Introduction
"When you get hungry enough, you find yourself speaking Spanish pretty well." This quote, attributed to African-American baseball legend Josh Gibson when temporarily plying his trade in 1930s Cuba, vividly captures the notion that motivation is important for cognition. When faced with the prospect of a cognitively demanding task, we ask ourselves not only whether we are able to perform it, but also whether it is worth the effort. In an elegant new study, Kouneiher et al.1 employed functional magnetic resonance imaging (fMRI) to assess how regions of the human lateral and medial prefrontal cortex (PFC) mediate the relationship between cognitive demands and motivational considerations. Their results suggest that motivational incentives engage areas in medial PFC that in turn energize regions of the lateral PFC, which are involved in selecting task-appropriate behavior. Furthermore, both medial and lateral areas appear to obey an anterior-to-posterior (rostro-caudal) functional organization, according to whether representations of cognitive and motivational task parameters stem from temporally distal or proximate cues.
The lateral PFC has been thought to mediate aspects of high-level cognition since the birth of the localization movement in the latter half of the nineteenth century. However, a parcellation of this large cortical territory according to some basic functional organizing principle, such as the type of working memory content2 or the distinction between temporary representation versus manipulation of information3, has proved challenging. Nevertheless, a notion that has garnered empirical support in recent years is that the lateral PFC may be organized along a rostro-caudal gradient of decreasing abstraction of representations4, 5, 6, 7, 8. One well-developed view of this kind is the hierarchical 'cascade model' of cognitive control by Koechlin and colleagues7, 8. Two levels of this hierarchy are central to their new study1: contextual control and episodic control. Contextual control refers to the use of a current cue (context) for selecting task-appropriate behavior. For example, although you would normally proceed to eat your meal once the waiter places it in front of you, contextual control allows you to overcome this urge on noting that the person you are dining with has not yet been served. Episodic control, on the other hand, refers to the use of a temporally distal (past) cue that determines, for an extended period of time (an episode), the way that current stimuli and contextual cues are interpreted. If the person you are dining with had previously announced that they were not going to eat anything, episodic control would allow you to adjust your responses to the food stimulus and contextual cue accordingly (and eat away without qualms).
Thus, both contextual and episodic control signals allow us to transcend habitual stimulus-response associations, but they do so in different temporal frames and they are arranged hierarchically: episodic control affects contextual control, but not vice versa. In a previous study, Koechlin et al.7 reported that this hierarchy of control processes maps onto the functional organization of lateral PFC, in that activity in posterior lateral PFC increased additively with episodic and contextual control signals, whereas activity in more anterior, mid-lateral PFC was driven by episodic control processes alone. In their new experiment, Kouneiher et al.1 assessed neural correlates of contextual and episodic control, but they also introduced a motivational manipulation by varying the amount of monetary reward or penalty at stake in applying these control processes. During fMRI scanning, subjects performed an intricate task that required the selection of responses on the basis of independently varied contextual and episodic cue information. In the vital extension to these manipulations of cognitive demands, half of the task trials were also associated with a motivational cue that signaled the chance to earn or lose money (for a correct or incorrect response, respectively). Furthermore, the nature of these bonus trials differed across blocks of trials: in low-incentive blocks, potential gains and losses were negligible, whereas they were substantial in high-incentive blocks. These manipulations produced contextual motivational cues that signaled a transient increase in stakes (bonus trials) and episodic motivational cues that indicated a sustained high or low stakes environment (high- versus low-incentive blocks). This enabled the authors to identify regions coding for sustained or transient motivational signals and to assess their relation to the lateral PFC regions mediating episodic and contextual control processes.
The results concerning the cognitive control variables replicated previous data7; the mid-lateral PFC was only activated by episodic control, whereas posterior lateral PFC activation increased with both episodic and contextual control demands. Furthermore, effective connectivity analyses indicated that episodic control was associated with top-down excitation of posterior lateral PFC by mid-lateral PFC. The analysis of the motivational factors' effect produced fascinating results; increased episodic motivation was associated with enhanced activity in the dorsal anterior cingulate cortex (dACC) and in the mid-lateral PFC region implicated in episodic control. Increased contextual motivation, on the other hand, activated the pre-supplementary motor area (preSMA) and the posterior lateral PFC region involved in contextual control. Additional effective connectivity analyses suggested that the motivational effects observed in lateral PFC originated in medial regions; episodic motivation enhanced connectivity between dACC and mid-lateral PFC, as well as between the latter and the posterior lateral PFC, and contextual motivation enhanced connectivity between the preSMA and the posterior lateral PFC. No interactive effects of cognitive and motivational cues were observed, suggesting that they affected frontal activity independently. Notably, activity in medial PFC was modulated by factors beyond the motivational manipulations. Activity in the dACC generally increased with performance difficulty in a block-wise fashion (irrespective of trial type) and preSMA activation was generally higher in trials that involved response competition, irrespective of block types.
These results paint an intriguing new picture of the interactions between medial and lateral PFC in supporting goal-directed behavior (Fig. 1). The functional role of the dACC in this framework echoes a classical view of this structure as a conduit for motivational, limbic influences on attentional selection9 and is consistent with the recent proposal that this area computes effort-based cost-benefit analyses, employing past action outcomes to guide current action selection10. The current data extend this notion by providing empirical evidence for how this evaluative process in the dACC may drive the selection of goal-relevant stimulus-response associations in lateral PFC. This motivational view of dACC–lateral PFC interactions differs notably from the influential conflict-monitoring model, which posits that the dACC detects conflict between competing response representations and consequently signals lateral PFC regions to impose top-down biasing processes that help to resolve the conflict11. Kouneiher et al.1 argue that, instead of detecting conflict, the dACC codes for anticipated benefits, as reflected by the effect of episodic motivation, and for anticipated cost (in terms of effort), as reflected by the dACC's susceptibility to block-wise performance difficulty in the current task. It should be noted though that tonic dACC activation during episodes of high stakes and high difficulty could, in principle, also be accounted for by non-evaluative functions, such as sustained attentional or autonomic processes12.
Figure 1: The functional architecture of medial and lateral PFC regions suggested by Kouneiher et al.
According to this conception, the dACC retains past cue information indicating costs and benefits associated with behavioral options over an extended period of time (an episode). When a cue initiates a high-stakes episode, the dACC provides sustained input to mid-lateral PFC (midPFC), which represents cognitive demands stemming from past cues and imposes enduring (episodic) top-down biases on the selection of behavior, via its input to the posterior lateral PFC (postPFC). The latter employs concurrent contextual cue information to transiently guide the selection of stimulus-response associations in premotor cortices7. The posterior lateral PFC is also energized by the preSMA, which evaluates costs and benefits stemming from current contextual cues. As a consequence of this motivated cascade of cognitive control, activity in the posterior lateral PFC (and thus, the contextually driven selection of behavior) is modulated by past and current motivational and cognitive cues.
Full size image (45 KB)With respect to the preSMA, Kouneiher et al.1 argue that this region assesses current contextual cues of incentives, as indicated by the effect of contextual motivation, and gauges the costs associated with behavioral selection, as reflected by the preSMA's registering of response conflict in the current study. If potential benefits are deemed to outweigh associated costs, the preSMA energizes contextual control operations in posterior lateral PFC (Fig. 1). This view of preSMA function bears a close resemblance to a recent proposal for dACC function, derived from an extension of the conflict-monitoring model, where response conflict is conceptualized as one component of the costs associated with a current action or strategy that the dACC may consider in evaluating cost-benefit tradeoffs13. Findings from the current study, in combination with other recent data14, raise the possibility that some form of the conflict-monitoring hypothesis may eventually migrate from the dACC to the preSMA. However, the transient activation of the preSMA in response to incentive cues and response competition found in the current study could alternatively be attributed to non-evaluative processes. For example, as the authors duly acknowledge, preSMA activity during response competition might reflect an active role of the preSMA in resolving response conflict, by inhibiting prepotent, but inappropriate, responses15. To conclusively separate the appraisal of the costs and benefits associated with a current stimulus from subsequent processes related to response selection and strategic processing, adjustments will probably require the combination of the type of sophisticated experimental protocol devised by Kouneiher et al.1 with more time-sensitive measurements of neural activity than fMRI.
The study by Kouneiher et al.1 represents an important and exciting attempt at integrating cognitive and motivational determinants of goal-directed behavior. As with all of our current ideas about how the PFC orchestrates highly complex behavior, the details of the model may well turn out to be wrong. However, as with all the best models, it provides a principled and explicit framework that can be readily tested and refined by future studies.
