While it is widely recognized that thinking is somehow costly, involving cognitive effort and producing mental fatigue, these costs have alternatively been assumed to exist, treated as the brain's assessment of lost opportunities, or suggested to be metabolic but with implausible biological bases. We present a model of cognitive cost based on the novel idea that the brain senses and plans for longer-term allocation of metabolic resources by purposively conserving brain activity. We identify several distinct ways the brain might control its metabolic output, and show how a control-theoretic model that models decision-making with an energy budget can explain cognitive effort avoidance in terms of an optimal allocation of limited energetic resources. The model accounts for both subject responsiveness to reward and the detrimental effects of hypoglycemia on cognitive function. A critical component of the model is using astrocytic glycogen as a plausible basis for limited energetic reserves. Glycogen acts as an energy buffer that can temporarily support high neural activity beyond the rate supported by blood glucose supply. The published dynamics of glycogen depletion and repletion are consonant with a broad array of phenomena associated with cognitive cost. Our model thus subsumes both the “cost/benefit” and “limited resource” models of cognitive cost while retaining valuable contributions of each. We discuss how the rational control of metabolic resources could underpin the control of attention, working memory, cognitive look ahead, and model-free vs. model-based policy learning. While it is widely recognized that thinking is somehow costly, involving cognitive effort and producing mental fatigue, these costs have alternatively been assumed to exist, treated as the brain's assessment of lost opportunities, or suggested to be metabolic but with implausible biological bases. We present a model of cognitive cost based on the novel idea that the brain senses and plans for longer-term allocation of metabolic resources by purposively conserving brain activity. decision making optimal control glycogen energy cognitive control mental effort pmc-status-qastatus 0 pmc-status-live yes pmc-status-embargo no pmc-status-released yes pmc-prop-open-access yes pmc-prop-olf no pmc-prop-manuscript no pmc-prop-legally-suppressed no pmc-prop-has-pdf yes pmc-prop-has-supplement no pmc-prop-pdf-only no pmc-prop-suppress-copyright no pmc-prop-is-real-version no pmc-prop-is-scanned-article no pmc-prop-preprint no pmc-prop-in-epmc yes pmc-license-ref CC BY 1. As such, it is not surprising that hypoglycemia is known to cause performance decrements in cognitive tasks. Moderate hypoglycemia, such as that arising from fasting, can impair cognitive performance short-term verbal (Martin and Benton, 1999 ) and spatial (Benton and Parker, 1998 ) memory, and the speed of mental computation (Benton and Sargent, 1992 ; Donohoe and Benton, 1999 ; Kennedy and Scholey, 2000 ). We suspect that these differences can be explained by specifics of task design and reward structure, though there is some evidence for an effect of individual differences during mental fatigue (Hoffman et al., 1989 ; Evans et al., 2000 ). See (Feldman and Barshi, 2007 ) for a thorough review on the effects of glucose levels on cognitive function. Subjective mental fatigue and time-on-task can produce similar performance decrements to ( 2003 ) subjected rats to insulin-induced hypoglycemia. As brain glucose approached zero, brain glycogen content (as measured using 13-C NMR) declined gradually and sustained brain activity for 2 h. Glycogen utilization has also been shown to increase by tactile stimulation of rats (Swanson, 1992 ; Swanson et al., 1992 ; Dienel and Cruz, 2006 ). Glycogen stores have been shown to increase during sleep (Swanson, 1992 ), anesthesia and sustained levels of high blood sugar (Nelson et al., 1968 ), and to decrease during sleep depravation (Karadzic and Mrsulja, 1969 ) and one-trial learning (Hertz et al., 1996 ). Glycogen accumulates faster in regions of the brain that have highest synaptic density (Phelps, 1972 ) and has a high concentration in the cerebellum, hippocampus, thalamus, and striatum (Sagar et al., 1987 ). Finally, glycogen does not appear to be a passive reservoir, utilized only when energy need exceeds resources. Glycogen can be synthesized and degraded simultaneously (Brown and Ransom, 2007 ), and glycogen turnover rate increases in the presence of nearby neural activity (Pentreath and Kai-Kai, 1982 ; Swanson et al., 1992 ). On the other hand, increasing blood sugar levels increases instantaneous transfer of glucose from capillaries to astrocytes and neurons, providing energy to support normal levels of firing even in the case of glycogen depletion. Glucose in the blood is treated as an infinitely large resource pool from which resources are extracted at a set rate. In principle, blood glucose could be modeled as finite pool larger than that of astrocytes. With the exception of Brown et al. These tasks share a reliance on the use of executive function and working memory resources. For our purposes, the important finding is that in a specific but useful set of task types, monetary incentives can explicitly increase the utilization of the exact cognitive faculties that are considered cognitively costly. There is currently insufficient data to enable us to identify a specific molecular mechanism for the proposed control. The cognitive benefits of physical exercise may result in part from better metabolic regulation, and cognitive training may produce increases in glycogen similar to the impact of physical training on muscle glycogen. Congruent with this possibility, the time scale of glycogen repletion in the brain is similar to that in muscle also overcompensation afterwards (Matsui et al., 2012 ), and glycogen depletion leads to super-compensation of glycogen levels in astrocytes (Choi et al., 2003 ).