Introduction
In order to recommend appropriate guidance and treatment to adults with autism spectrum disorders (ASD), it is important to be aware of the specific impairments and coping mechanisms of these individuals. Knowledge about their strengths and impairments enables the search for occupations in which they can use their strengths and be restricted only minimally by their impairments. Local information processing has been frequently mentioned as a strength of individuals with ASD (Frith
1989,
2003; Happé and Frith
2006; Jolliffe and Baron-Cohen
1997; Shah and Frith
1993 and others). However, it is yet undetermined whether this also applies to adults with ASD.
Local versus global information processing in children with autism has been a topic of extensive research since 1989 (Frith
1989,
2003; Happé
1996; Jarrold et al.
2000; Morgan et al.
2003; Mottron et al.
2003; Ropar and Mitchell
2001 and others). Whereas global information processing has been characterized as processing information for meaning and gestalt, local information processing can be described as having a bias for featural and detailed information (Happé and Frith
2006). Individuals with autism appear to have a local perceutal bias, since they focus more on the elemental parts of a stimulus and have a strength in detail-focused information processing (Happé
1999).
The local information processing style in individuals with ASD is thought to be underlying areas of talent like memory for exact pitch (Bonnel et al.
2003) and superior visual search (Plaisted et al.
1998). However, the body of research that examined whether and to what extent adults with high functioning autism (HFA) or Asperger syndrome (AS) have a local information processing style is limited and the results of these studies are contradictory (Jolliffe and Baron-Cohen
1997; Kaland et al.
2007; Minshew et al.
2008; Pring et al.
1995; Rumsey and Hamburger
1988). Previous studies used both neuropsychological tests and self-reports to assess local information processing, although it has never been examined whether the two measure a similar underlying construct.
In the present study, local information processing by adults with HFA, AS and a neurotypical adult group will be investigated using both neuropsychological tests and self-report questionnaires. Furthermore, the relationship between the neuropsychological tests and the self-reports will be assessed.
Frith (
1989,
2003) was the first to examine local versus global information processing in individuals with autism. In her ‘weak central coherence account’, she described strengths in local information processing combined with a failure to integrate information into a meaningful whole as characteristic for autism. Throughout the years, the idea of a core deficit in central coherence has been replaced by the suggestion that local, fragmented information processing can be seen as a bias or cognitive style in individuals with autism spectrum disorders (ASD), which can be overcome in tasks that demand global processing (Happé and Frith
2006; Wang et al.
2007). Currently, two prevailing frameworks in local information processing in ASD are the ‘Enhanced Perceptual Functioning (EPF) hypothesis’ (Mottron et al.
2006), and the ‘Empathizing-Systemizing (E-S) account’ (Baron-Cohen et al.
2002). The EPF hypothesis states that people with autism display a local bias without evidence of a global deficit (Mottron et al.
2007). According to the E-S account, individuals with autism are more likely to use systemizing strategies. Systemizing can be described as the tendency to analyze information and to construct systems that are lawful. Although the E-S approach is not a local versus global theory of cognition theory per say, it does consider excellent attention to detail as a core characteristic of autism.
Studies that examined local information processing specifically in adults are limited and results are contradictory. Although there are no tests developed specifically to examine local information processing, the Embedded figures test (EFT: Witkin et al.
1962) and the Block design subtest of the WAIS III (Wechsler
1997) have been used the most frequently in this account. Research showed that performance on an adapted Block Design task is positively related to autistic traits (Stewart et al.
2009) and generally, superior performance on both tasks is interpreted as a strength in local information processing (Jolliffe and Baron-Cohen
1997; Shah and Frith
1993). However, to our knowledge, only a few studies examined EFT performance in adults with HFA or AS. In one study, superior functioning was found for adult groups with HFA and AS (Jolliffe and Baron-Cohen
1997), while another study in a similar group reported no strengths on this task (Minshew et al.
2008). For the Block Design task, superior performance by adult ASD groups was demonstrated in two studies (Rumsey and Hamburger
1988; Pring et al.
1995). Yet, Kaland et al. (
2007) reported no differences between adolescents with AS or HFA and a neurotypical group. In the present study, we use both the EFT and the Block Design task in relatively large adult groups with HFA or AS in order to more thoroughly examine local information processing in these groups.
A recent development in autism research is the use of self-reports to examine cognitive and behavioral features. In order to assess self-perceived local information processing and systemizing tendencies in adults with ASD, the Autism Spectrum Quotient (AQ: Baron-Cohen et al.
2001) and the Systemizing Quotient (SQ: Baron-Cohen et al.
2003) have been developed. Research demonstrated that adults with ASD obtained higher scores for both questionnaires compared to neurotypical adults (Baron-Cohen et al.
2001,
2003; Goldenfield et al.
2005; Hoekstra et al.
2008). Furthermore, AQ performance appears related to SQ performance in an autism spectrum condition group and, in a lesser degree, in a typical group (Wheelwright et al.
2006). Although the use of self-reports in individuals with autism is controversial, adolescents and adults with average verbal ability and a relatively high level of functioning seem able to describe their strengths and weaknesses adequately (Blackshaw et al.
2001; Frith and Happé
1999; Hobson et al.
2006; Spek et al.
2009). However, it has never been formally investigated whether self-report questionnaires and neuropsychological tasks that aim to measure local information processing actually measure similar underlying constructs. Therefore, the present study will examine the relationship between self-reports and neuropsychological tests that that used to measure local information processing.
When examining local information processing, it may be relevant to differentiate between HFA and AS, although it is questionable whether HFA and AS can be differentiated. The validity of AS as a distinct diagnostic entity, separate from other pervasive developmental disorders has not been established or disproved (Eisenmayer et al.
1998; Leekam et al.
2000; Wing
2005; Kamp-Becker et al.
2010). Furthermore, research shows that there are only few qualitative distinctions between HFA and AS; most features appear to be shared or overlapping to some degree (Ghaziuddin and Mountain-Kimchi
2004; Macintosh and Dissanayke
2004; Ozonoff and Griffith
2000). Still, the difference in degree of impairment and in language skills between HFA and AS (Kamp-Becker et al.
2010; Klin et al.
2005; Ozonoff et al.
2000; Spek et al.
2008) convinced us to study the two groups separately, especially since the self-report questionnaires rely on verbal skills.
Two factors that may be relevant to the use of the EFT and the Block Design task are speed of information processing and motor demands. Regarding information processing speed: Both tasks make use of a time limit: bonus points can be earned when less time is spent on resolving the items. The impairment in speed of information processing that has been found for children and adults with ASD (Calhoun and Mayes
2005; Mayes and Dickerson
2008; Spek et al.
2008; Yoran-Hegesh et al.
2009) may influence their performance of the EFT and the Block Design task negatively. Motor demands may also influence outcome on these two tasks (Wechsler
1997; Witkin et al.
1962). Therefore, in the present study we used processing speed as a covariate and chose two processing speed tasks that also incorporate motor demands.
Hypotheses of the Present Study
The present study aimed to examine local information processing in a relatively large group of adults with HFA and AS, using the EFT, the Block Design task, the AQ subscale ‘attention to detail’ and the SQ. We compared the performance of the HFA and AS groups with an IQ-matched control group of neurotypical adults. In line with the ‘enhanced local information processing’ theories in autism, we expected that the adult HFA and AS groups would perform better on the EFT and the Block Design task and would receive higher scores on the AQ and the SQ, compared to the neurotypical group. We investigated the relationships between the neuropsychological instruments (Block Design task and EFT) and the self-reports (AQ and SQ) in the research groups, in order to examine whether and to what extent these instruments measure similar phenomena. Furthermore, since we expect the speed of processing information to influence performance on the EFT and the Block Design task, specifically in the HFA group, we used the processing speed factor scale of the WAIS III as a covariate.
Results
Differences in EFT Response-Time and Block Design Performance
The mean scores and standard deviations of local information-processing as measured by the EFT and the Block Design task for the HFA group, the AS group and the neurotypical group are presented in Table
2.
Table 2
Means and standard deviations for the neuropsychological tests and the questionnaires
AQ subscale | 25.52 (6.06) | 25.44 (5.79) | 21.07 (4.79) | .000 | AS, HFA > NT |
SQ | 36.00 (11.52) | 34.24 (11.25) | 25.32 (9.56) | .000 | AS, HFA > NT |
Block design | 12.12 (3.63) | 12.56 (3.67) | 12.93 (2.25) | .528 | |
EFT | 38.71 (21.33) | 35.65 (22.17) | 25.99 (14.08) | .010 | AS > NT |
Processing speed | 100.19 (19.11) | 109.44 (17.10) | 112.24 (15.62) | .005 | AS, NT > HFA |
To test the hypothesis of differences in performance on the EFT and the Block Design task between the three groups, two-one-way between-group analyses of variance (ANOVA) were performed, using the diagnosis as the independent variable and the two neuropsychological tests as the dependent variables, respectively. The assumption of homogeneity was met, however, Levene’s test (Levene
1960) indicated that the assumption of equality of variance was violated in the analysis. Therefore a more conservative alpha of .025 was set (Tabachnick and Fidell
2007).
For mean response time in the EFT, the results displayed a statistically significant main effect of diagnosis (
F(2,121) = 4.76,
p = .01, partial eta squared = .07) with a moderate effect size (Cohen
1988, states that a partial eta squared of more than .06 can be described as a moderate effect size). For the Block Design task, no statistically significant main effect of diagnosis was found (
F(2,121) = .642,
p = .53). Post-hoc Tukey comparisons revealed that the neurotypical group was significantly faster in the EFT than the HFA group (
p = .01). The AS group did not differ in response time from either the neurotypical group or the HFA group.
To test the hypothesis of differences in self-perceived local information processing and the tendency to systemize, two-one-way between-group analyses of variance (ANOVA) were performed with the diagnosis as the independent variable or factor and the AQ and the SQ scores as the dependent variables, respectively. The assumptions of homogeneity and equality of variance were met. Wilks’ Lambda was used to measure group differences. For the AQ subscale, the results displayed a statistically significant main effect of diagnosis (
F(2,121) = 8.578,
p < .01, partial eta squared = .12). The effect size can be interpreted as moderate (Cohen
1988). For the SQ, a large and statistically significant main effect of diagnosis was found (
F(2,121) = 11.57,
p < .01, partial eta squared = .16). Post-hoc Tukey comparisons revealed that the neurotypical group scored significantly lower on the AQ subscale then the individuals with HFA (
p < .01) and the AS group (
p < .01). Furthermore, the neurotypical group obtained lower scores on the SQ compared to the HFA (
p < .01) and the AS group (
p < .01). There were no significant differences between the two disorder groups in the AQ and the SQ. The findings thus support the hypothesis that adults with HFA or AS report higher levels of local information processing and systemizing tendencies compared to the neurotypical adult group (Baron-Cohen et al.
2001,
2003; Goldenfield et al.
2005; Hoekstra et al.
2008).
The Relationship Between the SQ, the AQ Subscale, the EFT the Block Design Task
To investigate whether the self-assessments on the two self-report questionnaires and the performance on the two neuropsychological tasks are related, Pearson product-moment correlation coefficients were calculated. Table
3 presents the results.
Table 3
Correlation coefficients
1. AQ subscale | – | | | | |
2. SQ total score | .58** | – | | | |
3. Block design task | .10 | .19* | – | | |
4. EFT | −.01 | −.07 | −.63** | – | |
Strong and significant correlations were found between the SQ and the AQ subscale (r = .58, p < .01) and between the EFT and the Block Design task (r = −.63, p < .01). The correlation between the SQ and the Block Design task was significant but small (r = .19, p = .03). Other correlations were not significant. To investigate possible group differences, the correlation analysis of the AQ and the SQ was also done within the three groups separately. Strong and significant correlation between the AQ subscale and the SQ existed in all three groups (Autism group r = .57, p < .01; Asperger group r = .41, p < .01; and Neurotypical group r = .58, p < .01). This shows that the high correlations hold out in each group separately.
The finding of a strong association between the two neuropsychological tasks and between the two self-report assessments on the one hand and the lack of association between the neuropsychological tasks and self-report local information processing on the other, raises the question whether the two instruments assess a similar underlying construct.
This issue of construct validity was further explored by performing a factor analysis with the two neuropsychological tasks and the two self-report questionnaires as the variables. If all four measures point towards the same underlying construct, this points to the emergence of one factor (Gregory
2007).
Analysis yielded a KMO value above .5, and Barlett’s Test of Sphericity was significant at <.01, suggesting satisfactory conditions for factor analysis to proceed (Field
2005). In the analysis (method: Principal Components) two components emerged with eigenvalues exceeding 1, explaining 48 and 36% of the variance, respectively. The Oblimin rotated structure matrix of the two principal components is presented in Table
4.
Table 4
Principal component analysis: factor loadings (rotated component matrix)
Embedded figures test | −.907 | |
Block design task | .894 | |
SQ total score | | .892 |
AQ subscale | | .883 |
As Table
4 shows, the EFT and the Block Design task loaded predominantly on component 1, while the AQ and the SQ assessments loaded predominantly on component 2, with both components being only loosely associated (
b
between factors = .11).
The findings of the analysis indicate that the neuropsychological tasks and the self-reports do not point towards a similar underlying construct, but refer to two different constructs.
Exploration of the Predictive Validity of the SQ, the AQ Subscale, the EFT and the Block Design Task
To examine the ability of the neuropsychological test and self-report questionnaires to predict whether a person belonged to the neurotypical or to one of the diagnostic groups, a discriminant analysis was performed. The Asperger group and the HFA group were merged into one group and a two-group discriminant analysis was performed with the neurotypical group and the merged AS/HFA group as the dependent variable. This analysis yielded a statistically significant function (χ2(4) = 32.18, p < .01). Overall the discriminant function successfully predicted outcome for 77% of the cases, with accurate predictions being made for 77% of the HFA/Asperger group and 78% of the neurotypical group. The correlations between the predictor variables and the discriminant function showed that the SQ score (r = .72) and the AQ score (r = .63) are highly relevant in order to determine whether an individual belonged to either the HFA/Asperger group or the neurotypical group, while the EFT (r = .36) and the Block Design task (r = −.18) are less relevant in this respect.
The Influence of Processing Speed on Embedded Figures Test Performance
A one-way between-groups analysis of covariance was conducted to investigate whether the differences in Embedded Figures Test performance between the three groups can be attributed to processing speed differences. After adjusting for the processing speed scores, there was no significant difference between the neurotypical and the HFA group in the Embedded Figures Test (F(2,120) = 2.84, p = .06). This suggests that processing speed, as was expected, is an underlying factor of EFT performance in adults with HFA.
Discussion
The present study aimed to investigate local information processing in adults with HFA or AS and the usefulness of neuropsychological instruments and self-report questionnaires in this respect. We expected to find superior performance on the EFT and the Block Design task in the HFA and the AS group; however, the data of the present study did not support this hypothesis. The three groups did not differ in performance in the Block Design task and the neurotypical group even outperformed the two disorder groups on the EFT. Although the impairment in the EFT in the HFA group can be attributed to their relatively low processing speed group, this does not explain why the expected strengths were not found in the disorder groups.
Although these results are in contrast to previous studies of children and adults with ASD that used the EFT and the Block Design task (Jolliffe and Baron-Cohen
1997; Pring et al.
1995; Rumsey and Hamburger
1988; Shah and Frith
1993), two studies of adolescents and adults with ASD reported similar results (Kaland et al.
2007; Minshew et al.
2008).
As opposed to the results of the neuropsychological tests, the findings of the self-report questionnaires were in line with what we expected to find. The two disorder groups obtained higher scores for both the SQ and the AQ compared to the neurotypical group. Apparently, individuals with HFA or AS perceive themselves as being more detail-oriented and report the use of more systemizing strategies compared to the neurotypical group. These results replicate previous findings for adults with HFA or AS and are in line with the Enhanced Perceptual Functioning (EPF) hypothesis’ and the ‘Empathizing-Systemizing (E-S) account’ (Baron-Cohen et al.
2001,
2003; Hoekstra et al.
2008; Mottron et al.
2006; Wakabayashi et al.
2007).
The contrast between the results of the self-reports and the findings of the neuropsychological tasks is striking. Moreover, the analyses pointed to different underlying constructs. Previous studies reported similar results in other cognitive areas (Veenman
2005). If local information processing is an unitary concept, the results evokes the following explanations: either the neuropsychological tasks or the self reports are valid indicators of local information processing. If, according to the first possibility, the results of the neuropsychological tasks are a valid representation of local information processing, then adults with ASD would not differ from neurotypical adults in this respect. This would indicate that they have ‘overgrown’ their local information processing bias. It would also suggest that the relatively high level of self-reported local information processing that was found for the disorder groups is not valid. We can think of two possible explanations for this: first, the disorder groups may have adjusted their answers to what, in their opinion, corresponded to their diagnosis. However, this explanation seems unlikely because most of the participants were unaware of their diagnosis until after the neuropsychological testing process took place. Second, it could be argued that a lack of insight influenced the results of the self-report questionnaires for the individuals with ASD. However, this would imply that healthy adults are also unable to determine their level of local information processing, since in this group correlations between the neuropsychological tasks and the self-reports were also low or absent. Although it is theoretically possible, it does not seem likely that neurotypical adults with average intellectual capacities have so little insight into their cognitive functions.
According to the second possibility, the self-reports are a valid indicator of local information processing, which implies that the EFT and the Block Design task measure different cognitive features. In favor of this hypothesis is the fact that the performance on two self-report questionnaires appeared to be highly indicative of whether a person belonged to one of the disorder groups or to the neurotypical group, while the neuropsychological tests were less specific in this respect. Furthermore, it is important to note that the EFT and the Block Design task were not developed to measure local information processing. Research indicated that performance in the two tasks can be affected by multiple cognitive features (Happé and Frith
2006; Lezak et al.
2004; Witkin et al.
1962,
1971). For example, right and left hemisphere problems can influence performance on the Block Design task (Lezak et al.
2004). From this perspective, it is possible that the performance by our research groups in the EFT and the Block design subtest was influenced by other cognitive features than local information processing. Following this line of thought, the present data add to a recent discussion about whether cognitive task performance corresponds to performance in the real world, which has been referred to as ecological validity (Chaytor et al.
2006). It appears that a large amount of variation in everyday cognitive and behavioral skills cannot be accounted for in neuropsychological tests. In addition, factors such as compensation strategies, age and environmental characteristics influence test performance and can have a negative impact on ecological validity (Chaytor et al.
2006; Kenworthy et al.
2008). Research in autism showed that various neuropsychological tasks, aiming to examine executive functioning, have questionable ecological validity: emphasis has been put on improving and developing ecological valid tasks in this area (Kenworthy et al.
2008). Based on the results of the present study, parallels can be drawn for local information processing.
Although it seems most plausible that the self-reports provide the most valid representation of local information processing, our proof is only indirect. For instance, self-report questionnaires are verbal tasks in which comprehension of the various questions is essential. Therefore we need to be careful with conclusions in this respect. It is clear, however, that adults with HFA or AS report to be more detail-prone and more inclined to use systemizing strategies. It is important to take this into account when searching for an optimal educational and work environment where these individuals can use their strengths and abilities.
Although more research on this subject is needed, the results of the present study raise questions about the ability of the EFT and the Block Design task to measure local information processing in adults. There are alternative neuropsychological tasks which can be used to assess local information processing, as for instance a modified Block Design task (Shah and Frith
1993). However, research in high-functioning adults with ASD is limited and there is no information about the ecological validity.
If our results are replicated in future studies in adults, self-reports might be considered first choice for examining local information processing in adults, at least until valid neuropsychological instruments are developed specifically to measure this feature.
With regard to the self-reports, the present study showed that the correlation between the SQ and the AQ subscale is medium to strong in all three groups, which is in line with previous results of Wheelwright et al. (
2006). The two questionnaires share a considerable proportion of the variance. Local information processing is apparently related to the use of systemizing strategies. This is in correspondence with the E-S approach, which states that for systemizing, local processing is inevitable because a high systemizing mechanism needs to record each data-point (Baron-Cohen
2006). People with autism appear to use these lawful systems to keep an overview of all the details they are perceiving. This hypothesis supports recent ideas that individuals with autism are able to process information globally when necessary or when instructed to do so (Plaisted et al.
1999). It is interesting that the SQ and AQ subscale are also closely related in the neurotypical group. Systemizing strategies may also be used by healthy individuals as a way of organizing details and predicting change. This indicates that local information processing can be seen as a cognitive style and not as a defect, which is not only present in ASD but also in the general population. The idea of local information processing as a style rather than a deficit lends itself to a continuum approach, which is in line with recent perspectives on autism (Rapin
2005). In this view, individuals with ASD can be placed at the extreme end of the continuum, whereas people with impaired local information processing are placed at the opposite end of the same continuum.
In the present study, we differentiated the individuals with HFA group from those with AS, since research has shown that the degree of impairment in various areas is different in the two groups (Klin et al.
2005). Contrary to our expectations, no differences in the neuropsychological test results or in the self-report measures were found between the HFA and the AS group. It may be possible that, because of the relatively high level of functioning, differences in impairment between individuals with HFA and AS diminish during their lifetime. The results of the present study confirm previous studies which stressed the questionable validity of identifying autism and AS as separate disorders (Volkmar and Klin
2005).
Limitations
In the present study, all participants had at least average verbal ability. It is possible that the verbal ability of the participants have influenced performance on the tasks and questionnaires. Therefore, the results of the present study cannot be generalized to individuals with ASD who are less verbally capable. Furthermore, the relatively late diagnosis of a proportion of the participants characterizes our research group. A relatively late diagnosis has been hypothesized to be related to milder symptoms (Vermeulen
2002). However, all the individuals in the disorder groups matched criteria for HFA or AS and individuals with relatively mild symptoms were not included in the present study because they were, generally, diagnosed with PDD-NOS.
The present study used two self-report questionnaires to assess local information processing and systemizing tendencies. An adequate understanding and interpretation of the questions used in the questionnaires relies on semantic capacities. Although the two disorder groups were carefully selected and all participants had at least average verbal abilities, deficiencies in semantic processing which characterize individuals with ASD may have influenced performance in the two questionnaires.