Project 3: The Differential Roles of the Prefrontal Cortex in Real-World Problem Solving

Patients with lesions to the prefrontal cortex have well-documented difficulties in decision-making and problem-solving in real-world, open-ended situations, often despite average or even excellent performance on neuropsychological test batteries (Eslinger & Damasio, 1985; Goldstein, Bernard, Fenwick, Burgess, & McNeil, 1993; Shallice & Burgess, 1991; Stuss & Benson, 1986). It is becoming increasingly clear that neuropsychological test batteries are failing to capture some crucial aspects of cognitive and social functioning in the world (Bechara, Damasio, Damasio, & Anderson, 1994; Shallice & Burgess, 1991). There are a range of opinions and intuitions as to the correct characterization of what is missing from neuropsychological tests, and the implications of this for a theory of frontal lobe functions (Damasio, 1994; Goldman-Rakic, 1987; Grafman, 1989; Levine et al., 1998; Shallice, 1988; Shallice & Burgess, 1996; Sirigu et al., 1995).

Our own research in this area – guided by an interdisciplinary background in cognitive, computational, patient, and neuroimaging studies – has led to different theoretical ideas and methodology (Goel, in review; Goel & Grafman, 1995, 2000; Goel, Grafman et al., 1997; Goel, Pullara, & Grafman, 2001). It seems to us that one critical difference between the laboratory tasks – that frontal lobe patients often perform well on – and the real-world tasks – that they often stumble on – is one of structure. Laboratory tasks are well-structured while real-world problems have both ill-structured and well-structured components. The Tower of Hanoi, in which a set of disks are moved from one peg to another, is a typical exam-ple of a well-structured problem. In such puzzles the start state is completely specified, as is the goal state and the set of legal transformations (though generating or selecting the "best" transformation at any given point may be a difficult task). Planning a meal for a guest is an example of an ill-structured task. The start state is incompletely specified (e.g. how hungry will they be?; how much time and effort do I want to expend? etc.). The goal state is also incompletely specified (e.g. how much do I care about impressing the guest?; should there be 3 or 4 courses?; would salmon be appropriate?; would they prefer a barbecue or an indoor meal?; etc.). And finally, the transformation function is also incompletely specified (e.g. should I have the meal catered, prepare it myself, or ask everyone to bring a dish?; if I prepare it, should I use fresh or frozen salmon? etc.). This is an important distinction in the cognitive science problem solving literature and is further developed in Goel (1995).

Based upon the theoretical ideas presented in (Goel, 1995) and the patient data presented in (Goel & Grafman, 1995, 2000; Goel, Grafman et al., 1997; Goel et al., 2001) I have developed a theory of frontal lobe function calculated to explain the discontinuity in patient performance in the lab and world. The general idea is that real world problems have both ill-structured and well-structured components. Ill-structured problem-solving typically involves four phases: problem structuring, preliminary solutions, refinement, and detailing of solutions. Each phase differs with respect to the type of informa-tion dealt with, the degree of com-mit-ment to generated ideas, the level of detail attended to, the number and types of trans-formations en-gaged in, the mental representations needed to support the different types of information and transformations, and the corresponding computational mechanism (Goel, 1995). As one progresses from the preliminary phases to the detailing phases, the problem becomes more structured.

Our preliminary data (Goel & Grafman, 2000) suggests that a major issue for frontal lobe patients is movement across the preliminary solution, refinement, and detailing phases of the problem. Preliminary solution generation is a classical case of cre-ative, ill-structured problem solving. It is a phase of “cognitive groping” or “way-finding”, a phase of concept construction, where a few kernel ideas are generated and explored through transformations. This generation and exploration of ideas/concepts is fa-cili-tated by the abstract nature of infor-ma-tion being con-sidered, a low degree of commitment to gen-erated ideas, the coarseness of detail, and a large number of lateral transformations. A lateral transformation is one where movement is from one idea to a slightly differ-ent idea rather than a more detailed version of the same idea. Lateral transformations are necessary for the widening of the prob-lem space and the exploration and development of ker-nel ideas. The rules underlying lateral transformations are difficult to articulated (Goel, 1995).

The refinement and detailing phases are more con-strained and structured. They are phases where preconstructed concepts are manipulated. Com-mitments are made to a particular solution and propa-gated through the problem space. They are character-ized by the concrete na-ture of in-formation being con-sidered, a high degree of commitment to generated ideas, attention to detail, and a large number of verti-cal transformations. A vertical trans-formation is one where movement is from one idea to a more detailed version of the same idea. It results in a deepening of the prob-lem space. The rules underlying vertical transformations can often be articulated (Goel, 1995).

Goel (1995) has argued that the ability to engage in lateral transformations is underwritten by a mechanism that supports ill-structured mental representations and computation. Ill-structured representations are imprecise, ambiguous, fluid, indeterminate, vague, etc. The ability to engage in vertical transformations is underwritten by a mechanism that supports well-structured mental representations and computation. Well-structured representations are precise, distinct, determinate, and unambiguous. Furthermore, there is a computational dissociation between these two mechanisms (Giunti, 1997; Goel, 1995). Laboratory problems emphasize well-structured mental representations while real-world problems require both ill- and well-structured mental representations. Ill-structured and well-structured representations differ with respect to modes of inference and computational mechanisms. On this account, one explanation for the general dissociation exhibited by frontal lobe patients between their lab and world performance is that there is an anatomical dissociation corresponding to the computational dissociation.

Given that most patients that exhibit a dissociation between world and lab performance have lesions to the right PFC (Burgess, 2000), we propose that well-structured representations and computations may map onto left hemisphere processes while ill-structured representations and computations map onto right hemisphere processes. That is, the processing of inarticulate, ill-structured information may involve the right prefrontal cortex, while the processing of articulate well-structured information is underwritten by the left prefrontal cortex.

In the current phase of this project we are undertaking to a large study with 80 neurological patients to test hypothesis H1. We are examining and comparing the performance of patients with LH (20), RH (20), and bilateral (20) focal lesions in the prefrontal cortex, patients with lesions in areas other than the prefrontal cortex (20), and normal controls (20) in the execution of complex real-world travel planning task. We are administering ill- and well-structured versions of the planning task to all subject groups and looking for differential performance as a function of lesion location, as per hypothesis H1.

Specifically, it should be possible to find, not only (right PFC) patients, who have difficulty with preliminary, explorative, ill-structured phases of problems, but also (left PFC) patients who can negotiate the lack of structure, but have difficulty in the later, more concrete phases where they are required to make commitments and follow through. Both types of patients will have difficulty with coping in the world, but with different cognitive signatures. Right hemisphere patients will have difficulty in the early, explorative phases, while left hemisphere patients will have difficulty in the later, well-structured phases, where commitments need to be made and propagated. For example, in terms of planning, the former would have difficulty with plan formulation, and the latter would have difficulty with plan execution.

In addition to the patient study, we have also tested the hypotheses about the differential involvement of left and right PFC in ill- and well-structured problem solving situations with two fMRI studies of normal controls. In one study we scanned 12 normal volunteers as they engaged in mental set shifts involving the meaningful juxtaposition of unrelated categories (e.g. imagine “a chair that is a vacuum cleaner”) and compared the results to mental operations involving within-category superordinate/subordinate relationships (e.g. imagine “furniture that is a chair”). As predicted in H1, we found right prefrontal activation in the juxtaposition of unrelated categories compared to the familiar, related categories condition. In the second study we scanned 12 normal volunteers as they engaged in the well known “matchstick” task from the convergent/divergent literature. Once again we found activation in right prefrontal cortex in the divergent condition compared to the convergent condition. More specifically, the right lateral/ventral (BA 47) was activated in both studies, suggesting that this region plays a critical role in dealing with novel, ill-structured situations, such as found in coping/problem-solving in real-world situations. These results will allow us to revisit our patient data with refined hypotheses.

Basic operating funds for this study are covered by my CIHR/MRC grant (Title of Project: “The Differential Roles of the Prefrontal Cortex in Real-World Problem Solving”).


Nov. 10, 2003