Fluid intelligence and naturalistic task impairments after focal brain lesions

Classical executive tasks, such as Wisconsin card-sorting and verbal fluency, are widely used as tests of frontal lobe control functions. Since the pioneering work of Shallice and Burgess (1991), it has been known that complex, naturalistic tasks can capture deficits that are missed in these classical tests. Matching this finding, deficits in several classical tasks are predicted by loss of fluid intelligence, linked to damage in a specific cortical “multiple-demand” (MD) network, while deficits in a more naturalistic task are not. To expand on these previous results, we examined the effect of focal brain lesions on three new tests – a modification of the previously-used Hotel task, a new test of task switching after extended delays, and a test of decision-making in imagined real-life scenarios. As potential predictors of impairment we measured volume of damage to a priori MD and default mode (DMN) networks, as well as cortical damage outside these networks. Deficits in the three new tasks were substantial, but were not explained by loss of fluid intelligence, or by volume of damage to either MD or DMN networks. Instead, deficits were associated with diverse lesions, and not strongly correlated with one another. The results confirm that naturalistic tasks capture cognitive deficits beyond those measured by fluid intelligence. We suggest, however, that these deficits may not arise from specific control operations required by complex behaviour. Instead, like everyday activities, complex tasks combine a rich variety of interacting cognitive components, bringing many opportunities for processing to be disturbed.

It has long been thought that complex, open-ended tasks may capture aspects of frontal "executive" impairment that are missed in the more constrained setting of conventional neuropsychological testing. In a pioneering study, Shallice and Burgess (1991) introduced two tasks designed to mimic the open-ended character of everyday problem solving. In the 6-element task, the patient had to divide a period of 15 minutes between six different tasks, with freedom to switch tasks whenever they chose, but with several additional rules concerning task order and time allocation. In the multiple-errands task, the patient undertook a list of activities in a street of shops, again organizing the entire performance to respect a list of rules and requirements. In three frontal patients, Shallice and Burgess (1991) showed major impairment in these tasks, despite generally good performance on a battery of more conventional executive tests such as Wisconsin card sorting (Milner 1963), verbal fluency (Benton 1968) and Trails B (Reitan 1955).
In previous studies, we have investigated the link between executive tests and fluid intelligence, measured with a standard test such as the Culture Fair (Institute for Personality and Ability Testing 1973). For a number of conventional tests, including card sorting, fluency, and Trails, deficits in several patient groups are largely explained by a loss of fluid intelligence; once fluid intelligence is partialled out, performance is largely equivalent for patients and controls (Roca et al. , 2012(Roca et al. , 2013(Roca et al. , 2014. Fluid intelligence deficits are linked to damage in a distributed cortical "multipledemand" or MD network, incorporating specific regions of lateral frontal, dorsomedial frontal, insular and parietal cortex (Woolgar et al. , 2018Barbey et al. 2012); for evidence on white matter connections (see Gläscher et al. 2010). Performance on tests such as card sorting, fluency, and Trails may largely reflect the functions of this network. Findings are different for a more openended task, the Hotel, designed to mimic the Shallice and Burgess (1991) 6-element task in a more realistic setting (Manly et al. 2002). For the Hotel task, we have repeatedly found that performance is only weakly related to fluid intelligence, and partialling fluid intelligence does not remove patient deficits (Roca et al. , 2012(Roca et al. , 2013(Roca et al. , 2014. These findings suggest less specific dependence on MD functions. In the present work, we extended these prior findings using three new tests, administered along with the Culture Fair to a group of patients with focal lesions in different regions of cortex. First, we used a version of the Hotel, somewhat shortened compared with previous versions. Second, we designed a new test of everyday comprehension and problem-solving, based on short vignettes describing real-life situations and their accompanying decisions. Third, we designed a new taskswitching test to mimic just one aspect of the complex processing requirements of the Hotel. It has often been suggested that, in this test, patients may fail to divide their time between component sub-tasks because, having become immersed in one, they forget the larger requirement to give some time to all of them (Manly et al. 2002). To investigate this kind of immersion as a possible key factor in naturalistic behaviour, we modified a standard task-switching paradigm (Rogers and Monsell 1995) to manipulate the length of time before a task switch. Compared to our other two tasks, this one focused on a quite specific cognitive requirement, but one that we thought might be important in temporally extended, open-ended behaviour.
As predictors of cognitive impairment, we considered damage to two a priori networks. The first was the MD network, predicted to be important for fluid intelligence but less so for the other tasks.
The second was the default-mode network (DMN). In functional brain imaging, the DMN is well known as a set of brain regions with strong functional connectivity (Yeo et al. 2011), deactivation in many tasks compared to rest (Shulman et al. 1997), but positive activation linked to mind-wandering (Mason et al. 2007;Christoff et al. 2009), self-related thought (D'Argembeau et al. 2005), and sometimes large externally-directed task switches (Crittenden et al. 2015;Smith et al. 2018). As regards the Hotel, one specific reason for suspecting DMN involvement came from the previous results of Roca et al. (2010Roca et al. ( , 2011, who linked Hotel deficits to anterior frontal lesions. On the medial surface, anterior frontal cortex includes a large region of the DMN, and though Roca et al. (2011) combined lateral and medial patients, it seemed possible that the DMN component was responsible for Hotel deficits. More broadly, a large body of imaging work links the DMN to both social cognition and imagination of cognitive scenes or episodes (Frith and Frith 2003;Amodio and Frith 2006;Addis et al. 2007Addis et al. , 2009Andrews-Hanna et al. 2010). Several authors have suggested that situation representations in the DMN place ongoing cognition in a broader context (Hassabis and Maguire 2007;Ranganath and Ritchey 2012;Smith et al. 2018). Reference to a larger context could be especially important in naturalistic problem-solving, including behavioural management over an extended time period. For each patient, we measured volume of damage in MD, DMN and other cortical regions, and linked these lesion measures to behavioural impairment.
A different possibility is that naturalistic task deficits do not reflect specific control requirements associated with complex behaviour. Rich, varied tasks based on the requirements of everyday cognition are likely by definition to have many different cognitive components, dependent on multiple brain processes (Burgess et al. 2000). These components, furthermore, are unlikely to act independently; if one component is weakened through brain damage, it may send delayed or inaccurate input to others, or compete for mental resources. Of course, all tasks are potentially influenced by multiple sensory, motor and cognitive functions, but for rich tasks, it may be especially unlikely that one or a few core processes are responsible for the bulk of neuropsychological impairments. On this account, complex tasks are highly sensitive to brain damage simply because there are many opportunities for their processing to be disturbed. The prediction would be deficits that are not well explained by any one focal cognitive impairment, or by damage to any one cortical region or network. Patients were selected on the basis of having chronic, focal lesions from mixed aetiology excluding traumatic brain injury, and aged between 18-80 years old. There were no other formal exclusion criteria for region of lesion or specific cognitive deficit. Two patients were unable to complete more than 1 task and were excluded from further analysis. Demographic and lesion information for the remaining 32 patients is presented in Table 1. 30 non-brain-damaged control participants were also tested. To ensure good correspondence between patient and control groups on factors of age and years of educations, data for the patient group was collected first and control participants were then selected to approximate the patient distribution. Accordingly, patient and control groups did not differ significantly in age (patients mean =58.4 years, SD=15.3; controls mean=52.6, SD=18.9; t(61)=1.35, p=0.18) or years of education (patients mean=13.9, SD=2.3; controls mean=14.7, SD=3.9; t(61)=1.04, p=0.30). Testing Participants were given a battery of computer-based and other tasks described below. The test battery was completed in a single session lasting around 90-120 minutes. The battery consisted of 7 tasks, presented in fixed order. In this paper we present data for Culture Fair (presented first in the session), Situations (second), Hotel (fourth and seventh; see below), and Switch Time (sixth). The remaining three tests were included for other purposes, not concerned with naturalistic decision making. They were a more conventional task switching paradigm (presented third), adapted from Smith et al. (2018), an object in place concurrent discrimination memory task (presented fifth), similar to Gaffan (1994), and a comparison of attention control by scene or object cues (presented last). Four patients did not complete the Switch Time task due to fatigue. Additionally, for one of these four patients, the battery was split into two shorter sessions. Computer-based tasks were given on a Dell 1280x1024 resolution laptop, controlled using Psychophysics Toolbox for MATLAB (Brainard 1997).

Culture Fair
Participants were given the standardised version of the Cattell Culture Fair, Scale 2 Form A (Institute for Personality and Ability Testing 1973) consisting of four timed subtasks (series completion, odd one out, matrices, topological relations). At the start of each subtask, the experimenter read aloud the rules to the participant and went through 3 examples with them. Total correct scores were calculated and then converted into IQs from the standardised table of norms. To match the rest of the data set, IQ scores were inverted such that higher numbers would correspond to poorer performance.

Situations
The Situations task was designed to test social and non-social decision-making in real-life vignettes.
During the task, participants were shown 12 short stories on the computer screen, and after each story asked one social judgement question, one emotion judgement question and one executive judgement question. An example story and question set is presented in Figure 1. The full set of stories and questions is presented in the Supplementary Materials. For each item, participants were first shown the story text and asked to read through the story. After reading, participants were asked to press a button. With the story text still present on the screen, the questions were then presented one by one, along with 3 possible answers each. The answer options were designed such that one was correct, one was very incorrect and the third was plausible but less optimal than the correct option. Participants were asked to respond as quickly as they could using buttons "1", "2" or "3" on the keyboard, corresponding to the 3 possible answers. The order of story presentation was randomised and the order in which the questions were presented was pseudorandomised such that each type of question (social, emotional and executive) was presented equally often first, second and third. The position of the correct answer was counterbalanced across question types such that it appeared in positions 1, 2 and 3 on an equal number of trials.
We scored proportion error and median response time (RT) for correct trials. RTs for the three different question types were strongly correlated for both patients (for the three possible pairs of question types, r = .89 to .95) and controls (.49 to .71). Proportion error scores for the three question types were also correlated in patients (.22 to .59), though not in controls (-.03 to .11), perhaps in part because error proportions were low. As overall measures of performance, we averaged RTs and proportion error scores across the three question types.

Hotel
The Hotel task used materials laid out on a table in front of the participant. An example of the task layout is presented in Figure 2. Participants were asked to imagine they were in a job interview for a position at a hotel and were asked to try 3 different hotel activities, each one involving sorting a stack of sheets of paper. Participants were told that they would have 9 minutes to work on the three activities and that it would be impossible to finish any of them completely in the time limit. Instead, they should ensure that some time was allotted to each activity. To keep track of the time participants were given a clock. Throughout the task the clock was turned away from the participants, but participants could choose to check it at any time before returning it to its backward-facing position.
Participants were given two variants of the task, which varied the form of periodic interruptions. The order of the two variants was counterbalanced across participants, with two other tasks performed between the two variants. Interruptions were motivated by the work of Manly et al. (2002), who found that performance improved when patients were given an occasional auditory alert designed to break focus on the current activity and reorient attention to the overall goal. In our version, interruptions were designed to reorient participants' attention either within the current activity (internal interruption) or to the external environment (external interruption). In the internal interruption condition a yellow sheet of paper was placed after every 7 task items in each activity.
Participants were told to place the yellow item in the same pile as the previous item. In the external interruption condition a written instruction, asking participants to perform an action directed towards an aspect of the surrounding environment, was placed after every 7 task items. Participants were asked to follow the written instruction (e.g. point to a window) and then place the instruction sheet to the side.
For the two task versions, there were two separate sets of 3 activities. The activity set paired with each interruption condition was counterbalanced across participants. Set A consisted of sorting conference name tags by alphabetical order, sorting invoices into piles according to the vendors, and sorting bills into piles according to customer name. Set B consisted of sorting staff name tags by alphabetical order, sorting restaurant lists into piles by their location, and sorting spa receipts into piles by treatment.
During the task, the experimenter kept a continuous record of which activity the participant was working on, using computer keys to indicate each time an interruption was encountered and each time the activity was switched, along with which new activity was begun. In line with previous work (Manly et al. 2002;Torralva et al. 2009;Roca et al. 2010), the score was the summed deviation from optimal time (180 seconds per activity) across the 3 activities. Preliminary analysis showed no differences between interruption types. However, there was evidence that participants (especially controls) improved their strategy when performing the task for the second time. Accordingly, scores were based just on the first version performed, whichever interruption type it involved.

Switch Time Task
One difference between the Hotel task and other classical switching tasks is the long period (~180s) between switches. The Switch Time task was designed to test whether patients are particularly impaired at task switching after longer rather than shorter periods. As noted above, only 28 patients took part in this task. Task events are illustrated in Figure 3. For each trial, participants were presented with a picture to the left of the screen and a word with a letter missing on the right of the screen. Participants were required to make yes/no judgements on one of these stimuli based on a task rule cued by a central shape. If the rounded corners of the cue pointed towards the left, then participants were asked to do the picture task. If the rounded corners of the cue pointed towards the right, then participants were asked to do the word task. The two tasks used were taken from Crittenden et al. (2015). For pictures, the decision was whether the item would fit in a shoebox; for words, it was whether addition of a letter 'a' in the blank position would create a real word.
Participants were asked to make a "yes" or "no" response, by left or right keypress respectively.
Performance was self-paced, with the stimuli remaining until a key was pressed, and participants were asked to respond as quickly as possible without making mistakes. An inter-trial interval (ITI) of 1.5 seconds followed each response.
Pilot studies revealed that switch performance systematically declined with increasing numbers of repeated trials prior to a cue change, reaching an asymptote at around 12 repeats. Accordingly, trials of the same task were repeated in blocks of 4, 6 or 12 trials, with block length presented in a pseudorandom order. Trials could thus be sorted into 4 transition types: stay trials (task trials preceded by the same task), switch 4 trials (task switch after 4 of the same task), switch 6 trials (task switch after 6 of the same task), and switch 12 trials (task switch after 12 of the same task). The experiment consisted of 1 run of 182 trials. The run contained 4 blocks of each length for each task, plus a final run of 6 trials after the last switch. A break was inserted into the run, approximately halfway through the trials, immediately after a switch trial. After the break, the task block continued with the same task cued as in the switch before the break. The last trial before the break and the first trial after the break were included for purposes of calculating run length before the next switch. Photoshop. By using only a subtle difference between the two sides of the cue, we aimed to avoid a salient visual change that would alert participants to task switches.
There were equal numbers of "yes" and "no" trials for each combination of task and switch type. At the task switch trial and the first stay trial after a task switch, correct answers for the word task and the picture task were always different. For all other stay trials, the correct response for the cued task was randomised so that it was the same as the uncued task on half of the trials and different on the other half of the trials.
Before the start of the task, participants were shown 2 trials on paper and asked which task they would do given the cue shape and what button they would press in response to the cued stimulus.
Participants continued with the main task only when the experimenter was satisfied that the participant understood the instructions. No subject needed more than two repetitions of the instructions before continuing with the main task.
Scores concerned proportion error and median correct RT on switch trials. To capture the effect of time since last switch, scores were proportion error on switch 12 trials minus proportion error on switch 4 trials, and similarly RT on switch 12 minus RT on switch 4. Neuroradiological Assessment MRI T1 and T2 structural scans were acquired for all patients. Lesions were traced on structural images by a neurologist, blind to the experimental results, using MRIcron (Rorden and Brett 2000) before normalising to MNI space using SPM software (Wellcome Department of Imaging Neuroscience, London, England; www.fil.ion.ucl.ac.uk) with cost-function masking to mask the lesion from the calculation of the normalization parameters (Brett et al. 2001).

Regions of Interest
Volumes for the MD network, DMN and all other grey matter regions were constructed in the following steps. The resulting MD network (red) and DMN (blue) regions of interest (ROIs) are presented in Figure 4a. Figure 4b shows patient lesion overlap.
The As the MD volume AI/FO and the DMN volume vmPFC showed slight overlap, the region of overlap was removed from both ROIs.
The Other region was created using custom scripts for SPM 12 (Wellcome Department of Cognitive Neurology, London, UK). First a whole grey matter volume was created by concatenating all grey matter regions included in the AAL Atlas (Tzourio-Mazoyer et al. 2002). Then grey matter included in the MD or DMN ROIs, and all grey matter 5 mm or less from these volumes, was excluded. An additional occipital-temporal region is sometimes associated with the MD network but also strongly related to visual processing, and therefore this region (Figure 4a, yellow) was also excluded.
Remaining grey matter was assigned to the Other ROI. For each patient, volumes of damage were separately measured in MD, DMN and Other ROIs.

Differences between patients and controls
Scores for each task are shown in Table 2 (left), separately for patients and controls. One-way analyses of variance (ANOVAs) showed that patients performed significantly worse than controls on all but one measure (  Figure 5, with regression lines for each group calculated from the ANCOVA and constrained to have the same slope in the two groups. As a direct comparison with Roca et al. (2010) these analyses were re-run for the 14 patients (12 for the Switch Time task) whose lesions were restricted to the frontal lobe. Results were similar to those obtained in the full group, except that, in the ANCOVA, the significant group difference for Situations, proportion error was removed.
For Switch Time, our primary scores concerned the effects of preceding block length on switch trials (difference between switch 12 and switch 4 trials). In this task, however, we note that patients also showed substantial impairments even on stay trials (mean proportion errors 0.06 and 0.04 respectively for patients and controls, p < 0.08; median RT 2.40 and 1.57 sec, p < 0.004).

Effects of lesion volume
The next analysis examined the relationship between behavioural scores and lesion volumes. This analysis was restricted just to the patient group. For MD, DMN and Other ROIs, the mean (standard deviation) volumes of damage were 8.39 (7.32), 8.43 (11.15) and 10.65 (7.76) ml respectively.
Across patients, volumes of damage in the 3 ROIs were close to independent (maximum correlation 0.14).
Following Woolgar et al. (2010Woolgar et al. ( , 2018, we asked first whether Culture Fair IQ was predicted by volume of MD lesion. Scatterplots relating IQ to ROI lesion volumes are shown in Figure 6. To account for multiple comparisons (ROIs), significance threshold for correlations was set to p<0.017, one-tailed. Consistent with Woolgar et al. (2010Woolgar et al. ( , 2018, patient IQ was found to be significantly related to MD lesion volume (r=0.53, p<0.001). There was no relationship to DMN lesion volume (r=-0.21, p=0.88), and only a weak association with Other lesion volume (r=0.35, p=0.03). Total lesion volume was also not significantly related to IQ (r=0.19, p=0.15).

Figure 6. Scatterplots relating Culture Fair IQ to MD, DMN and Other lesion volumes (ml).
As 9 of the current patients were also tested in Woolgar et al. (2010), the correlation between Culture Fair IQ and MD lesion volume was re-calculated with those 9 patients excluded. Results for this independent patient group replicated the Woolgar et al. (2010) findings (r=0.39, p=0.03).
Similar analyses were then carried out for naturalistic task scores. The results are shown in Table 3.
No measure was significantly related to lesion volume in either MD or DMN regions. For the RT score (switch 12 -4) from the Switch Time task, there was a significant correlation with lesion volume in the Other ROI. The naturalistic measures were also unrelated to total lesion volume. shown with an asterisk (*).

MD DMN Other
Hotel ( To test whether the naturalistic tasks were related to each other, between-task correlations were performed separately in patient and control groups. Results are shown in Table 4

Hemispheric asymmetry
To check for differential involvement of left and right hemispheres in our naturalistic tasks, we compared performance in patients with left (n = 11) and right (n = 18) hemisphere lesions (see Table   2). For naturalistic tasks there was no evidence of hemispheric differences (maximum t = 0.99), contrasting with worse performance for the right hemisphere group in the Culture Fair, t = 2.39, p < .02.

Residual patient impairment
In a final examination of lesion effects, we calculated average performance across naturalistic tasks, after accounting for effects of fluid intelligence. For each patient, standardized residual scores, calculated from the above ANCOVAs, were derived for all 5 naturalistic-task scores and then averaged. Signs were set such that high scores reflected poorer performance than predicted from Culture Fair IQ. For the four patients who had not completed the Switch Time task, the average residual score was generated from Hotel and Situations scores. After correcting for multiple comparisons, the average residual was not correlated with volume of damage in any ROI, with the strongest correlation for Other (r=0.31). Figure 7a shows the lesions of the 6 patients with the highest average residual scores, reflecting naturalistic task performance worse than predicted from fluid intelligence. Rather than implicating a specific brain region, the figure illustrates the diversity of lesions associated with naturalistic task impairment.
In order to compare with Roca et al. (2010Roca et al. ( , 2011 we recalculated patient residuals after restricting the patient sample to patients with frontal lesions only. Figure 7b shows the lesions of the 4 frontal patients with the highest average residual scores. Unlike Roca et al. (2010), residual impairment after accounting for IQ was not restricted to the anterior PFC but spread across the frontal lobe. It should be noted however, that this analysis was done on a small sample of 14 patients only.
As the Switch Time RT score was significantly related to damage in the Other ROI, we conducted a final similar analysis to search for a common lesion location among patients with the greatest deficits on this score only. The analysis was the same as for average residual scores (Figure 7a), but this time using only the Switch Time residual. Again the results showed scattered brain lesions among the 6 patients with the greatest deficit ( Figure 7c).

Discussion
In previous work, we have found that "executive" tasks vary widely in the degree to which deficits are explained by fluid intelligence (Roca et al. , 2012Woolgar et al. 2010). For several classical tests, such as Wisconsin card sorting, impairments in diverse patient groups are eliminated when fluid intelligence is partialled out. The Hotel task of Manly et al. (2002), however -based on the 6element task of Shallice and Burgess (1991) -shows a very different result. For this task there is only a weak correlation with fluid intelligence, and patient impairments are not removed by partialling out this effect. These results match the long-held belief that complex, naturalistic tasks can reflect aspects of cognitive deficit that are missed in classical tests.
In the present work, we confirmed these findings for a new version of the Hotel task. Again, deficits in a diverse group of patients with focal brain lesions were not accounted for by fluid intelligence.
Results were similar in a second test of understanding and decision-making in complex, life-like scenarios. In a third test, we attempted to isolate one critical factor in the Hotel task and potentially other real-world situations -the need to break out from a lengthy period of immersion in a single task. While task switching in the Hotel is spontaneous, with no external cue, we used an explicit switch cue but varied the length of the previous task block. Here too we observed deficits in the patient group that were not explained by fluid intelligence. Though fluid intelligence may account for deficits in many "executive" tasks, our results extend the list of more naturalistic tasks for which this is not true.
Also replicating previous work, we found that deficits in fluid intelligence were predicted by the extent of lesions to the frontoparietal MD network. MD lesions, however, were not strong predictors of deficits in the naturalistic tasks. While MD lesions may explain many aspects of classical executive deficit, again these data suggest a different explanation for naturalistic deficits.
As a second potential predictor of naturalistic deficits, we considered lesions to the DMN. Influential accounts link the DMN to representation of broad cognitive contexts, including spatial, temporal and social aspects (e.g. Ranganath andRitchie, 2012, Andrews-Hanna et al., 2010). Plausibly, broad contextual representations will be especially important in temporally extended, open-ended behaviour. As we found for the MD network, however, naturalistic impairments were not strongly predicted by DMN lesions.
Instead, we found that naturalistic deficits could arise from lesions scattered through multiple regions of the brain, in left or right hemispheres, in frontal, parietal or occipito-temporal cortex, and including or not sections of MD and DMN networks. Though performance was not significantly associated with either MD or DMN lesions, our interpretation is not that these networks make no contribution to naturalistic behaviour. Rather, no one network is strongly predictive of deficit simply because deficits can arise from lesions of many different kinds.
Our results for Hotel and Situations are firmly in line with the finding that complex, naturalistic tasks capture cognitive deficits missed in many executive tests, but suggest that these tests will be poor at isolating any specific cognitive process. In line with the arguments of Burgess et al. (2000), complex, real-life decision making rests on many processes, likely dependent on many different cortical regions. The particular processes that are critical may vary between tasks, matching our finding of generally nonsignificant correlations between them. Perhaps not surprisingly, complex, real-life decision-making may depend on much of the cortex, making these tests highly sensitive to brain lesions, but not highly diagnostic of any specific cognitive deficit.
For Switch Time this argument is less clear, since the test is ostensibly much more focused in its cognitive demands. Results however were similar, with deficits not explained by fluid intelligence, and unrelated to volume of damage in either MD or DMN regions. For this test there was a significant correlation with damage to the residual, Other ROI, but given widely scattered lesions in the most impaired patients (Figure 7c), along with differences in total size of MD, DMN and Other regions this result should be interpreted with caution. Further work would be needed to analyse deficits in this task. By design, the cue indicating a change of task had low visual salience, perhaps requiring participants to maintain a sustained attentional set for its possible occurrence. The duration of a block of 12 trials depended on individual RTs, with typical durations around 40 s for controls, but appreciably greater for patients given slow responses even on stay trials. The data suggest that escape from a period of immersion in one task indeed calls on cortical functions unlike those captured in fluid intelligence and, by extension, other classical executive tests. Again, however, deficits may have mixed causes, not simply related to specific lesion locations.
In principle, the source of a deficit in any one task and patient might be clarified by extensive neuropsychological testing. Given the low correlations between our naturalistic tasks, however, it seems unlikely that any small set of cognitive processes will be broadly predictive of "naturalistictask" deficits. Instead, we suggest that each such task has its own, rich cognitive profile, likely impaired for different reasons in different patients.
On this interpretation, naturalistic tasks do not identify specific control deficits required only in complex, open-ended behaviour. Instead, their sensitivity to brain damage arises simply through the many cognitive processes that must be combined, bringing many opportunities for performance impairment. In line with the original work of Shallice and Burgess (1991), such tests may be especially useful as an indication of difficulties that patients may face in return to everyday life. To predict such difficulties, clinicians should use not only measures that target specific cognitive abilities, such as classical executive tasks which are associated with lesion to the MD system ), but also tasks -like the ones described here -that tackle more complex, naturalistic scenarios. This way, we can provide a more comprehensive cognitive assessment which could better reflect real life deficits. Regarding neuropsychological rehabilitation, the findings shed light on the complicated task that neuropsychologists face when dealing with brain lesions. As clinical experience shows, difficulties in returning to everyday life are not easily predicted from damage to specific brain regions or networks, reflecting the complexity of cortical involvement in everyday behavioural management and decisions.
Michael´s son Trevor had been asking him to teach him how to drive for a very long time now. When Trevor turned 19 and got his first job Michael thought it was time and started teaching him with his brand new Toyota Corolla. Trevor was a very good driver, respectful and prudent. On their way back from Trevor's driving test, Trevor asked Michael if he could drive the last block. Michael agreed and changed places with him. As soon as Trevor started driving a dog appeared, he tried to dodge it and crashed into his neighbor's car. Michael could not immediately afford the cost of repairs, and had been planning to drive friends to Edinburgh in two weeks' time.