Is macrocephaly a neural marker of a local bias in autism?

Highlights

• •Local processing bias common to individuals with Autism Spectrum Condition.
• •Recent research suggests a link between a local processing bias and macrocephaly.
• •Results of this study of children with ASC further support this link.
• •Abnormal neural connectivity possibly the causal factor.
• •Characteristics represent an endophenotype in a subgroup of individuals with ASC.

Abstract

Previous research has suggested that the local processing bias often reported in studies of Autism Spectrum Condition may only be typical of a subgroup of individuals with autism also presenting with macrocephaly. The current study examined a group of children with autism, with and without macrocephaly, on the Children's Embedded Figures Test (CEFT), a well-established measure of local processing bias. The results demonstrated that the children with autism and macrocephaly performed significantly better on the CEFT than children with autism without macrocephaly, indicative of a local bias. These results lend support to the proposal that both macrocephaly in autism and a local processing bias may arise from the same underlying neural processes and these characteristics represent an endophenotype in a subgroup of individuals with ASC worthy of further investigation.

1. Introduction

The bias or preference to focus on local rather than global details has long been proposed as a characteristic of Autism Spectrum Condition (ASC; Frith, 1989, Happé, 1999). This bias towards a local processing style, often referred to as Weak Central Coherence (Frith, 1989), is proposed to account for the superior performance demonstrated by individuals with autism, mostly on visuospatial tasks involving attention to detail such as the embedded figures test and the block design task (Shah and Frith, 1983, Shah and Frith, 1993). In Shah and Frith's (1983) study, children with autism performed significantly better than the matched control group and the matched moderate learning disabilities group on the embedded figures test, indicating a natural bias for processing information locally. Effort is required to ignore the global picture and to focus on the local detail in this task; an innate bias towards local processing in ASC would mean that less effort is required to ignore the Gestalt than in typically-developed populations in which global processing is predominant, thus resulting in the noted superior performance (Shah and Frith, 1993). On the other hand, individuals with ASC are able to process information for global meaning when test measures explicitly require it (e.g. Plaisted et al., 1999), supporting the idea of a local bias or preference rather than a global processing impairment (Happé and Frith, 2006).

Although processing for local and global detail per se is intact in ASC populations, a suggested mechanism behind this local processing bias is a difficulty switching from one mode of processing to another, specifically from local into global; once individuals with ASC are processing in a local style, it may be more costly for them to broaden their spread of visual attention or zoom out to a global processing style (Mann and Walker, 2003, Rinehart et al., 2001, Ronconi et al., 2012, Wang et al., 2007). This is not related to general attention deficits as the difficulty appears to be unidirectional. Previous studies have also indicated that executive dysfunction is unrelated to a bias for local processing (Booth et al., 2003, Booth and Happé, 2010) and again, such an impairment would certainly not predict a unidirectional difficulty.

While an ever increasing body of research supports the idea of a local processing style in ASC populations (Jolliffe and Baron-Cohen, 1997, Morgan et al., 2003, Pellicano et al., 2006, Ropar and Mitchell, 2001), many other studies equally report null results, instead finding a global processing bias similar to control groups (Brian and Bryson, 1996, Mottron et al., 1999, Ozonoff et al., 1994, Spek et al., 2011; see Happé and Frith, 2006 for a review). One possible reason for this discrepancy in findings between studies is due to differences in methodology and task design; another could be due to heterogeneity within the ASC population. For instance, Booth et al.’s (2003) investigation of a bias towards a local processing style in autistic populations, indicated that only 60% of their autistic participants actually showed a preference for local processing, suggesting that this bias is restricted to a subgroup of the ASC population. Recent research by White et al. (2009) has suggested a neural marker for this bias towards local processing in a subgroup of individuals with ASC. White et al. (2009) found that this subgroup of autistic children showed a greater cost when switching from local into global processing, supporting the mechanism proposed by previous studies (Mann and Walker, 2003, Rinehart et al., 2001, Ronconi et al., 2012). More specifically, this greater cost of switching from local into global processing was restricted to children with ASC and macrocephaly; the same processing cost was not noted in children without macrocephaly, whether or not they had ASC, or in typically-developing children with increased head size (White et al., 2009).

Macrocephaly is a term used to describe abnormally large head circumference, greater than the 97th percentile (z > +1.88 SD) in the typical population, which occurs in approximately 20–30% of individuals with ASC (Bailey et al., 1995, Davidovitch et al., 1996, Lainhart et al., 1997). However, Wass's (2011) review reports that around 70% of children with ASC have a head circumference of 1.5 standard deviations above the norm as a result of abnormally accelerated brain growth. There are believed to be two possibly complementary reasons for this increased head size: an overabundant proliferation of synapses between nerve cells and a lack of pruning during early childhood (Frith, 2004, Happé, 1999). This lack of pruning would result in the preservation of unnecessary connections and a reduction in the reinforcement of essential neural connections (Frith, 2004). Frith (2004) suggested that this lack of pruning affects top-down processing and results in a poor ability to control bottom-up systems, which could present behaviourally as a local information processing style. Couchesne and Pierce (2005) propose that it is the neurons responsible for cortico-cortical communication that are particularly affected by this over growth and lack of pruning; this results in impaired communication between brain regions but enhanced connectivity locally which they propose would explain this detailed-focused processing style. Supporting this idea, a growing body of imaging studies have shown reduced long range connectivity (Bird et al., 2006, Damarla et al., 2010, Just et al., 2007, Kleinhans et al., 2008) and increased local connectivity (Belmonte and Yurgelun-Todd, 2003) in individuals with autism compared to typically-developed populations (see Wass, 2011 for a review).

Furthermore, fMRI studies examining cerebral activation while performing the EFT in ASC and typically developing populations have found between group differences (Damarla et al., 2010, Manjaly et al., 2007, Ring et al., 1999). In the study by Ring and colleagues, the typically developing control group showed greater activation of the right dorsolateral prefrontal cortex and bilateral dorsal parietal regions, while the ASC group demonstrated greater activation of the right ventral occipitotemporal area. Ring et al. (1999) propose that this supports the notion that ASC and controls solve the EFT task differently; with controls employing working memory strategies and ASC group employing a local processing approach evidenced by greater activation of the primary and associated visual areas. Damarla et al. (2010) and Manjaly et al. (2007) also found that their ASC groups demonstrated greater activation of brain regions responsible for visual processing in comparison to controls, suggesting enhanced local processing in the ASC group, with the study by Damarla et al. also showing lower functional connectivity between frontal and visuospatial regions. However although these three studies found differences in brain activation during performance of the EFT, behavioural results did not find the ASC group to show enhanced performance on the EFT in comparison to controls.

Studies attempting to link the neurocognitive profile of autistic individuals with macrocephaly have not found head size to be associated with verbal or non-verbal intelligence, executive function or language functioning (Davidovitch et al., 1996, Deutsch and Joseph, 2003, Lainhart et al., 1997). However a study by Deutsch and Joseph (2003) relating cognitive characteristics to macrocephaly revealed that ASC participants with superior non-verbal compared to verbal IQ had significantly larger head circumferences than either ASC participants with relatively superior verbal IQ or ASC participants with equivalent scores on verbal and non-verbal IQ. Deutsch and Joseph suggest that this superior performance in non-verbal IQ may be related to the enhanced visuospatial perceptual skills reported in autism, potentially reconciling this finding with White et al.’s (2009) study.

The current study aimed to replicate the relationship between autism, macrocephaly and a local processing bias using a well-established task that is also an independent test of local processing bias to that used by White et al. (2009): the Children's Embedded Figures Test (CEFT). It was expected that children with ASC and macrocephaly would demonstrate superior performance on the CEFT compared to ASC children without macrocephaly. Furthermore, the relationship between head size and IQ was also investigated, following the study of Deutsch and Joseph (2003). It was expected that relatively high non-verbal compared to verbal IQ scores would relate to enlarged head circumference.

2. Methods

Ethical approval for the study was received from the Joint UCL/UCLH committee on the Ethics of Human Research and written informed consent was obtained from the parents of all participants before inclusion in the study. 39 children (33 male) with Autism Spectrum Condition (ASC), aged 6–12 years, took part in the study (see Table 1). All these children had been previously diagnosed by a qualified clinician and their diagnoses were confirmed through paper reports at the time of testing. In addition, the Communication Checklist (Frith, unpublished; see Abell et al., 1999 and Hill and Bird, 2006) was employed to assess autistic symptomatology in terms of verbal and nonverbal communication; this is based on observation and scores are agreed by two raters. Thirteen items are rated on a 3-point scale and high scores indicate more severe autistic symptomatology. Data was available for 24 children and scores ranged from 16 to 34, comparable to the adults with Asperger Syndrome scoring 14–29 in the study by Hill and Bird (2006).

Table 1.

Participant characteristics: means (and standard deviations).

	ASC macrocephaly (N = 8)	ASC non-macrocephaly (N = 31)
Gender (M:F)	7:1	26:5
Age (years)	9.74 (1.49)	9.25 (1.53)
Communication checklist	24.40 (7.16)	19.58 (2.41)
Head size z-score**	2.93 (.84)	0.47 (1.06)
Non-verbal IQ	99.25 (15.85)	96.35 (17.29)
Verbal IQ	82.00 (12.01)	87.23 (12.17)
CEFT*	18.38 (6.14)	12.13 (5.36)

^*p < .05.

^**p < .001.

The head circumference of each child was measured in centimetres with a flexible tape measure and, using the Farkas (1994) database, was converted to standardised z-scores adjusted for age and gender. All participants were Caucasian, in keeping with the population through which these norms were defined. Macrocephaly is defined as a head circumference greater than the 97th percentile, that is more than 1.88 Standard Deviations above the normative mean. The children were classified as either macrocephaly or non-macrocephaly and this determined the between groups allocation. 21% (8/39) of children in the current sample met this criteria for macrocephaly.

The macrocephaly and non-macrocephaly groups were matched in age (t(37) = 0.82), verbal IQ (t(37) = 1.09) and non-verbal IQ (t(37) = 0.43). Both groups displayed greater mean non-verbal IQs compared to verbal IQs and this difference was significant within both groups (Macro: t(7) = 4.19, p = .004; Non-Macro: t(30) = 2.97, p = .006). The groups differed significantly in head size z-scores (t(37) = 6.10, p < .001) but not on the Communication Checklist (U = 29.5, p = .19) (see Table 1).

Verbal and Non-verbal IQ scores were calculated using the British Picture Vocabulary Scale 2nd edition (BPVS–II; Dunn et al., 1997) and Ravens Standard Progressive Matrices (RSPM; Raven, 1958) respectively. The Children's Embedded Figures Test (CEFT; Witkin et al., 1971) was administered according to the manual as a measure of local processing bias. In this test the child is required to find a small shape hidden within a larger image. Accuracy in finding the hidden shapes out of a maximum of 25 was recorded in this task. The recommended 180 s time limit was used although most children either located the target or gave up looking well before this limit was reached.

3. Results

All variables were normally distributed and displayed homogeneity of variance, except for scores on the Communication Checklist. Firstly the relationship between head size and IQ was investigated across the whole sample. There was no significant relationship between head size and verbal IQ (r = .01) or non-verbal IQ (r = .22). Verbal IQ scores were then subtracted from Non-Verbal IQ scores (NVIQ–VIQ) to investigate if the magnitude of the discrepancy between these scores was related to head size; again this was non-significant (r = .22). There was similarly no significant difference in the IQ discrepancy scores (NVIQ-VIQ) when comparing the macrocephaly and non-macrocephaly groups (t(37) = 1.26, p = .215).

The relationship between head size and CEFT score reached borderline significance across the whole sample (r = .30, p = .068), with a trend for increasing head size to be associated with increasing scores on the CEFT. Furthermore, the two groups did differ significantly in their performance accuracy on the CEFT (t(37) = 2.85, p = .007). The ASC children with macrocephaly scored significantly higher on this measure than the ASC children without macrocephaly (see Fig. 1).

Fig. 1.

Relationship between head size z-scores and performance on the Children's Embedded Figures Test by children with ASC with and without macrocephaly.

CEFT score was not correlated with verbal IQ (r = .23) or IQ discrepancy score (r = .15), however there was a borderline correlation between CEFT score and non-verbal IQ (r = .30). The above difference in CEFT scores between groups remained when NVIQ was covaried (p = .005).

To further examine the relationship between increased head size and a local processing bias, a median split was performed on head size across the entire group (n = 39), with 19 children in Group 1 (head size z-score >1.16) and 20 children in Group 2 (head size z-score <1.16). By definition, Group 1 had significantly greater head size (t(37) = 7.92, p < .001) but the two groups did not differ in age (t(37) = .097), verbal IQ (t(37) = 1.42), non-verbal IQ (t(37) = 1.61), IQ discrepancy (t(37) = .580) or on the Communication Checklist (U = 60.5, p = .57). As expected, Group 1 did still score significantly higher on the CEFT compared to Group 2 (t(37) = 2.93, p = .006).

From Fig. 1 it can be seen that one child with macrocephaly did not demonstrate this superior performance on the CEFT. This was the one female child within the macrocephaly group. This child had a borderline verbal IQ of 76 and a non-verbal IQ in the normal range at 98. This borderline VIQ score is unlikely to explain her performance on the CEFT, as other children within the Macrocephaly group had similarly low VIQ scores (range 63–103) and as previously reported VIQ was not correlated with performance on CEFT (r = .23).

4. Discussion

The current study aimed to investigate the relationship between a bias towards local processing and macrocephaly in autism. The Children's Embedded Figures Test was utilised as a measure of local processing, as performance on this task requires the individual to focus on the local detail of the image and to ignore the gestalt, thus resulting in superior performance in individuals predisposed to this processing style. Children with ASC and macrocephaly were more able at the task than the remaining children with ASC; furthermore, head size and task performance were weakly correlated. These results replicate previous findings (White et al., 2009) in an independent sample of children and with an independent measure of local bias; they therefore lend stronger support to the hypothesis that macrocephaly in autism is associated with a bias towards local processing and that the same neural process may underlie both of these features. Macrocephaly may therefore be a neural marker for a local processing bias.

While previous research has demonstrated that individuals with ASC show enhanced performance on measures of local processing (Jolliffe and Baron-Cohen, 1997, Pellicano et al., 2006, Shah and Frith, 1983), other studies have not replicated this finding (Brian and Bryson, 1996, Mottron et al., 1999, Ozonoff et al., 1994, Spek et al., 2011). As White et al. (2009) suggested that this local processing style is only seen in a subgroup of individuals with ASC, those also presenting with macrocephaly, this could potentially explain the inconsistencies in the past literature. The current study adds further support to this idea that a local bias is not a universal feature of autism and is therefore unlikely to be causal to the defining features of the disorder.

Furthermore there is the question whether the enhanced performance in local processing noted in individuals with ASC is due to local processing bias or a deficit in global processing (Happé and Booth, 2008). Happé and Frith's (2006) review highlights the mounting research in support of the notion of a local bias rather than a global processing deficit, as global processing has found to be intact in individuals with ASC (e.g. Plaisted et al., 1999). The study by White et al. (2009), showed that there was no significant difference between children with ASC and macrocephaly in comparison to children with ASC without macrocephaly or typically developing children on tasks examining local and global processing separately, demonstrating both intact local and global processing in all groups. The difference between groups is only apparent in the cost of switching, whereby those with ASC and macrocephaly demonstrate a greater cost of switching from local into global processing compared to children with ASC without macrocephaly and typically developing children. The children with macrocephaly in the current study presumably displayed superior performance on the EFT as they experienced less interference from global level detail; once their visual attention was ‘zoomed in’ on the local detail they had difficulty in broadening their attention and switching to global processing thus experiencing less interference. However, as suggested by Shah and Frith (1993), in children without this local processing bias more effort is required to ignore the global image thus resulting in inferior performance.

Frith (2004) suggests that the increased head size seen in children with ASC may be due to overgrowth of synapses and a lack of neural pruning in childhood and that this affects top-down processing and results in a poor ability to control bottom-up systems resulting in a local processing bias. Studies by Damarla et al. (2010), Manjaly et al. (2007) and Ring et al. (1999) found greater activation in the visual cortex in ASC participants compared to controls while performing the EFT task. As the authors suggest this indicates enhanced local processing in participants with ASC. These authors did not find an overall advantage in the ASC participants on performance of the EFT, perhaps due to the heterogeneous nature of the ASC participants. It would be interesting to examine brain activation during performance of the EFT in ASC participants with and without macrocephaly, and to examine both behavioural and functional differences in performance.

Wass (2011) reported that 70% of children with ASC have a head size of 1.5 standard deviations (SD) above the norm due to abnormally accelerated brain growth. When our participants were spilt into two even groups, with a head size greater or less than 1.16 SDs, the strong difference in CEFT score between the groups remained. This suggests that this bias for local processing might not be restricted to those who meet the clinical criteria for macrocephaly but perhaps extends to a larger proportion of ASC children with above average head size. Indeed, the weak correlation found here between head size and CEFT indicates that this relationship may be continuous, with the degree of macrocephaly being related to the strength of the local bias in ASC. To a lesser extent therefore, excessive synaptogenesis and a lack of neural pruning may be present in a larger subgroup of individuals with ASC than the 20–30% currently identified as displaying macrocephaly. As suggested by Couchesne and Pierce (2005), this increased head size may result in reduced long range connectivity, and imaging studies do support this proposal of reduced connectivity in individuals with ASC (Bird et al., 2006, Damarla et al., 2010, Just et al., 2007, Kleinhans et al., 2008); however, the impact of head size on connectivity was not investigated in these studies. Again investigating the relationship between the extent of reduced connectivity, the degree of increased head size and performance on tasks requiring local processing would help to shed light on whether this is a continuous relationship and on the neural processes behind this processing style.

The only other previous study to find a relationship between macrocephaly and a behavioural or cognitive feature of autism used IQ discrepancy (Deutsch and Joseph, 2003). Specifically, children with relatively high non-verbal IQ in comparison to their verbal IQ were found to have larger head circumferences. Here, we similarly found no relationship between absolute verbal or non-verbal IQ and head size but also no association between IQ discrepancy and head size. In fact, individuals with ASC both with and without macrocephaly displayed greater non-verbal than verbal IQs. Our study and that of Deutsch and Joseph did use very different IQ measures however and it is possible that their non-verbal IQ measure tapped more accurately into local processing tendencies. It seems likely therefore that it is a local processing bias rather than relatively high non-verbal IQ per se that shares a common neural substrate. In addition, Shah and Frith (1993) found that individuals with ASC demonstrated superior performance on the block design task of the Weschler Scales compared to other non-verbal subtests, while typically developing children displayed a balanced performance across all non-verbal subtests. The block design task requires the participant to segment a larger image into parts and with these parts reconstruct the global image, and so is consider a task of local processing. The ASC group in this study outperformed a learning disability group similar in age and nonverbal IQ. This suggests that the ASC group had superior ability on this task irrespective of nonverbal IQ abilities (Shah and Frith, 1993). In the current study a borderline correlation between nonverbal IQ and CEFT score was found, again suggesting that local processing ability as measured by the CEFT is partially independent of non-verbal abilities.

In summary, in this study 21% of an unselected sample of children with ASC had a head circumference indicative of clinical macrocephaly, a finding in keeping with the proportions reported in the ASC literature (Lainhart et al., 1997, Davidovitch et al., 1996, Bailey et al., 1995). This increased head size has been proposed to result from a period of excessive neural growth in infancy combined with a lack of neural pruning, leading to poor modulation of top-down processes and enhancement of perceptual bottom-up processes (Frith, 2004, Happé, 1999). This is suggested in turn to give rise to a bias towards local processing. We found further evidence to support this hypothesis: children with ASC and macrocephaly displayed enhanced performance on a well-established task that taps into this local processing bias. Macrocephaly may be a neural marker of a local processing bias and may therefore be a useful endophenotype for future genetic studies of ASC.

Conflict of interest statement

None declared.