[1] 12w High functioning children with autism spectrum disorder: A novel test of multitasking

[1] 119w High functioning children with a diagnosis of autism or Asperger's syndrome (HF-ASD) often experience diYculties organising goaldirected actions in their day-to-day lives, requiring support to schedule daily activities. This study aimed to capture these everyday diYculties experimentally using multitasking, a methodology that taps into the cognitive processes necessary for successful goal-directed activities in everyday life. We investigated multitasking in children with HF-ASD using a novel multitask test, the Battersea Multitask Paradigm. Thirty boys participated in the study, 14 with HF-ASD and 16 typically developing controls, matched for age and IQ. Group diVerences in multitasking were observed. Participants with HF-ASD were less eYcient at planning, attempted fewer tasks, switched inXexibly between tasks and broke performance rules more frequently than controls.

[1] 19w "I bet he'll become a rocket scientist, but I'll probably have to dress him and drive him to work."

[2] 229w Mother speaking of her son with high functioning autism, cited in OzonoV, Dawson, and McPartland (2002, p. 18). This mother's comment accurately summarizes the paradox faced by many high functioning children with autism spectrum disorder (HF-ASD). By deWnition, children with high functioning autism have normal or above normal intelligence and relatively well developed structural language and cognitive skills (OzonoV et al., 2002). However, even though some children with HF-ASD achieve milestones such as forming a career or getting a university degree, they con-tinue to have diYculties with the demands of everyday life and may struggle to live independently as adults (Howlin & Goode, 1998). One of the reasons that individuals with HF-ASD Wnd it hard to live independently is because they have diYculties organising and coordinating everyday activities. Children with HF-ASD are commonly reported to have diYculties with time management, organising the materials necessary to perform an activity and sequencing activities; generally reXecting a deWcit in the ability to plan ahead (OzonoV et al., 2002). This impacts upon day-to-day life: at school children can fall behind in class due to poor time management and diYculties organising their workload, homework is all too often left at school instead of being brought home. At home, activities of daily living such as getting dressed or getting ready for bed take longer to perform, often leading to frustration on all sides (OzonoV, 1998).

[3] 287w The question arises of how to capture these everyday problems experimentally. A number of studies have investigated executive control processes in children with HF-ASD. This research has consistently identiWed executive function (EF) impairments in individuals with autism (OzonoV, 1998;Pennington & OzonoV, 1996). The executive proWle of children with ASD is one of 'high level' diYculties (Hughes, 2001). Executive deWcits in ASD are typically more pronounced than those observed in other developmental disorders (Pennington & OzonoV, 1996) and may occur across a range of domains of EF (Geurts, Verte, Oosterlaan, Roeyers, & Sergeant, 2004). Planning and cognitive Xexibility stand out as areas of EF that present particular diYculties for individuals with ASD. Children with ASD plan poorly on tasks such as the Tower of Hanoi (TOH) relative to both clinical (e.g., children with Attention DeWcit Hyperactivity Disorder and children with Tourette's syndrome) and typically developing control groups (Hughes, Russell, & Robbins, 1994;OzonoV & McEvoy, 1994;OzonoV, Pennington, & Rogers, 1991). On tests of cognitive Xexibility, participants with ASD demonstrate impaired cognitive Xexibility (OzonoV et al., 1991;OzonoV, Strayer, McMahon, & Filloux, 1994;Prior & HoVman, 1990;Szatmari, TuV, Finlayson, & Bartolucci, 1990), engage in highly perseverative and inXexible strategies (Hughes et al., 1994) and show impaired performance when shifting response set (OzonoV et al., 1994). In comparison, deWcits in inhibitory control are less pronounced in ASD, but may depend upon the measure used to assess inhibitory skills. Participants with ASD often perform as well as controls on traditional tests such as the Stroop test (Eskes, Bryson, & McCormick, 1990;OzonoV & Jensen, 1999;Russell, Jarrold, & Hood, 1999). However, inhibitory dysfunction in ASD has been reported in studies employing diVerent measures such as Go-NoGo paradigms (Geurts et al., 2004;Nyden, Gillberg, Hjelmquist, & Heiman, 1999).

[4] 105w Although it is probable that this proWle of executive dysfunction has a signiWcant impact upon the everyday lives of children with HF-ASD and their families, few studies have sought to measure this eVect. Executive dysfunction in children and adults with HF-ASD has been shown to correlate signiWcantly with measures of adaptive behaviour (OzonoV et al., 2004). Performance on the Tower of London task relates to communication symptoms in school age children with autism (Joseph & Tager-Flusberg, 2004). Poor cognitive Xexibility may be related to the everyday repetitive behaviours that characterize ASD (Hughes, 2001), however this relationship has not consistently been reported (Joseph & Tager-Flusberg, 2004).

[5] 74w It is possible that so little evidence exists to support relationships between EF and everyday diYculties in ASD because many tests of EF involve planning or solving a single problem within highly structured, clearly deWned limits. In contrast, multitask tests assess an individual's ability to organise and coordinate the performance of multiple activities in a more Xuid environment which is more representative of everyday life (Burgess, Veitch, Costello, & Shallice, 2000;Shallice & Burgess, 1991).

[6] 113w In a multitask test, the participant is required to perform a number of tasks within a given time period. The tasks are interleaved, meaning that they cannot be performed sequentially. Success is constrained by a set of rules which typically restrict the order in which tasks can be performed. These time and rule-based constraints emulate practical restrictions placed upon the organisation of multiple activities in everyday life, such as performing an activity at or within a certain time or performing one activity in advance of another. Indeed, multitask tests have a high 'ecological validity' as test performance reXects real life diYculties (Alderman, Burgess, Knight, & Henman, 2003;Burgess, Alderman, Evans, Emslie, & Wilson, 1998).

[7] 102w Adult neurological patients with frontal lobe damage can demonstrate signiWcant impairments organising activities in their day-to-day lives. For example, Shallice and Burgess (1991) report a patient who shopped for food sequentially, returning to his car after purchasing each individual item, because coordinating buying multiple items at one time was too challenging. Such patients perform poorly on multitask tests, demonstrating poor time management (i.e., spending too long on one task), failing to attempt all tasks assigned (despite being aware of the requirement to do so), breaking the rules and carrying out subtasks incorrectly (Burgess & Shallice, 1996;Burgess et al., 2000;Shallice & Burgess, 1991).

[8] 144w The key diYculty of these patients is an impaired ability to create and activate delayed intentions (Burgess et al., 2000). In a multitask test, multiple intentions (to perform multiple tasks) are created, but the execution of the majority of these intentions must be delayed, as it is not possible to perform all the tasks simultaneously. Moreover, during this delay, attention is focused on another activity (the current task) rather than the 'to-be-performed' (delayed) tasks. When an intention is delayed, an 'intention marker' must be created. When this marker is subsequently activated, it 'brings to mind' the intended action and switches the focus of attention to performing the intended task. These processes of marker formation, activation, and intention execution are believed to be impaired in adult frontal lobe patients who perform poorly on multitask tests and in their everyday lives (Burgess, 2000;Burgess et al., 2000).

[9] 216w Recent investigations into the cognitive processes involved in multitasking have placed the organisation of prospective actions Wrmly within the context of the executive control of behaviour (Burgess et al., 2000;Kliegel, Martin, McDaniel, & Einstein, 2002). The ability to create and activate delayed intentions has been deWned as prospective memory, PM (Burgess et al., 2000;Einstein & McDaniel, 1996;Ellis, 1996). Successful prospective remembering is inXuenced by retrospective mnemonic processes and various executive functions (Burgess et al., 2000;Kliegel et al., 2002;Shallice & Burgess, 1996). Retrospective memory is important for storing the content of an intended action. We not only need to remember that we intend to do something, we must also remember what it was that we intended to do. Planning is the EF most involved in the creation of delayed intentions, and the success with which an intention is executed is inXuenced by the quality of the plan through which it was set up (Gollwitzer, 1999). Switching attention from a current task to the intended task requires inhibiting attention to the current task and shifting attention to the intended task, thus involving executive functions such as inhibitory control and cognitive Xexibility (Kliegel et al., 2002). Each of these executive functions has been identiWed as dysfunctional in HF-ASD, providing a strong cognitive basis for investigating multitasking in this group.

[10] 194w Parallels can be drawn between the everyday organisation diYculties reported in adult patients with frontal lobe damage and the problems children with HF-ASD experience in their daily lives. Indeed, multitasking is strongly associated with the prefrontal cortex, and atypical frontal lobe development has been proposed to play a role in autism (Pennington & OzonoV, 1996). The frontal lobes are extensively connected to other regions of the brain, and hence are well placed to co-ordinate and sequence complex cognitive activity (e.g., Fuster, 1989Fuster, , 1998;;Stuss & Benson, 1987). The ability to organise multiple activities has been associated with the prefrontal cortex in behavioural studies (Bechara, Tranel, & Damasio, 2000;Burgess, 2000;Eslinger & Damasio, 1985;Fortin, Godbout, & Braun, 2002;Shallice & Burgess, 1991), imaging studies (Burgess, Quayle, & Frith, 2001;Yamadori et al., 1997) and neurophysiological studies (Leynes, Marsh, Hicks, Allen, & Mayhorn, 2003;West, Herndon, & Ross-Munroe, 2000). Regions which have consistently been highlighted include the left ventromedial prefrontal cortex and the right dorsolateral prefrontal cortex (Brodmann areas 8, 9, 46, and especially 10). Roles attributed to these regions include maintenance of intention over delay, switching attention from the current to the intended action, intention retrieval, and intention execution.

[11] 289w Observations that the cognitive and behavioural impairments found in autism resemble those of adult patients with frontal lobe injuries led to the hypothesis that atypical frontal lobe development plays a causal role in autism (e.g., Damasio & Maurer, 1978;Pennington & OzonoV, 1996;Rogers & Pennington, 1991). There is evidence to support atypical structural and functional frontal lobe development in autism (see Volkmar, Lord, Bailey, Schultz, & Klin, 2004 for a review). Maturation of the frontal cortex may be delayed in autism (Zilbovicius et al., 2004). Increased white matter in the whole brain has been reported in infants and toddlers with autism (Courchesne, 2002), and it has been suggested that this reduces overall functional integration and connectivity (Horwitz, Rumsey, Grady, & Rapoport, 1988). Reduced dopaminergic activity (Ernst, Zametkin, Matochik, Pascualvaca, & Cohen, 1997) and increased levels of serotonin (Chugani et al., 1997) have also been reported in the prefrontal cortex of individuals with autism. The dorsomedial prefrontal cortex may be critical for social cognition, and its disruption may lead to the social impairments observed in autism (Schultz et al., 2000). Likewise the ventromedial prefrontal cortex forms part of the brain system claimed to be specialized in social processing and emotional learning (Dawson, MeltzoV, Osterling, & Rinaldi, 1998;Rolls, 1990;Stone, Baron-Cohen, & Knight, 1998). The dorsolateral prefrontal cortex (DLPFC), in particular, has been closely linked to multitasking (Burgess et al., 2000); reduced activation in this region has been reported when individuals with autism perform cognitive tasks known to be dependent upon this region (Luna et al., 2002). The reliance of the organisation of multiple activities on the prefrontal cortex, together with evidence supporting the atypical development of this region in autism, provides a sound neuroanatomical basis for investigating multitasking in children with HF-ASD.

[12] 97w In sum, children with ASD have diYculty organising their everyday activities and typically need a lot of support to do this eVectively (OzonoV et al., 2002). Multitasking is an experimental test of everyday organisational skills which has high ecological validity (Burgess et al., 1998) and is dependent upon the functioning of the prefrontal cortex (Burgess et al., 2000(Burgess et al., , 2001)). The cognitive processes underlying multitasking include executive functions, so it would seem to be a good methodology to use to investigate everyday organisational diYculties in children with ASD, which have hitherto rarely been investigated experimentally.

[13] 145w To the best of our knowledge, only one investigation touching on multitasking in ASD has been reported. This was done as part of the validation of a clinical test battery designed to measure executive dysfunction in children: the Behavioural Assessment of the Dysexecutive Syndrome for Children (BADS-C, Emslie, Wilson, Burden, Nimmo-Smith, & Wilson, 2003). A group of children who had been referred for clinical assessment and had a variety of clinical diagnoses, including 13 children with ASD, were compared to typically developing children and children with clinical diagnoses who had not been referred for clinical assessment. As part of the BADS-C, children performed the 'Six Part Test' in which they had to attempt six tasks in 10 min with the order of task performance limited by rules. The general results pointed to performance deWcits in children with ASD, but no detailed analyses were carried out.

[14] 64w The aim of the present study was to investigate in depth the presence of multitasking impairments in high functioning children with ASD. We designed a novel paradigm to assess multitasking skills in children with HF-ASD and matched controls. Our paradigm measures several of the cognitive processes contributing to multitasking, enabling us to investigate which aspects of multitasking present particular diYculties for children with HF-ASD.

[1] 72w The groups were well matched for measures of IQ. Full Scale: HF-ASD mean D 105.6 (SD D 12.9), TD mean D 109. 1 (SD D 12.9). Verbal: HF-ASD mean D 105.8 (SD D 17.4), TD mean D 110.2 (SD D 16.4). Performance: HF-ASD mean D 104.8 (SD D 11.8), TD mean D 105.4 (SD D 11.2). One-way analyses of variance revealed no signiWcant diVerences between the groups on any of these measures.

[2] 600w The BMP generates six composite variables. Group mean scores and standard deviations for each of these variables are detailed in Table 1. Group comparisons were analysed using non parametric Mann-Whitney U tests, except for the plan follow variable where Analysis of Covariance (ANCOVA) was used to co-vary for the eVect of planning. Participants in both groups were equally able to learn and remember the rules of the multitask paradigm, as indicated by the mean scores in Table 1. No signiWcant diVerences were observed for rule learn (z D ¡0.90, ns) or rule memory (z D ¡1.86, ns). There was a signiWcant group diVerence in terms of planning how to perform the BMP (z D ¡2.93, p D .01); participants with HF-ASD produced less complex plans than TD controls. This diVerence did not transfer to plan implementation as no group diVerences in plan follow were observed. Note that as the plan follow variable represents the extent to which a participant enacted their original plan, it is yoked to the planning score. Hence group comparisons of plan following are considered after covarying for the eVect of planning group diVerences in planning (ANCOVA): F (1, 27) D 2.86, ns; estimated marginal means HF-ASD mean D 9.48 (standard error, SE D .46), TD mean D 9.70 (SE D .42). These results indicate that despite the seemingly large group diVerences in the means for plan following, children in both groups were equally able to implement the plan they had formed. Children with HF-ASD performed the multitask paradigm less eVectively than The multitask variable perform is a composite score and we speculated that an examination of the sub-scores contributing to this composite might illuminate which aspects of multitask performance present diYculties for children with HF-ASD. In terms of: (a) The 'number of tasks attempted'-three children with HF-ASD failed to perform all three tasks (mean D 2.71, SD D 0.61) compared to controls who invariably attempted all three tasks (mean D 3.00, SD D 0.00). This diVerence narrowly missed being signiWcant: Mann-Whitney U, z D ¡1.92, p D .055. (b) The 'strategic performance' score reXects how well children applied the rules of the paradigm to gain points during performance. No signiWcant group diVerences were observed for this measure: HF-ASD mean D 13.28 (SD D 4.03), TD mean D 15.37, (SD D 2.27), Mann-Whitney U, z D ¡1.44, ns. (c) Scores for 'penalty performance' indicate that only one child in each group made a placement error. However, 7 of 14 children with HF-ASD broke the rules of the paradigm compared to 2 of 16 TD participants. Fisher's exact test was used to calculate group diVerences in task switching as values in one cell were less than Wve and this diVerence was signiWcant, p D .032. (d) The 'task switch performance' score measures the eVectiveness with which participants switch Xuently between tasks. The most eVective way to perform the interleaved tasks of the BMP is to weigh up performance options across tasks (i.e., to identify the smallest yellow items to Wll in all three tasks). Therefore, moving sequentially from task 1-2-3 only represents an ineVective performance strategy, and participants who employ this strategy are penalized, despite having attempted all three tasks. We hypothesized that children in the HF-ASD group would switch between tasks less eYciently than children in the TD control group and this was supported by the pattern of our results: 6 children with HF-ASD (43%) adopted an ineYcient switch strategy, compared to only 2 TD control children (13%). Group diVerences in task switching were investigated using Fisher's exact test, results indicate a non-signiWcant trend, p D .07.

[3] 295w Parent ratings on the BRIEF were collected for all participants in the HF-ASD sample. Scores for the eight BRIEF subscales, with a mean of 50 and a standard deviation of 10, are reported in Table 2. Scores above 65 are considered within the clinical range (Gioia et al., 2000). Our results support the BRIEF proWle identiWed in previous research (Gioia et al., 2002), of generally elevated subscale scores with a noticeably high score on the 'shift' subscale indicative of cognitive inXexibility. We also investigated hypothesized a priori relationships between speciWc multitask scores and BRIEF subscale scores, as there are theoretical parallels between some of these variables. Five Pearson correlations were conducted, with age in months and Full-scale IQ partialled out. Multitask plan and BRIEF plan/organise both measure planning skills but did not correlate: r D ¡.04, ns. Multitask perform, multitask switch and BRIEF initiate all measure the ability to independently self-initiate tasks; signiWcant negative relationships were identiWed between these variables: multitask perform and BRIEF initiate r D ¡.59, p D .04, multitask switch and BRIEF initiate r D ¡.64, p D .02. Multitask switch and BRIEF shift did not correlate signiWcantly r D ¡.24, ns, although both are thought to measure cognitive Xexibility. Finally, no relationship was observed for measures of the ability to track goal-oriented behaviour, multitask monitor and BRIEF monitor, r D ¡.26, ns. Despite having a priori theoretical grounds for predicting two of the relationships observed, we interpret these correlations cautiously. This is because when we correct for family-wise error and adopt a more stringent alpha level (Bonferroni method, alpha is corrected for the Wve relationships investigated, thus is set at .05/5 D .01), the two signiWcant correlations we reported between multitask and BRIEF variables fall below the level of statistical signiWcance.

[1] 160w We observed multitasking deWcits in children with HF-ASD compared to TD controls; indicating that children with HF-ASD were less eYcient at planning, organising and coordinating the performance of the multiple activities of the BMP. Although not all comparisons involving performance sub-scores reached signiWcance, trends in the data indicate that attempting all tasks, task switching and rule breaking appear to be problematic for children with HF-ASD. Our results correspond with those of one previous study in which children with HF-ASD exhibited impaired performance on a diVerent and sequentially rigid multitask paradigm (Emslie et al., 2003). Moreover, our results extend this previous research as we left the sequence of task performance to be Xexibly inferred and we investigated in depth what might underlie impaired multitask performance. Interestingly, multitasking deWcits could not be attributed to group diVerences in participant's knowledge of the task before them, as children in both groups demonstrated an equivalent ability to learn and retain the rules governing multitask performance.

[2] 24w Group diVerences in planning were observed: children with HF-ASD provided signiWcantly less strategic plans than TD controls. This Wnding is in accordance with the

[3] 40w Table 2 BRIEF scores for the HF-ASD group BRIEF subscale Mean (SD) Inhibit 71.78 (9.61) Shift 77.00 (8.90) Emotional control 70.21 (7.63) Initiate 63.14 (9.69) Working memory 63.93 (8.76) Plan/organise 64.78 (9.69) Organisation of materials 60.93 (8.13) Monitor 69.86 (6.90)

[4] 151w diYculties planning ahead that children with HF-ASD are reported anecdotally to have at home and at school (Ozo-noV et al., 2002). Although many studies have reported planning deWcits in ASD (e.g., Geurts et al., 2004;OzonoV et al., 1991), these deWcits tend to be on tasks with clearly deWned goals such as tower tasks. Our multitask paradigm taps the ability to plan multiple tasks in a complex environment in which the goals are largely self-determined, which is more characteristic of planning in everyday life. On a cautionary note, planning scores in this study were reliant on verbal output, and verbal Xuency can be reduced in ASD (Hughes et al., 1994). The possibility that the planning deWcits reported here may be due to group diVerences in the articulation of verbal plans, rather than impaired planning per se, cannot be discounted at this juncture in the absence of an independent measure of verbal Xuency.

[5] 260w Participants with HF-ASD switched between subtasks less eYciently than TD controls. In fact, three participants with HF-ASD failed to attempt all three subtasks, whereas all controls managed to fulWl this basic requirement. These three children engaged in the game as enthusiastically as other participants-but failed to attempt all three tasks. Two children (A and B) ran out of time before switching to the third task, and the third (C) became Wxated on one task and failed to switch at all. All three were disappointed when asked to stop by the examiner, in fact children B and C protested that they had not yet tried all the tasks. These three children were able to learn the rules of the paradigm as well as other ASD and control participants, and additionally to recall these rules post performance. Therefore, it is diYcult to attribute their poor performance cannot to a lack of comprehension of task instructions, or to poor motivation, hence performance diYculties are attributed to an impaired ability to organise the performance of multiple tasks. Providing further support for this interpretation, even those children with HF-ASD who did attempt all tasks switched between them less often and tended to get 'stuck' on task. In total, 50% of children with HF-ASD were inXexible switchers. We had expected task switching to be impaired in children with HF-ASD on the basis of their poor cognitive Xexibility (Geurts et al., 2004;OzonoV & Jensen, 1999) and from results of studies indicating their perseveration in switching set (OzonoV et al., 1991(OzonoV et al., , 2004;;Szatmari et al., 1990).

[6] 149w Although very few participants in either group made performance errors, signiWcantly more children with HF-ASD broke the rules than controls. In adult patients with frontal lobe damage, rule breaking in multitasking has been attributed to poor inhibitory control (Shallice & Burgess, 1991). Although not typically associated with HF-ASD, deWcits in inhibitory control have been reported in this population (e.g., Geurts et al., 2004;Nyden et al., 1999). It is therefore possible that this could account for the rule breaking observed. The strategic rule use score represents how eVectively an individual maximizes performance given the constraints of the situation. No signiWcant group diVerences in strategic rule use were observed. Although children with HF-ASD have been reported to have an impaired ability to use high-level, complex rules (Pennington & OzonoV, 1996), the rules of the BMP were designed to be very simple and the lack of group diVerences potentially reXects this aim.

[7] 474w The BRIEF scores observed in our clinical sample indicate that the HF-ASD children do indeed have signiWcant diYculties with executive control in their daily lives. The proWle of elevated scores across all eight subscales with the 'shift' subscale being the most elevated is entirely consistent with proWle generated for children with HF-ASD in two other studies (Gioia et al., 2000(Gioia et al., , 2002)). We investigated relationships between BRIEF sub scores and multitask variables hoping to provide further support for the ecological validity of our paradigm. Although two of the relationships we anticipated were observed, we interpret this Wnding cautiously as the correlation between variables failed to achieve signiWcance using more rigorous statistical criteria. The BRIEF initiate score correlated with both multitask performance and switching; 'initiate' measures the ability to begin tasks independently and to generate thoughts and ideas and this could relate to the requirement to independently begin and switch to many tasks during multitask performance. However, many other relationships we had anticipated were not observed; for example, multitask switch was expected to correlate with the BRIEF shift because task switching is thought to rely upon shifting attention from the current to the prospective activity (Kliegel et al., 2002). Likewise, multitask monitor was expected to correlate with BRIEF monitor as these measures both concern the ability to keep track of goal-directed behaviour. Finally, multitask plan and BRIEF plan/organise failed to show any relationship, despite the fact that planning deWcits were observed in the multitask paradigm and planning diYculties were apparent in the elevated BRIEF plan/organise subscale score. Other recent studies have failed to Wnd associations between performance on cognitive tasks and ratings of everyday behaviour (e.g., OzonoV et al., 2004). One potential limiting factor that also applies in the present study is the lack of variability in BRIEF scores, reXecting the fact that the majority of participants with HF-ASD scored clearly in the impaired range. Another possible explanation for this lack of observed diVerences is that we only obtained BRIEF ratings from parents, when in fact the questionnaire is designed to be completed independently by both parents and teachers. Parent reports may focus less than teachers on skills such as planning and monitoring, in part because parents are likely to do a lot of the day-to-day organising on behalf of their child to simply get things accomplished. These abilities might be more readily assessed in an environment where the child is expected to be more independent, such as school. Investigations using both parent and teacher questionnaires would be necessary to explore this possibility. A second explanation is that BRIEF subscales and multitask variables may measure diVerent aspects of the same ability; for example planning in the multitask paradigm involves planning in complex, less well-deWned situations, whereas the majority of questions in the BRIEF plan/organise subscale refer to planning ahead towards a single goal.

[8] 91w The inclusion of the BRIEF measure enables us to demonstrate that children with HF-ASD who have impaired multitask performance also have diYculties organising and coordinating activities in everyday life. For the most part these results add to the ecological validity of multitasking as a test of real life organisational diYculties. Further, they contribute to the growing practice in autism research of attempting to link performance on experimental paradigms to ratings of behaviour in everyday life (e.g., Dawson et al., 2002;Joseph & Tager-Flusberg, 2004;OzonoV et al., 2004;Pellicano, Maybery, Durkin, & Maley, 2006).

[9] 200w Although deWcits in planning how to perform the multitask paradigm were observed in children with HF-ASD, these same children did not demonstrate diYculty implementing the plans they had formed. Plan following scores, though yoked to planning scores, did not diVer signiWcantly between groups. Similarly, no signiWcant group diVerences on the multitask monitor score were found, indicating that children in both groups were able to monitor their ongoing performance. Both these results are interesting as they provide a possible platform for intervention. If children with HF-ASD can implement the plans they form and monitor the progress of this implementation, perhaps improving initial planning would facilitate subsequent performance. However, as discussed above, planning diVerences could be due to group diVerences in verbal Xuency. Likewise group comparisons on the monitor measure could be confounded by the measure being too simple, as the TD control group scored at ceiling. The ability to monitor ongoing performance may rely on metacognitive skills, and these may or may not be impaired in individuals with autism (Farrant, Boucher, & Blades, 1999). Future investigations of multitasking in HF-ASD would beneWt at least from independent measures of verbal Xuency and comprehension, and at best from minimizing verbal demands where possible.

[10] 146w In sum, the results of the present study highlight an overall deWcit in multitask performance in children with HF-ASD relative to age, gender and IQ matched controls, replicating and extending limited previous research. SpeciWcally, planning to perform multiple tasks, performing and switching between multiple tasks and inhibiting rule-breaking behaviour have been shown to be challenging for children with HF-ASD. These results are consistent with the cognitive proWle of executive dysfunction in autism, which is one of poor planning, cognitive inXexibility and possible inhibitory dysfunction. We interpret these results as experimental evidence that children with HF-ASD have problems in the prospective organisation of actions. We further supported this interpretation with evidence from a parent report measure which shows that children with HF-ASD who perform poorly on our multitask paradigm demonstrate diYculties organising future-oriented activities in their everyday lives, supporting the ecological validity of our novel multitasking paradigm.

[1] 270w Since autism is far more prevalent in males, 30 boys participated in this study, 16 typically developing (TD) boys and 14 boys with HF-ASD. TD participants were recruited from two mainstream schools. Participants with HF-ASD were recruited from two clinical centres specialising in the diagnosis of social communication disorders. All had a current clinical diagnosis of childhood autism (N D 9) or Asperger's syndrome (N D 5), according to ICD-10 (WHO, 1993) classi-Wcations. Nine of the participants with HF-ASD were taking part in an ongoing prevalence study of autism spectrum disorders (Chandler et al., 2005). In addition to having an ICD-10 (WHO, 1993) clinical diagnosis from experienced clinicians, these nine children also met research criterion for a consensus diagnosis of autism where participants were required to score above algorithm cut oV point on two of the following three measures: the Autism Diagnostic Interview (ADI-R; Lord, Rutter, & Le Couteur, 1994), the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000), and the Social Communication Questionnaire (SCQ; Berument, Rutter, Lord, Pickles, & Bailey, 1999). Although ADI-R and ADOS assessments were not completed for the remaining Wve children, these children were recruited from one of the leading diagnostic centres in the United Kingdom and diagnosis was made by an experienced clinical team. Additional inclusion criteria for HF-ASD participants included IQ within the normal range and no co-morbid diagnoses. Written parental consent and informed child assent were obtained for each participant prior to assessment. The groups were well matched for chronological age: TD mean D 11 years 11 months (SD D 6.36) and HF-ASD mean D 12 years 0 months (SD D 8.89).

[2] 75w The Battersea Multitask Paradigm (BMP) is a children's multitask test that we designed following the principles of the Greenwich Multitask Test for adults (Burgess et al., 2000). The BMP consists of three interleaved but very simple tasks, which children must perform in a time limit of 3 min, with performance constrained by four rules. The rules ensure that to achieve the most eVective performance, children must switch frequently between the tasks, rendering sequential performance ineVective.

[3] 223w The three tasks are bead sorting, counter sorting, and caterpillar colouring (Fig. 1). In each task, the child must sort or colour yellow and blue items (beads, counters, caterpillar circles) to Wll a 'cluster' (a pot of beads, a grid of counters or a whole caterpillar). In the beads task, children sort blue and yellow wooden beads (1.5-cm diameter, 150 of each colour) from a large box into eight smaller transparent containers embedded in a wooden tray: 2 small, 4 medium, and 2 large. Half the containers (some of each size) are marked for blue beads and half for yellow beads. In the caterpillar task, children use crayons to colour parts of blue or yellow caterpillars presented on an A3 sheet of paper. Each caterpillar is constructed of circles (2-cm diameter) with one coloured circle representing the 'head' and indicating which colour the 'body' circles should be. There are 12 caterpillars of varying length: 4 short, 6 medium, and 2 long, half to be coloured yellow (some of each length) and half blue. In the counters task, children sort Xat blue and yellow counters (1.5-cm diameter, 150 of each colour) from a large tub onto 10 grids of varying size: 4 small, 4 medium, and 2 large. Half the grids are marked for yellow counters (some of each size) and half for blue.

[4] 105w The tasks must be performed in a total time limit of 3 min, displayed visually to the child by a giant sand timer. Task performance is governed by four rules: (1) try all three games before the sand runs out, (2) yellow items get more points than blue, (3) full clusters get extra points, and (4) items must only be picked up or coloured one-by-one. The object of the game is to score maximum points without breaking any rules. The optimal way to perform the game is to Wll up small clusters of yellow items in all three tasks; this requires moving Xexibly between tasks.

[5] 52w To measure the cognitive processes underlying multitasking, Burgess and colleagues (2000) administered their Greenwich paradigm according to a six-stage invariant behavioural sequence. Each stage generated a dependent variable representing one of the cognitive processes involved, prospective memory, executive functions, and retrospective memory. We adopted this sequence of administration in the present study.

[6] 97w The child was introduced to the concept of the multitask game, the experimenter demonstrated each Fig. 1. Birds Eye View of Battersea Multitask Paradigm Test Apparatus. In all tasks, yellow and blue items are sorted to diVerent locations (all blue item locations are shaded). Caterpillars (left) of varying length to be coloured yellow or blue. Bead pots (centre) of varying size (small, medium, and large) to be Wlled with blue or yellow beads, stored in the box below. Counter grids (right) of varying size to be Wlled with blue or yellow counters, stored in the pot beside.

[7] 181w task and allowed the child to practice sorting and colouring items one-by-one and Wlling up clusters. Next the child was shown the giant sand timer and informed that they had only 3 min in which to try all three tasks. At this point the four rules of the paradigm were introduced (see above). It is important to note that tasks and task rules were introduced gently, using demonstrations as well as verbal instructions, to ensure that children were not confused by the simultaneous presentation of multiple verbal instructions. After the rules had been explained and demonstrated, the child was asked to free recall them. The free recall phase ended when the child had correctly recalled all four rules (2 points awarded for each rule free recalled, therefore rule learning criterion score D 8). Finally, to test cued rule recall, the child was asked nine questions about the rules, e.g., 'how many games must you try?' (see Appendix A for full list of questions). Free and cued rule learning are summed to form the composite score 'rule learn.' Possible range 8-18 points.

[8] 105w Next the child was asked to generate a plan of how they intended to perform the task, gaining as many points as possible and without breaking any rules. Plans were self-paced and recorded verbatim and scored according to the presence or absence of the following elements: plan to try all three tasks (0-3 points, 1 point per task), plan to prioritise yellow items (0-3 points, 1 point per task in which prioritising yellow was planned), and plan to Wll up clusters (1 point per task in which Wlling clusters was planned). The composite score 'plan' represents the sum of these scores. Possible range 0-12 points.

[9] 330w There was a time delay of approximately 5-7 min between the introduction of the rules and performance of the task, during this time children had consolidated their knowledge and understanding of the rules and formed a plan of how they would perform the task. Immediately after planning, the child was reminded of the rules and performed the multitask paradigm in the 3-min time limit. Four aspects of performance were scored: (a) 'Number of tasks attempted' measuring whether the child attempted to perform all three tasks (1-3 points, 1 point awarded per task attempted), sub-score range D 1-3 points. (b) 'Strategic performance' reXecting how well the child applied the rules of the paradigm to gain points; selecting a yellow item Wrst (0-3 points, 1 point for each task on which this was done), selecting signiWcantly more yellow items than blue items (0-3 points, 1 point per task on which this was achieved), selecting smaller, easy to Wll clusters Wrst (0-6 points: 2 points for a small cluster, 1 point for a medium cluster, 0 points for large cluster-on each task), prioritising Wlling a cluster of items (0-3 points, 1 point for each task in which a cluster was Wlled before moving on to another cluster or task). Sub-score range 1-18 points. (c) 'Penalty performance' representing breaking the 'one-byone' rule (¡3 to 0 points, subtract 1 point for each task in which this rule is broken) or placement errors made during performance (¡3 to 0 points, subtract 1 point for each task in which a coloured item is incorrectly placed, e.g., blue bead in yellow pot). Sub-score range D ¡6 to 0 points. (d) 'Task switch performance' reXecting how eVectively the child switched between the three tasks, categorized as eYcient (credited 2 points) or ineYcient (debited 2 points). EYcient switchers move Xuently between tasks, weighing up the best point-scoring options across as well as within tasks. IneYcient switchers fail to move Xuently between tasks, or make multiple switches in a haphazard fashion.

[10] 188w The composite score 'perform' is the sum of these four aspects of performance. Composite perform score range ¡3 to 20 points. 1 2.2.1.4. Plan follow. Assessed how well children followed their original plan whilst performing the paradigm. Failure to implement a planned intention was penalized. Plan follow is scored using similar elements to plan: number of tasks planned compared to number of tasks attempted (0-3 points, 1 point per task planned that was subsequently performed), number of tasks in which the child planned to prioritise yellow items compared to number of tasks in which yellow items were prioritised (0-3 points, 1 point per task in which prioritising yellow was planned and subsequently performed), number of tasks in which the child planned to Wll up clusters compared to number of tasks in which clusters were Wlled (0-3 points, 1 point per task in which clusters were planned and subsequently Wlled), and Wnally the order in which tasks were planned compared to the order in which they were performed (0-3 points, 1 point for each task where planned order matches performed order). Composite 'plan follow' score, possible range 0-12 points.

[11] 75w 2.2.1.5. Monitor. After performance the child was asked to tell and show the experimenter what he or she had done and why. Responses were recorded verbatim and scored for the following elements: recount of number of tasks attempted (0-3 points, 1 point per task attempted correctly recounted) and recount of order in which tasks were attempted (0-6 points, 2 points per task recounted in the same order as tasks were performed). Possible range 0-9 points.

[12] 158w Finally the child was asked to free recall the four rules of the paradigm (0-8 points, 2 points per rule correctly recalled) and to answer the same cued questions as when learning the rules (0-10 points, see Appendix A for list of questions). Free and cued recall 1 The lower bound of the possible performance score range is ¡3, due to worst case scenario. For example, if a child attempted only one task they would score 1 point. If, while performing this task they failed to prioritise yellow, or to Wll items etc., no strategic performance points would be awarded. Furthermore, this child could break the one-by-one rule and place coloured items incorrectly, thus being penalized ¡2 points. Finally, ineYcient switching would also be penalised by subtracting 2 points. Therefore summing +1, ¡2, and ¡2 gives a lower bound value of ¡3 points. It should be noted that no children performed the multitask in such an ineYcient way.

[13] 10w contribute to the composite score 'rule memory'-possible range 0-18 points.

[14] 206w These variables tap into the diVerent cognitive processes underlying multitasking. SpeciWcally, 'rule learn' and 'rule memory' measure the participant's knowledge of what they are supposed to do, hence retrospective memory. 'Plan' is dependent on executive planning skills. Multitask performance variables such as 'number of tasks attempted' and 'task switch' require both the self-initiated execution of delayed intentions (tapping prospective memory) and executive skills of inhibition and cognitive Xexibility. Similarly 'strategic performance' and 'penalty performance' rely upon executive strategy formation and inhibitory skills respectively. 'Plan follow' measures the successful implementation of delayed intentions and as such also represents prospective memory. 'Monitor' measures the child's awareness of what has been achieved and what has still to be achieved and is considered an executive monitoring process. In addition, all these aspects of multitasking are constrained by working memory as participants process multiple components of information in a limited capacity processing space. Given the anecdotal evidence that children with HF-ASD have diYculties organising activities in everyday life and have a proWle of executive dysfunction, predictions about their multitasking can be made. SpeciWcally, children with HF-ASD are expected to have diYculties planning compared to controls, and also to demonstrate multitask performance impairments, attempting fewer tasks and switching between tasks less eVectively.

[1] 91w The WISC-III UK (Wechsler, 1992) was used to assess general intellectual ability. Twenty-one participants were tested using a short form of the WISC-III UK in which scores from two verbal sub tests (arithmetic and similarities) and two performance sub tests (block design and picture completion) were prorated to give Full Scale, Verbal and Performance IQ estimates (following Kaufman, Kaufman, Balgopal, & McLean, 1996). Nine children with HF-ASD had been assessed on the full WISC-III UK within the last 12 months, so rather than repeat this assessment their previous scores were used.

[2] 159w Parents of HF-ASD children also completed the BRIEF, an 86-item questionnaire designed to assess EF behaviours in everyday environments (Gioia, Isquith, Guy, & Kenworthy, 2000). The BRIEF yields eight subscales representing domains of executive functioning: inhibit, shift, emotional control, initiate, working memory, plan/organise, organisation of materials and monitor (for further information, see Gioia et al., 2000). Each subscale is reported as a T score, with a mean of 50 and a standard deviation (SD) of 10. T scores at or above 65 are considered as having potential clinical signiWcance. The BRIEF proWle of children with HF-ASD is one of elevated scores on all subscales, particu-larly the shift subscale which measures cognitive Xexibility (Gioia, Isquith, Kenworthy, & Barton, 2002). We used the BRIEF: (a) to investigate whether any multitasking deWcits observed in children with HF-ASD were also reXected in their everyday environments and anticipated elevated scores and (b) to explore the relationship between speciWc multitask variables and some BRIEF subscales.