Kurzban et al.'s provocative and compelling theory of effortful behavior links the psychological literature on self-control with a parallel literature in cognitive neuroscience. Their opportunity cost model suggests that the anterior cingulate cortex (ACC), prefrontal cortex, and other frontal brain areas compose a neural substrate that prioritizes mental actions based on their learned costs and benefits, an assertion that should be uncontroversial given this system's well-known role in high-level decision making (Silvetti & Verguts Reference Silvetti, Verguts and Bright2012). Yet the rapidly evolving literature on cognitive control suggests that aspects of their proposal require further development.

In particular, although the conflict-monitoring theory of ACC function provides much of the neural foundation for Kurzban et al.'s proposal, accruing evidence appears inconsistent with it (Mansouri et al. Reference Mansouri, Tanaka and Buckley2009; Nachev Reference Nachev2011; Rainer Reference Rainer2007; Yeung Reference Yeung, Ochsner and Kosslyn2013). The conflict theory was motivated largely by functional neuroimaging data (Botvinick et al. Reference Botvinick, Cohen and Carter2004), but other neuroimaging findings have been less supportive of the theory (e.g., Erickson et al. Reference Erickson, Milham, Colcombe, Kramer, Banich, Webb and Cohen2004; Roelofs et al. Reference Roelofs, van Turennout and Coles2006). Studies in nonhuman primates have also failed to reveal conflict-related activity in ACC neurons, and ACC damage in monkeys and humans tends to spare conflict processing (Mansouri et al. Reference Mansouri, Tanaka and Buckley2009). By contrast, recent human functional neuroimaging (Dosenbach et al. Reference Dosenbach, Fair, Miezin, Cohen, Wenger, Dosenbach, Fox, Snyder, Vincent, Raichle, Schlaggar and Petersen2007; Hyafil et al. Reference Hyafil, Summerfield and Koechlin2009; Kouneiher et al. Reference Kouneiher, Charron and Koechlin2009), human lesion (Picton et al. Reference Picton, Stuss, Alexander, Shallice, Binns and Gillingham2007), and nonhuman primate (Hayden et al. Reference Hayden, Pearson and Platt2011; Johnston et al. Reference Johnston, Levin, Koval and Everling2007) studies suggest that the ACC is responsible for task initiation and maintenance and for motivating or “energizing” behavior.

We have recently proposed a novel theory of ACC function that seems more amenable to the opportunity cost model (Holroyd & Yeung Reference Holroyd and Yeung2012). This idea links a previous suggestion that the ACC acts as a high-level decision-making mechanism that learns to choose between action plans according to principles of reinforcement learning (Holroyd & Coles Reference Holroyd and Coles2002) with recent advances in reinforcement learning theory that utilize a hierarchical mechanism for action selection called hierarchical reinforcement learning (HRL) (Botvinick Reference Botvinick2012). According to the HRL account, the ACC supports the selection and execution of context-specific sequences of goal-directed behavior, called “options,” over extended periods of time (Holroyd & Yeung Reference Holroyd and Yeung2012). This view holds that the ACC integrates rewards and punishments across time to learn not whether individual actions are worth performing, but rather, whether the task itself is worth carrying out. Thus, the ACC would be responsible for motivating subjects to participate in a psychology experiment until its completion, as opposed to implementing subtle behavioral adjustments along the way.

Options are comparable to mental actions to the extent that both represent extended, task-related activities such as playing a board game, doing math homework, and jogging. Both are also selected (prioritized) based on their learned costs and benefits. Yet the two theories have an important difference: Unlike the opportunity cost theory, the HRL theory does not set the resource-depletion and cost-benefit accounts of effortful behavior in opposition. Recent HRL computational work from our laboratory (unpublished) simulates a “dual system” approach to behavioral regulation (Heatherton & Wagner Reference Heatherton and Wagner2011; Hofmann et al. Reference Hofmann, Friese and Strack2009) whereby “top-down” control is applied by the ACC over a relatively impulsive, basal ganglia mechanism for action selection. Control is maintained via an energy factor that depletes with use (Ackerman Reference Ackerman and Ackerman2011; Van der Linden et al. Reference van der Linden, Frese and Meijman2003) such that optimal task performance is maintained with the minimal level of control necessary (Kool et al. Reference Kool, McGuire, Rosen and Botvinick2010; Yeung & Monsell Reference Yeung and Monsell2003). Contrary to assertions in the target article, our simulations illustrate that – at least in principle – momentary increases in control can occur in the presence of a strictly decreasing resource (Muraven et al. Reference Muraven, Shmueli and Burkley2006; Muraven & Slessareva Reference Muraven and Slessareva2003).

But do mental resources actually exist? The opportunity cost model would seem to invoke separate resource-dependent and resource-independent mechanisms for physical versus mental control, respectively. This distinction may be artificial: Even when actions involve only minimal energetic costs, people still prefer doing nothing over something (Baumeister et al. Reference Baumeister, Bratslavsky, Muraven and Tice1998; Brockner et al. Reference Brockner, Shaw and Rubin1979), and when the costs between actions are equated, they choose actions that minimize control – indicating that mental actions, like physical actions, exact costs (Kool et al. Reference Kool, McGuire, Rosen and Botvinick2010). Doubts about glucose utilization notwithstanding (Schimmack Reference Schimmack2012), mental costs must reflect in part the simple fact that the brain is a biophysical system that obeys thermodynamic laws. For instance, metabolic processing of the neurotransmitter glutamate is a highly energy-consuming process, so synapses operate on a principle of resource optimization that maximizes the current released per glutamate molecule (Savtchenko et al. Reference Savtchenko, Sylantyev and Rusakov2013). A parsimonious theory would posit a unitary mechanism for maintaining control over the task at hand, whether this entails overcoming neural fatigue in a chess marathon or muscle fatigue in a long-distance marathon (Boksem & Tops Reference Boksem and Tops2008).

It has been suggested that the resource-depletion theory originated as an ill-conceived metaphor for the essential role that energy played during 19th-century industrialization (Hockey Reference Hockey and Ackerman2011). Ironically, in this contemporary age of dwindling natural resources, the energy metaphor may be even more apposite than before. Natural resource deposits are finite entities that become increasingly difficult to mine as the easiest resources to develop are extracted first. The decline can be masked with economic incentives that temporarily increase production, but doing so comes at the expense of an ultimately faster depletion rate (Youngquist Reference Youngquist1997). By analogy, studies of resource depletion in humans have typically involved shorter time frames (i.e., minutes) when, presumably, the resource in question is still plentiful and easy to extract (Hagger et al. Reference Hagger, Wood, Stiff and Chatzisarantis2010a). The HRL account suggests that the ACC energizes behavior over extended periods – on the order of hours or longer – rather than on a moment-to-moment basis. Experiments that utilize longer time-horizons may discover that the short-term performance gains resulting from motivational incentives, response conflicts, and so on, come at the expense of longer-term decrements in performance once the resources upon which they draw are ultimately depleted.