Dynamic functional changes associated with cognitive skill learning of an adapted version of the Tower of London task

Abstract

In this study, we used a modified version of the Tower of London (TOL) planning task, in conjunction with positron emission tomography, to examine the neural substrates mediating cognitive skill learning. Twelve healthy, right-handed participants solved a total of 90 different TOL problems. They were scanned on four occasions during the fast learning stage as well as in a condition designed to control for internally guided movements. Practice of the TOL task resulted in a significant decrease in planning, execution, and total time taken to solve the problems. Consistent with the results of previous studies, early performance of the TOL task was associated with increased blood flow activity in the dorsolateral prefrontal, orbitofrontal, and parietal cortices on the left, as well as in the caudate nucleus, cerebellum, and premotor cortex, bilaterally. Interestingly, however, activity in the left caudate nucleus was maintained from the beginning to the end of the learning process, suggesting that this structure plays a role in this type of cognitive skill learning. In addition, correlational analyses revealed that improved performance on the TOL task was associated with a significant decrease of activity in the medial orbitofrontal and frontopolar cortices over the course of learning, areas thought to be involved in decision making, guessing, and monitoring of feedback information. In sum, the results lend further support to the idea that the learning of cognitive skills requiring planning and working memory capacities is mediated through a fronto-striatal network.

Introduction

Neuropsychological, neurophysiological, lesion, and imaging studies have demonstrated the involvement of the frontal lobes in a variety of cognitive skills including attention, inhibition, organised search, monitoring, flexibility of strategy, problem solving, working memory, and planning (e.g., Barbas, 2000a, Barbas, 2000b, Duncan and Owen, 2000, Funahashi, 2001, Fuster, 2000a, Fuster, 2000b, Koechlin et al., 1999, Levy and Goldman-Rakic, 2000, Miller, 2000, Miller and Cohen, 2001. Other studies have recently shown that the striatum, cerebellum, and related structures also play a part in these cognitive processes Andreasen et al., 1999, Cooper et al., 1991, Dagher et al., 2001, Decety et al., 1990, Desmond and Fiez, 1998, Diamond, 2000, Downes et al., 1989, Fiez et al., 1992, Fiez, 1996, Grafman et al., 1992, Kim et al., 1994, Leiner et al., 1995, Middleton and Strick, 2000a, Middleton and Strick, 2000b, Middleton and Strick, 2001, Owen et al., 1998, Rapoport et al., 2000, Wise, 1996. However, very little attention has been given to the changes in brain function associated with improvements in performance inevitably observed following repeated trials of tasks designed to measure higher cognitive functions. Failure to address these effects in imaging studies could lead to incorrect attribution of functional activations to the cognitive process investigated (Garavan et al., 2000).

Knowledge of practice effects on performance in higher cognitive tasks has come mainly from studies investigating the cognitive determinants and neural substrates of cognitive skill learning. This ability is defined as the process by which rules, procedures, and strategies, which are relevant to the performance of a task demanding mental operations, come to be combined and applied effectively following repeated practice Ouellet, 1998, Van Lehn, 1996. To date, previous studies of cognitive skill learning have used mirror reading Kassubek et al., 2001, Poldrack et al., 1998, Poldrack and Gabrieli, 2001, verb generation Andreasen et al., 1995a, Andreasen et al., 1995b, Petersen et al., 1998, Raichle et al., 1994, Raichle, 1998, puzzle solving (Roncacci et al., 1996), recognition memory (Madden et al., 1999), and artificial grammar Fletcher et al., 1999, Knowlton et al., 1992, Peigneux et al., 1999. Yet, no study has identified the brain structures and plasticity associated with the acquisition of improved planning and problem-solving abilities through practice using brain imaging techniques. We therefore used positron emission tomography (PET) to investigate the dynamic changes associated with cognitive skill learning using a modified version of the Tower of London (TOL) planning task. Performance of the TOL is thought to require successful planning, execution, monitoring, and revision of a series of actions using working memory, as well as the selection of counterintuitive moves, which relies on inhibitory processes Chochon et al., 1999, Dehaene and Changeux, 1997, Owen, 1997, Welsh et al., 1999.

Investigations in various patient populations and with sensitive brain imaging techniques have demonstrated the critical role of the prefrontal cortex and basal ganglia in solving TOL problems Baker et al., 1996, Carlin et al., 2000, Dagher et al., 1999, Dagher et al., 2001, Elliott et al., 1997, Lange et al., 1992, Morris et al., 1988, Morris et al., 1993, Owen et al., 1990, Owen et al., 1992, Owen et al., 1995, Owen et al., 1996a, Owen et al., 1998, Rezai et al., 1993, Robbins et al., 1994, Rowe et al., 2001. On one hand, previous lesion studies in humans have shown that patients with unilateral or bilateral damage to the frontal lobes are impaired on the TOL task, as they required more moves to complete the problems compared to matched control subjects Carlin et al., 2000, Owen et al., 1990, Owen et al., 1995, Shallice, 1982. Imaging studies have also corroborated the involvement of frontal lobes during planning, as activations have been observed in those regions, and particularly in areas 9 and 46 (Brodmann) of the mid-dorsolateral prefrontal cortex (dlPFC), using versions of the TOL task Baker et al., 1996, Dagher et al., 1999, Morris et al., 1993, Owen et al., 1996b; Rezai et al., 1993, Robbins et al., 1994. On the other hand, patients with basal ganglia disorders such as Parkinson's disease (PD) are also known to demonstrate performance deficits on the TOL task Morris et al., 1988, Owen et al., 1992. These deficits seem to be related to disease severity with mild PD symptoms affecting initial planning time, and more severe symptoms affecting problem solving accuracy as well Morris et al., 1988, Owen et al., 1992. However, the exact nature of the deficits remains a matter of debate as Hodgson and colleagues (2002) have also found significant differences in the gaze patterns used by patients when solving TOL problems, and suggest that impairments on the TOL may be due to a disruption of trans-saccadic working memory or attentional control.

Some investigators have suggested that these deficits are due to pathology of the frontal lobes (e.g., Morris et al., 1988), but others have proposed that they are caused by basal ganglia dysfunction, which disrupts frontal lobe function via the cortico-striatal loops (e.g., Lange et al., 1992, Owen et al., 1998. The latter hypothesis is based in part on PET studies in healthy subjects showing that cerebral blood flow in the caudate nucleus increases with task difficulty Dagher et al., 1999, Owen et al., 1996a, Owen et al., 1998. In addition, when PD patients are compared to age-matched controls on this task, the cerebral blood flow abnormalities are found in the basal ganglia but not the prefrontal cortex Owen et al., 1998, Dagher et al., 2001.

Finally, involvement of the cerebellum in planning is supported by imaging studies of the TOL task Baker et al., 1996, Elliott et al., 1997, Rowe et al., 2001 and by at least two other lesion studies using a similar planning task, the Tower of Hanoï (TOH) Fiez et al., 1992, Grafman et al., 1992. It is not clear, however, whether activations in the latter structure are associated with the cognitive components of the task or with confounding motor factors. Indeed, other investigators have proposed that the interpretation of cerebellar activations is complicated by the fact that in imaging studies, eye movements were not explicitly matched to their control condition (Baker et al., 1996), and that subjects make eye movements during mental planning (Rowe et al., 2001). On the other hand, Hodgson et al. (2000) have shown that good planners make very few eye movements on the TOL task, suggesting that cerebellar activations might not be related to eye movements per se.

Although the TOL task typically requires participants to solve several (often 12 or more) problems, this paradigm has mainly been used as a measure of cognitive performance. To study the learning effects in planning tasks, researchers have used the TOH and Tower of Toronto (TOT) (Goel and Grafman, 1995). Learning difficulties on both tasks have been shown in patients with striatal dysfunction, such as those seen in PD and Huntington's disease (HD) (Daum et al., 1995; Saint-Cyr et al., 1988; Salmon and Butters, 1995), hence supporting the notion of a fronto-striatal involvement in cognitive skill learning Daum et al., 1995, Salmon and Butters, 1995. However, the TOT and TOH were not used in this experiment because of the declarative nature of their solution. Indeed, once a participant has memorised the sequence of moves necessary to solve a problem, planning is no longer required. The TOL, on the other hand, allows for many different solutions at different levels of difficulty, thus making it impossible to rely on declarative memory to solve the problems.

The present study was designed to determine the functional neuroanatomy underlying implicit cognitive skill learning by combining [¹⁵O]H₂O PET with a modified “learning version” of the TOL task. Our main objective was to investigate the nature of the dynamic functional changes occurring during repeated practice on the TOL task. Significant activations in the prefrontal cortex and striatum were expected, given the established roles of these structures in planning and other forms of skill learning.