, the current study utilized fMRI with healthy young adults (n = 20, ages 20- 35), who were scanned 12 hours after their habitual wake time and again 36 hours after their habitual wake time in a counterbalanced design. For an attention component, results demonstrated activation in neural regions implicated in selective visual attention and frontal top-down control of the attentional system when participants were attending to the stimuli. For a rehearsal span component, results demonstrated activation in neural regions subserving rehearsal and episodic encoding when maintaining information in WM. The attention component parameter estimates from a mathematical model correlated with selective visual attention brain regions; whereas, the rehearsal span component estimates from the mathematical model correlated with brain regions subserving rehearsal processes. Following TSD, there was a decrease in behavioral performance for the attention component coinciding with decreases in activation (relative to when participants were rested) in the selective visual attention system. For the rehearsal span component, there was intact behavioral performance coinciding with increased activation (relative to when participants were rested) in brain areas subserving rehearsal processes. Results from the mathematical model demonstrated a shift from the use of both rehearsal span and episodic encoding when participants were rested to use of rehearsal span during TSD. These data suggest that a mechanism for decreased performance and activation in verbal WM following TSD lies in impairment of the supervisory attentional system. Alternatively, behavioral performance related to the phonological loop and associated neural substrates compensate in the face of TSD. The compensatory mechanism seems to be related to rehearsal processes as participants shifted from using episodic encoding to phonological rehearsal during TSD.