The special involvement of the rostrolateral prefrontal cortex in planning abilities: An event-related fMRI study with the Tower of London paradigm

Abstract

Planning abilities are essential for the successful management of everyday life activities. Although several neuroimaging studies provide evidence that the prefrontal cortex is crucially involved in planning, the differential roles of its subregions are still a matter of debate. The aim of this study was to investigate the neural correlates of planning by focusing on the functional differentiation between the dorsolateral and rostrolateral prefrontal cortex using the Tower of London (ToL) task and a parametric event-related functional MRI design. In order to control for activations unspecific to planning, two control conditions were presented, which were matched for the length of single events in the ToL task. Seventeen right-handed healthy subjects participated in this study. All statistics were reported with corrections for multiple comparisons (p < 0.05). Compared to control conditions, activations in the ToL task were observed in the dorsolateral prefrontal cortex bilaterally, the right ventrolateral and left rostrolateral prefrontal cortex along with the thalamus, as well as in the parietal and premotor cortex bilaterally. Task complexity dependent analyses revealed that only the left rostrolateral prefrontal cortex showed a BOLD signal increase over the four planning levels, which could not be observed in the control conditions. Hence, current findings suggest that planning involves an extensive fronto-parieto-thalamic network. Within this network, the rostrolateral prefrontal cortex seems to be the only region that is exclusively reactive to planning specific processes, which we described in terms of simultaneous monitoring of internally generated and externally presented information.

Introduction

Planning is essential to problem solving. A problem arises if one has a goal, but one's knowledge offers no clear path for achieving it. Thus, a problem comprises a dissatisfactory status quo: a desired goal and the barriers to achieving it (Duncker, 1945).

Understanding the dilemma creates a problem space, within which during planning a sequence of operations is executed enabling the transformation of the start-state into the goal-state. Planning depends both on the number of operations to be sequentially anticipated and the number of problem states which separate the start from the goal-state.

To reliably quantify planning demands, a transparent, well-defined problem with a circumscribed problem space is needed (Simon & Reed, 1976). The goal must be clear, the means for achieving it known and the problem space limited in size.

Although such a restricted problem has low external validity for describing complex problem solving processes in real life, it is suitable for investigating the neural correlates specific to planning. Given an appropriate design, the planning demands can be easily manipulated and controlled.

The Tower of London (ToL) task – in which the subject must rearrange a set of three colored balls on three pegs of varying size from the start-state into the goal-state in the minimum number of moves – has proven useful to measure planning abilities by systematically varying planning demands. In this test, all components of the problem space are well known, no complex rules have to be learned and several problem states can be defined depending on their complexity (Berg & Byrd, 2002; Shallice, 1982).

Additionally, the ToL task is especially appropriate for investigating the role of the prefrontal cortex (PFC), as it has been shown to predominantly tap executive processing and planning as one important aspect of it, which are known to be critically dependent on prefrontal cortex activation (Stuss & Benson, 1986). Ample evidence from neuropsychological, lesion and neuroimaging studies suggests that specific areas of the prefrontal cortex are significantly involved in higher executive function like decision making and planning of acts (Damasio, Tranel, & Damasio, 1990; Stuss & Benson, 1986). Specifically, two areas within it, the rostrolateral prefrontal cortex (RLPFC; BA 10) and the dorsolateral prefrontal cortex (DLPFC; BA 9, 46) constituting the anterior and posterior part of the middle and superior gyrus, respectively, were found to be essentially involved in higher executive function (Christoff & Gabrieli, 2000).

Activations in the RLPFC have often been observed in studies using complex tasks, like Wisconsin Card Sorting Test (WCST) (Goldberg et al., 1998) or inductive reasoning tasks (Goel, Gold, Kapur, & Houle, 1997). In an fMRI study using Ravens Progressive Matrices (RPM), Christoff et al. (2001) showed that the left RLPFC was activated when comparing complex with simple problems. Whilst activation of the dorsolateral prefrontal regions has mostly been reported in association with the maintenance and manipulation of externally presented information (Braver et al., 1997; Sakai, Rowe, & Passingham, 2002), it has been suggested that the RLPFC becomes specifically involved when evaluation, monitoring or manipulation is performed on internally generated information (Christoff & Gabrieli, 2000). Koechlin et al., 1999, Koechlin et al., 2000 described the function of the RLPFC with the term “cognitive branching”, which characterizes the process of keeping a higher order goal in mind while prosecuting sub-goals. Based on recent findings (e.g. Braver, Reynolds, & Donaldson, 2003) that demonstrated activation in rostrolateral prefrontal regions in association with task-switching (which demands the internal maintenance of one task while performing another), Gilbert, Frith, and Burgess (2005) expanded the concept of “cognitive branching” by suggesting that the switching between stimulus-oriented cognitive processes (i.e. processes provoked by external sensory information like the visual comparison of start and goal-state during solving of ToL problems) and stimulus-independent cognitive processes (i.e. processes detached from external stimuli such as the mental rearrangement of the stimuli in the ToL task) might be the determining factor for RLPFC activation.

However, not all previous studies with the ToL task found activations in the RLPFC. The most robust findings were activations in the DLPFC (BA 9, 46), the parietal lobe (BA 7, 40) and the premotor cortex (BA 6, 8) (Dagher, Owen, Boecker, & Brooks, 1999; Lazeron et al., 2000; Owen, Doyon, Petrides, & Evans, 1996; Schall et al., 2003, van den Heuvel et al., 2003). The most probable explanation for the inconsistencies in reported brain activations across studies with the ToL task might be differences in the task design. Whereas some studies (Dagher et al., 1999, Owen et al., 1996; Rowe, Owen, Johnsrude, & Passingham, 2001) used a motor version of a ToL task, in which subjects moved balls by touching a touch-sensitive computer monitor, other studies employed a mental version, in which subjects had to mentally work out the plan for the solution (Baker et al., 1996, Lazeron et al., 2000, Schall et al., 2003, van den Heuvel et al., 2003). In addition, the studies also differed in experimental setting. Most studies used a subtraction design to compare blocked ToL problems against a baseline or control condition. Baker et al. (1996) showed in their PET study with a ToL version – in which subjects were asked to mentally work out a solution – that the right RLPFC was more active in the difficult condition with four and five planning moves when compared to control conditions than in the easy condition with two and three planning moves.

In contrast, Lazeron et al. (2000) failed to show any significant differences between the easy and the difficult condition. However, for the difficult condition they used five to seven move problems to be mentally solved and two buttons for the choice of the correct answer. It is questionable whether subjects really mentally planned the entire sequence of such difficult problems.

Some studies employed a parametric design to correlate the task load with the rCBF and BOLD signal, respectively. In a PET study, Dagher et al. (1999) reported a correlation between the increase in planning amount over five levels and the increase in rCBF bilateral in the DLPFC, the premotor cortex, the anterior cingulate cortex, the right caudate nucleus along with the right RLPFC. The only one event-related fMRI study of van den Heuvel et al. (2003) showed only a small correlation (cluster size of 2 voxels) between task load and BOLD signal in the left RLPFC. Again, the strongest correlation occurred in the DLPFC, the premotor cortex, the striate cortex and additionally in the parietal lobe.

However, it is not always clear whether the increase in rCBF and BOLD signal in the reported areas is to be attributed to the higher planning demands, the increasing working memory demands, including short-term maintenance and manipulation of information, the higher amount of visuo-spatial processing and even unspecific effects as general task complexity.

To resolve this interpretation problem, we implemented two control conditions in the present study that were parametrically varied with regard to the amount of visuo-spatial information processing and working memory degree.

We used an event-related functional MRI design and a ToL version, in which subjects had to plan the entire solution in mind to minimize motor specific activations. We defined the problem complexity by the number of moves to plan ahead. Derived from the above mentioned literature, we expected prefrontal, cingulate, parietal and striatal areas to be activated during solving of ToL problems. Given the findings on the special involvement of the RLPFC in the manipulation of internally generated information, we hypothesized that the increase in planning complexity is associated with a parametric increase of BOLD signal in the rostrolateral PFC. For both control conditions, we did not expect to observe signal changes in RLPFC in dependency on their complexity levels. For the dorsolateral PFC, we hypothesized a BOLD signal increase in the ToL task as well as in both control conditions due to enhanced working memory and visuo-spatial demands.