Reductionism in the neurosciences has been incredibly productive, but it has been difficult to reconstruct how high-level behaviors emerge from the myriad biological mechanisms discovered with such reductionistic methods. This is most clearly true in the case of the motor system, which has long been studied as the programming of motor actions (at its least reductionistic). However, as pointed out by Mars et al., the brain is almost constantly enacting motor plans, and so the initiation of actions is more likely preceded by motor reprogramming than anything else. However, that process has only been studied in the domain of response inhibition, in the sense that previously-programmed actions must presumably be inhibited in order for "reprogramming" to take place.
To clarify the difference between the processes involved in motor programming and in motor reprogramming, Mars et al. used a fairly simple task. Basically, subjects saw one of 4 possible shapes, two of which indicated the need for a left button press, and the other two indicated the need for a right button press. However, subjects were instructed not to respond until they heard a tone, which followed the onset of the shapes by 1-5s on 60% of trials. However, in 15% of the trials (neutral trials), the shape did not appear until the tone occurred (a 0s delay). In another 15% of trials (switch trials), the first shape was followed by one or more different shapes; subjects were to respond according to the last shape seen. Thus, subjects learned to prepare their responses before the onset of the tone, although sometimes this wasn't possible (in the 15% of trials where the tone & shape were simultaneous) and sometimes this was actually a wasted effort (in the 15% of trials in which multiple shapes appeared). Subjects completed this task while being scanned with fMRI, having previously practiced these tasks in the following order: 40 standard trials, 40 neutral trials, 192 trials with all three trials types intermixed.
Along with some fancy statistics, this task allows Mars et al to dissociate the neural processes associated with motor programming (in terms of the neural response to the first shape), motor reprogramming (the neural response to the second shape on 15% of trials), and motor execution (the neural response to the tone). For the fMRI geeks, they used an interesting approach (with references to Toni et al.) in which the jitter between trials is not always an integer multiple of the TR, and they controlled for the differences in trial frequency by retaining only those voxels which showed differential effects in the frequency-matched conditions. (I have some gripes with this latter claim, since a voxel might show differential effects in frequency-matched conditions if those voxels are both frequency sensitive and sensitive to other variables.... anyway.)
Posterior parietal & premotor cortices were involved in movement programming, independent of reprogramming and execution (as assessed by the Nichols et al conjunction technique). On the other hand, switch trials - relative to standard trials - differentially activated left & right insula, right inferior frontal gyrus, and right precentral gyrus. Only the latter region survived a contrast designed to control for the "increased attention to action associated with selecting a response under time pressure" (although this contrast also controls for a frequency difference between the trial types). This has particular relevance for theories of response inhibition, in which right ventral prefrontal cortex is thought to be involved in the cancellation of responses. If that accurately characterized the function of right ventral PFC, we would expect right ventral PFC to be more active during switch than neutral trials (in which no response inhibition should occur). Since they are not, it indicates that right ventral PFC may be important for increased attention to responses under time pressure (to use their phrasing) or, more conventionally, in the processing of the infrequent stimulus.
Nonetheless, Mars et al did observe greater activity in the right prefrontal cortex during the switch trials than the neutral trials, indicating that trial frequency alone is not enough to explain these results. However, this region was in the premotor cortex, not the ventral prefrontal cortex, contrasting with some previous work on the locus of response inhibition.
No regions showed more activity in response to the standard than switch trials, and the authors conducted a control analysis to ensure that none of the above differences were due to between-trial response repetitions vs. switches - i.e., a task-switching effect.)
In summary, Mars et al demonstrate that parietal and premotor regions are involved in programming a response, and that distinct areas of parietal cortex are additionally recruited when reprogramming a response. In the prefrontal cortex, only right premotor regions were specifically involved in the reprogramming of a response after controlling for "time pressure" or frequency sensitivity. In my opinion, this is one of the finest studies to build on reductionism in cognitive neuroscience and reconstruct a more ecologically-valid and network-level picture of the link between brain and behavior.
R. B. Mars, C. Piekema, M. G. H. Coles, W. Hulstijn, I. Toni (2007). On the Programming and Reprogramming of Actions Cerebral Cortex, 17 (12), 2972-2979 DOI: 10.1093/cercor/bhm022