Damage to the ventromedial prefrontal cortex (VMPC) is well known to cause severe impairments of real-world behavior, leading in particular to disruption of social conduct and interpersonal behavior, impaired occupational functioning, and defective emotional regulation and decision making. Such deficits, however, have been notoriously difficult to capture in clinical and experimental laboratory settings. For example, it has been noted by many investigators that patients with VMPC lesions often—perhaps even typically—perform normally on tasks such as the Wisconsin Card Sorting Test, Trail Making Test, Tower of London (and its variants), Category Test, and Stroop Color–Word Interference Test. Not all VMPC patients pass all of these types of test, of course, but some of the patients with the most egregious real-world behavioral defects do have sterling laboratory performance profiles (cf. Eslinger & Damasio, 1985). One frequently cited explanation for this seeming paradox is that the laboratory tasks provide considerable structure, direction, and guidance, so that patients do not have to cope with the types of ambiguous, open-ended, unconstrained situations that are more commonly faced in the real world (e.g., Lezak, Howieson, & Loring, 2004, pp. 611–638).

Against this background, it is interesting to consider some notable task exceptions. One is the Iowa Gambling Task (IGT), which was developed by Bechara and colleagues (Bechara, Damasio, Damasio, & Anderson, 1994) in part to address the limitations of extant laboratory tasks of decision making. The IGT has proven remarkably effective at exposing decision-making deficits in VMPC patients and in a number of other populations with known or suspected dysfunction of VMPC and VMPC-related neural systems (e.g., Bechara, 2004; Bechara, Tranel, & Damasio, 2000; Denburg, Tranel, & Bechara, 2005; Fellows & Farah, 2005; Rogers et al., 1999; Stout, Busemeyer, Lin, Grant, & Bonson, 2004; for review, see Yechiam, Busemeyer, Stout, & Bechara, 2005). The IGT, however, is sine qua non a decision-making task, and despite its popularity in the research literature, the relationship between IGT performance and real-world behavior remains largely unexplored (see Anderson, Barrash, Bechara, & Tranel, 2006, for a partial exception).

Several other researchers have devoted considerable effort to developing tests that would be more ecologically valid as a means to assess the typical “dysexecutive syndrome” in patients with brain injuries. For example, Wilson, Evans, Emslie, Alderman, and Burgess (1998) developed a battery of tests known as the Behavioural Assessment of the Dysexecutive Syndrome (BADS). The BADS included a modified version of the Six Elements Test (SET); the SET, in turn, had originally been described by Shallice and Burgess (1991), along with another so-called “strategy application” task (the Multiple Errands Test, or MET), as a more sensitive means of flushing out the behavioral navigation defects of brain-injured patients. The MET, in particular, was designed to mimic various “real-world” demands and activities (e.g., shopping in a business precinct). These semiquantitative tasks provided relatively unstructured, open-ended situations with multiple subgoals, and they lacked the constraint and direction typical of clinical neuropsychological measures.

In the initial report on the SET and MET, Shallice and Burgess (1991) described three patients who had frontal lobe injuries as a consequence of severe head trauma, and who manifested impaired performances on the SET and MET despite relatively normal intelligence, language, and perception. Since then, these tasks (and various modifications) have been utilized by a number of researchers to investigate executive functioning in normals and various patient populations (e.g., Alderman, Burgess, Knight, & Henman, 2003; Burgess, Alderman, Evans, Emslie, & Wilson, 1998; Channon & Crawford, 1999; Duncan, Johnson, Swales, & Freer, 1997; Goel & Grafman, 2000; Goldstein, Bernard, Fenwick, Burgess, & McNeil, 1993; Jelicic, Henquet, Derix, & Jolles, 2001; Kafer & Hunter, 1997; Kliegel, McDaniel, & Einstein, 2000; Knight, Alderman, & Burgess, 2002; Levine, Dawson, Boutet, Schwartz, & Stuss, 2000a; Levine et al., in press Levine et al., 1998; Wilson et al., 1998), especially with an eye to rehabilitation planning (e.g., Levine et al., 2000b; Manly, Hawkins, Evans, Woldt, & Robertson, 2002). By now, there are consistent findings supporting the conclusion that these tasks are useful to characterize and quantify executive functioning deficits, and Burgess and his colleagues have made a convincing argument that “strategy application” types of task may be superior to traditional clinical neuropsychological tests (e.g., Wisconsin Card Sorting Test, WCST; Tower of London) at probing “real-world” deficits in executive functioning (Burgess et al., 2006).

Most of the published work involving the SET and MET (and their spin-offs) has focused on the construct validity of the tasks vis-à-vis executive functioning. Very few previous studies have given attention to neuroanatomical issues. The gist of most of the available work is that executive dysfunction—and failure on the SET and MET—is associated with frontal lobe damage. Beyond that, however, it is difficult to be more specific. For example, it is not clear if damage to the medial orbital and lower medial prefrontal sectors (the region that we have termed the ventromedial prefrontal cortices, or VMPC; see Tranel, Damasio, Denburg, & Bechara, 2005, for a precise description) is typical of patients who fail the SET and MET. In most previous studies, neuroanatomical issues were not addressed at all (e.g., Alderman et al., 2003; Burgess et al., 1998; Duncan et al., 1997; Jelicic et al., 2001; Kafer & Hunter, 1997; Kliegel et al., 2000; Knight et al., 2002; Wilson et al., 1998). Shallice and Burgess (1991) provided brief descriptions of the lesions in their original three cases, but other than mentioning frontal lobe damage, no specific details are provided, and it is not clear whether or not the VMPC was affected by the lesions (for two of these patients, a single computed tomography, CT, cut is provided in the recent review paper of Burgess et al. (2006), but this does not clarify whether there is VMPC involvement). Levine et al. (1998) included 10 patients with focal frontal lesions in their study (which involved a modified version of the SET), but at least half of these patients actually did not have VMPC damage (and lesion localization was based on acute clinical CT scans in most of the patients). Burgess, Veitch, de Lacy Costello, and Shallice (2000; also see Burgess, 2000), using mostly tumor cases (many of which were gliomas), undertook a neuroanatomical analysis of performance on a task that roughly resembled the SET and found that defects were associated with medial and polar left prefrontal structures (but not VMPC), and right dorsolateral prefrontal structures. An important caveat about many of the published studies is that the “frontal” patients tend to be severe head injury cases, making lesion localization somewhat hazardous to begin with.

The issue of whether or not the VMPC is specifically involved in tasks such as the SET and MET is important, because the real-world behavioral navigation impairments that are the real crux of the “dysexecutive syndrome” have been repeatedly linked to VMPC damage (e.g., Anderson et al., 2006; Barrash, Tranel, & Anderson, 2000; Shuren & Grafman, 2002; Stuss & Levine, 2002). Damage to superior mesial or dorsolateral sectors, for example, does not typically lead to blatant impairments of social conduct and associated functions (e.g., Anderson & Tranel, 2002). (A partial exception is the analysis by Burgess, 2000; Burgess et al., 2000, which linked the general construct of “multitasking” to medial and polar left prefrontal structures and right dorsolateral prefrontal cortex.) Moreover, there is a strong link between the VMPC and numerous facets of emotional processing (e.g., A. R. Damasio, 1994), and the role of emotional dysfunction in real-world behavioral incompetency in VMPC patients has been recently documented (Anderson et al., 2006; we return to this issue in the Discussion).

In the current study, we sought to replicate and extend the initial findings of Shallice and Burgess (1991), using a closely adapted version of their task and focusing on a carefully selected group of patients with focal, stable damage to VMPC. This group was contrasted with comparison groups of participants with (a) prefrontal brain damage outside the VMPC region; (b) nonprefrontal brain damage; and (c) no brain damage. We hypothesized that VMPC damage would be associated with defective performance on strategy application tasks, whereas damage outside the VMPC region would not. A secondary goal was to explore the extent to which defects on conventional neuropsychological tests of “executive functioning” would be manifest in VMPC patients and whether performance on conventional executive functioning tests would be related to performance on the strategy application tasks.

We formed four groups of participants, as follows:

1. Ventromedial prefrontal group (VMPC; n=9) This was the main target group for the study, and it included 9 patients with bilateral damage to the VMPC region. For the purposes of this study, the VMPC was defined as the region that encompasses the medial orbital sector and the lower medial prefrontal sector of the prefrontal lobes (Brodmann areas medial 11, 25, 12, ventromedial 10, and anteroventral 32; see Tranel et al., 2005, for a detailed illustration of this region). The lesions were caused by cerebrovascular accident (n=5) or resection of benign tumor (n=4).

2. Prefrontal comparison group (PFC; n=8) This was a brain-damaged comparison group that included patients with prefrontal damage outside the VMPC region: specifically, in left or right prefrontal sectors exclusive of those identified in (1) above. Participants in this group had lesions that were neuroanatomically heterogeneous; some were circumscribed to the right or left dorsolateral or lateral orbital region, and some were large frontoparietal lesions. Participants in the prefrontal comparison group were also selected so as to have demographic characteristics similar to those of the VMPC group. The lesions were caused by cerebrovascular accident (n=6); resection of arteriovenous malformation (AVM; n=1), or benign tumor (n=1).

3. Nonprefrontal comparison group (nonPFC; n=17) This was a brain-damaged comparison group that included patients with damage outside the prefrontal region: specifically, in left or right hemisphere structures exclusive of structures anterior to and including the premotor cortices. Participants in this group were also selected so as to have demographic characteristics similar to those of the VMPC group. The lesions were caused by cerebrovascular accident (n=11), resection of AVM (n=3), temporal lobectomy (n=2), or herpes simplex encephalitis (n=1).

For the three brain-damaged groups, the neuroanatomical classification was based on the following methods. Neuroanatomical analysis was derived from magnetic resonance (MR) data or, for a few patients in whom an MR could not be obtained due to the presence of metal clips, on computerized axial tomography data. All neuroimaging data were obtained at 3 or more months postlesion onset, and lesion locations were mapped according to our standard procedures (H. Damasio & Frank, 1992; Frank et al., 1997). Based on the neuroanatomical analysis, participants were classified into one of the three brain-damaged groups specified above (additional details regarding this type of neuroanatomical classification are provided in Anderson et al., 2006). The classification of participants into the different neuroanatomical groups was reviewed and confirmed by H. Damasio.

We used the MAP-3 and Brainvox techniques developed by H. Damasio (2000) to calculate lesion overlap maps for the three patient groups. In brief, this allows a determination of how many patients in each group have lesions at any particular voxel; this calculation is color coded and rendered in standard brain space (see H. Damasio, Tranel, Grabowski, Adolphs, & Damasio, 2004, for details of this procedure). Figure 1, Figure 2, and Figure 3 provide lesion overlap maps for the VMPC, PFC, and nonPFC groups, respectively. Figure 1 shows that the highest area of lesion overlap in the VMPC group (coded in red) is in the medial orbital and lower medial prefrontal region, bilaterally. In the PFC group (Figure 2), the most common area of damage in is right posterolateral orbital and left posterior dorsolateral regions; none of the patients had lesions encroaching on the VMPC region. Figure 3 shows that for the nonPFC group, none of the patients had lesions that encroached on prefrontal cortices.

Figure 1 Lesion overlap map for the 9 patients with ventromedial prefrontal cortex (VMPC) damage. The figure shows mesial and lateral views of the left (left upper panels) and right (right upper panels) hemispheres, and a ventral view (middle upper panel). The (more ...)

Figure 2 Lesion overlap map for the 8 patients with prefrontal cortex (PFC) damage outside the ventromedial prefrontal cortex (VMPC) region. The figure shows mesial and lateral views of the left (left upper panels) and right (right upper panels) hemispheres, and (more ...)

Figure 3 Lesion overlap map for the 17 patients with nonprefrontal (nonPFC) damage. The figure shows mesial and lateral views of the left (left panels) and right (right panels) hemispheres, and a ventral view (middle panel). The color bar indicates extent of lesion (more ...)

4. Normal comparison group (NC; n=20) This was a normal comparison group, composed of 20 participants who were free of neurological and psychiatric disease. The normal participants were recruited through the Department of Neurology and were compensated for their participation. These participants had to be unfamiliar with the local urban area (so that they would be operating from the same knowledge base, more or less, as the brain-damaged participants, most of whom are from areas outside of Iowa City), and this was ascertained by one of the experimenters (J.H.N.). Participants in the normal comparison group were also selected so as to have demographic characteristics similar to those of the VMPC group.

Demographic data for the four groups of participants are provided in Table 1. The groups were well matched on demographic characteristics, with similar ages, educational levels, gender ratios, and handedness distributions. The VMPC and PFC groups had somewhat longer chronicity (time since lesion onset) than the nonPFC group, but this was not statistically significant in a one-way analysis of variance (ANOVA).

TABLE 1 Participant demographic data

Neuropsychological assessment All brain damaged participants were administered a comprehensive battery of neuropsychological tests, in accordance with their participation in our general research program (see Tranel, in press). The tests probe various cognitive domains including orientation and attention, IQ, verbal and visual memory, speech and language, visuoperceptual and visuoconstructional abilities, and executive functions. For the purposes of the current study, the following specific tests were deemed most relevant, and specific data for these tests are provided (see Table 2):

• IQ: Wechsler Adult Intelligence Scale-R (Verbal IQ; Performance IQ)
• Memory: Verbal=Auditory−Verbal Learning Test; Visual=Benton Visual Rentention Test
• Executive functions: Wisconsin Card Sorting Test; Trail Making Test (Trail Making B minus Trail Making A)

TABLE 2 Neuropsychological data

Experimental measures

Six Elements Test This set of tasks was modeled closely on the Shallice and Burgess Six Elements Test protocol (Shallice & Burgess, 1991). Participants completed the tests in a laboratory setting, set up so that the participant performed the tasks under observation, but without the examiner in the room (via a one-way window). The participants were instructed to complete six tasks (two sets of three tasks) within a time limit of 15 minutes, in a manner that would maximize their scores. The following tasks were utilized:

• Naming A and Naming B subtasks. These subtasks involved writing down the names of objects presented in a pictorial format in two notebooks labeled “Naming A” and “Naming B.” Each notebook contained 50 pictures of common natural and artifactual entities.
• Arithmetic A and Arithmetic B subtasks. These subtasks required participants to examine flashcards with addition, subtraction, multiplication, and division problems, and to answer the problems. Each set of flashcards contained 30 different problems, ordered to be increasingly difficult across the set.
• Dictation A and Dictation B. These subtasks required participants to provide verbal descriptions of (A) how they arrived at the University of Iowa Hospitals and Clinics (UIHC), and (B) their intended route back home. The participants provided the information orally, and their responses were recorded on a tape recorder (the participants did not have to operate any equipment). (Many of the brain-damaged patients have had multiple visits to UIHC, which could give them an advantage over the normal comparison participants on the dictation tasks. However, this was probably at least partially offset by the fact that more of the normal comparison participants were from closer to Iowa City than the patients. Moreover, a qualitative analysis of the dictation protocols generated by the patient groups and the normal group did not suggest that any group had a particular advantage on these tasks.)

Participants were given five rules for these tasks (after Shallice and Burgess, 1991):

• You are not allowed to do the two subtasks of the same type one after the other.
• Each of the six subtasks is given equal weight.
• Within the Naming and Arithmetic subtasks, points will be given for correct answers.
• For the pictures/problems, earlier items will be given more credit than later ones.
• Errors and omissions will count against you.

Multiple Errands Test Subsequent to the completion of the Six Elements Test, participants were instructed about the second part of the experiment, which required in vivo participation at a mall located in a downtown urban campus. Participants were escorted to the mall via the University of Iowa’s Cambus transportation system. While still at the hospital, participants were provided with the following tasks and rules (presented in both written and spoken formats):

Tasks for Multiple Errands Test

• Buy one cookie.
• Buy one package of cough drops.
• Buy one Kleenex package.
• Buy one postcard.
• Buy one book marker.
• Buy one candle.
• You must meet up with the experimenter 15 minutes after starting your tasks. (This was explained as a “check-in point,” and participants were reminded that they had more than 15 minutes to complete all of the tasks.)
• You must gather the following pieces of information and write them down on the note card provided:

• The name of the store in the Old Capital Mall likely to have the most expensive item.
• The price of one dozen roses.
• The number of fast food eating establishments in the Old Capital Mall.
• The forecast high temperature for Denver, Colorado today.

Rules for Multiple Errands Test (the rules followed those used by Shallice and Burgess):

• You are to spend as little money as possible (within reason).
• You are to take as little time as possible (without rushing excessively).
• No store should be entered other than to buy something.
• Please tell the experimenter when you leave a store what you have bought.
• You are not to use anything not bought on your adventure (other than your own watch) to assist you.
• You may do the tasks in any order.

As defined by Shallice and Burgess (1991), errors during the execution of tasks were recorded as inefficiencies (when a more effective strategy could have been applied), rule breaks (where a specific rule is broken), interpretation failure (where the requirements of a particular task are misunderstood), and task failure (a task is either not carried out or not completed satisfactorily). The total mall errors were calculated by summing across these error types. Also, the experimenters monitored and recorded mall attempts (number of tasks attempted), mall completions without errors (number of tasks completed according to the rules), total mall completions (number of tasks completed, both with and without errors), and total time (the total time the participant spent on the mall tasks). For the purposes of this scoring, the total number of tasks was considered to be 11: This includes the six purchasing tasks, the check-in task, and the four information-gathering tasks. The mall completions without errors score is a subset of the total mall completions (with and without errors).

Data analysis The primary dependent variables were the various measures for the Six Elements Test and the Multiple Errands Test, as specified above. The groups were compared on these variables using multivariate analysis of variance (MANOVA) and correcting for multiple comparisons (Bonferroni) wherever appropriate. We also conducted correlational analyses, to explore the relationships between performances on the SET and MET and the various neuropsychological measures. The correlational analyses were conducted in the brain-damaged participants (overall N=34).

TABLE 3 Six Elements Test: Mean values as a function of group

TABLE 4 Multiple Errands Test: Mean values as a function of group

In sum, the data from the MET indicate that the VMPC group had a higher number of overall mall errors (7.6) than the PFC (5.9) and nonPFC (4.5) brain-damaged comparison groups and the normal comparison group (3.4). Compared to the normal group, the VMPC group had more than twice as many mall errors. The VMPC group tended to have higher error scores on the various error subtypes, too, especially rule breaks and task failures. The VMPC group had fewer mall attempts and fewer mall completions (both without errors and total) than the other brain-damaged groups and the normal comparison group. The total time was highest in the VMPC group, although this was not statistically significant and did not differ much from that of the nonPFC group.

TABLE 5 Correlations between strategy application tests and neuropsychological tests

The data from this study confirm previous findings from Shallice, Burgess, and colleagues (Shallice & Burgess, 1991; Burgess et al., 2006), indicating that failure on various “real-world” types of strategy application task—in particular, the Multiple Errands Test (MET)—is common following frontal lobe damage. Our data extend previous work by indicating that this pattern is especially characteristic of damage to the ventromedial prefrontal sector of the frontal lobes. Our findings also indicate that performances on the SET and MET were not correlated with IQ or memory measures, but there was a modest relationship between some SET and MET variables and scores on the Wisconsin Card Sorting Test and the Trail Making Test.

Our findings provide partial support for the main hypothesis. Specifically, it was found that damage to the VMPC sector was associated with defective performance on the MET, as hypothesized. Damage to the prefrontal cortex more generally (outside the VMPC sector) also tended to be associated with poorer performances on strategy application tasks—specifically, the PFC group had more errors on the MET, and fewer mall completions, than did the nonPFC and normal comparison groups, although the PFC group was less discrepant from the comparison groups than was the VMPC group. Overall, though, poor performances on the MET tasks were more characteristic of participants with VMPC damage, and the VMPC participants clearly had the worst MET performances of any of the other groups in the study.

As far as conventional neuropsychological tests such as the Wisconsin Card Sorting Test are concerned, we went into the study with two goals: One was to explore whether such “executive functioning” tests would discriminate participants with VMPC damage, and the other was to explore the extent to which SET and MET performances would correlate with neuropsychological measures. Regarding the first goal, we found that the VMPC group did not differ from the other brain-damaged groups on neuropsychological measures and, in particular, did not differ on the two executive functioning tests—namely, the Wisconsin Card Sorting Test and the Trail Making Test. In fact, as a group, the VMPC group was virtually indistinguishable from the nonPFC comparison group. This outcome confirms previous work that has demonstrated that conventional executive functioning tests are not particularly sensitive to VMPC integrity (e.g., Anderson, Tranel, Grabowski, Adolphs, & Damasio, 1991; Grafman, Jonas, & Salazar, 1990; Miyake et al., 2000; Stuss et al., 2000). As far as correlations with SET and MET variables are concerned, we did find a couple of consistent, modest relationships: (a) The Trail Making Test (calculated as time for Part B minus time for Part A) correlated significantly with two SET variables and one MET variable; (b) the MET mall attempts variable correlated significantly with two variables from the WCST and with the Trail Making Test. The degree of correlation was such that roughly 20 to 35% of the variance on certain SET and MET variables could be explained by WCST and/or Trail Making Test performances. Our results are generally consistent with data reported by Knight et al. (2002), who found correlations in the .5 to .6 range between the Modified Card Sorting Test percentage Perseverative Errors score and several MET variables (the investigators used a simplified version of the MET).

Overall, our results really do not provide compelling support for the utility of the SET and MET procedures for assessing behavioral navigation impairments in brain-damaged patients, at least insofar as clinical application is concerned. There are several reasons for this conclusion. As far as the SET is concerned, we actually did not find evidence of compelling task failures in the VMPC group. The only measure on which the group was below all of the other groups was number of tasks attempted, and this was not statistically significant. This result does not support the conclusion that the SET would be diagnostically useful on a case-by-case basis, especially in regard to linking up the test with specific neuroanatomical underpinnings. The MET yielded more encouraging results, and the VMPC group was consistently below the other comparison groups on nearly every measure from the MET, including total errors and numbers of attempts and completions. Even here, though, the between-group differences were not dramatic, and it would be hard to argue that the test would be diagnostically useful on a case-by-case basis. One also has to factor in the time and effort that these tasks require, not only from the participants but also from the experimenter. The MET, in particular, is time- and labor-intensive for all concerned. Thus, taking into account both the magnitude of between-group differences and the time and effort requirements of the tests, it seems unreasonable to suggest that the SET and MET could be easily incorporated as standard parts of a neuropsychological diagnostic procedure. The fact that two conventional procedures—the WCST and Trail Making Test—correlated moderately with some SET and MET variables also makes it difficult to stake a strong claim for the utility of these “strategy application” procedures.

We should hasten to add that we have no disagreement with the general objective of developing more “ecologically valid” measures of executive functioning, as characterized by the research program of Burgess and colleagues (see Burgess et al., 2006). Those researchers provide a detailed and convincing set of arguments for why extant clinical measures (such as the WCST) are lacking and why certain experimental measures (such as the MET) may offer hope for improvement. And it could be that applications to patients with frontal damage caused by other etiologies, such as head injury and malignant brain tumors (as is more common in the patient populations reported by Burgess and colleagues; see also Levin, Goldstein, Williams, & Eisenberg, 1991), would return a higher yield, especially when rehabilitation planning is factored into the equation. The patients we used in our study all had focal, stable brain injuries caused mainly by stroke or surgical resection of benign tumors, and this may account for some of the differences between our conclusions and those set forth by Burgess and colleagues (e.g., Burgess et al., 2006). In addition, if procedures such as the MET could be set up as virtual environments and run on a computer within the context of a neuropsychological testing room, this might greatly reduce the time and effort demands associated with the actual shopping mall visit and, in turn, improve the net diagnostic yield for the MET.

On the upside, the MET in particular was useful to capture “real-world” behavioral navigation defects in patients with frontal lobe damage, and our study shows that this relationship is strongest for bilateral VMPC damage. Prefrontal lesions outside the VMPC region did not lead to as severe of impairments, and lesions outside the prefrontal cortices were not associated with strategy application task failure. These results add neuroanatomical specificity to previous work, especially the elegant program of research by Burgess and colleagues on these and related “multitasking” measurements (see Burgess et al., 2006, for review). Recently, we have demonstrated that impairments in real-world competencies in patients with VMPC damage are substantially accounted for by emotional disturbances, especially heightened emotional reactivity and hypoemotionality (Anderson et al., 2006). In that study, too, there was a considerable degree of anatomical specificity in the findings, and prefrontal damage outside the VMPC region did not consistently lead to either emotional disturbance or impaired real-world competencies. Collectively, these studies add empirical support to the notion that the VMPC is critical for successful real-world behavioral navigation. The measurement of these competencies in the clinic or laboratory, however, remains a challenge.

This study was supported by Program Project Grant NINDS NS19632. We thank Dr. Hanna Damasio for reviewing the neuroanatomical classification for the brain-damaged participants, Dr. Eduardo Vianna for assistance preparing the figures, and Ken Manzel for help with the statistical analyses. D.T. and S.W.A. are supported by NINDS P01 NS19632.