Handedness and midsagittal corpus callosum morphology: a meta-analytic evaluation

Following a series of seminal studies in the 1980s, left or mixed hand preference is widely thought to be associated with a larger corpus callosum than right handedness, influencing the interpretation of findings and various theories related to interhemispheric processing, brain lateralisation, and hand preference. Recent reviews, however, find inconsistencies in the literature and cast doubt on the existence of such an association. The present study was conducted to clarify the relationship between hand preference and callosal morphology in a series of meta-analyses. For this purpose, articles were identified via a search in PubMed and Web Of Science databases. Studies reporting findings relating to handedness (assessed as hand preference) and corpus-callosum morphology in healthy participants were considered eligible. On the basis of a total of k = 24 identified studies and databases, random-effects meta-analyses were conducted considering four different group comparisons: (a) dominantly right- (dRH) and left-hand preference (dLH), (b) consistent right (cRH) and non-cRH preference, (c) cRH with mixed-hand preference (MH), and (d) cRH with consistent left-hand hand preference (cLH). For none of these meta-analyses did we find a significant effect of hand preference, and narrow confidence intervals suggest that the existence of population effects larger than 1% explained variance could be excluded. For example, considering the comparison of dRH and dLH (k = 14 studies; 1910 dRH and 646 dLH participants) the mean effect size was Hedge’s g = 0.016 (95% confidence interval: − 0.12 to 0.15; explained variance: < 0.001%). Thus, the common practice of assuming an increase in callosal connectivity based on mixed or left hand preference is likely invalid. Supplementary Information The online version contains supplementary material available at 10.1007/s00429-021-02431-4.

1 Supplement information to manuscript "Handedness and midsagittal corpus callosum morphology: A systematic meta-analytic evaluation" by René Westerhausen & Marietta Papadatou-Pastou 1. Table S1 to S4. Studies included in meta-analyses with details regarding sample characteristics and data extraction. Notes. n.a. = not available, a) classification criterion for all studies was a LQ larger or below zero for the dRH and the dLH group, respectively; or preference in selfidentification (study #9); b) in total sample; c) as original sample slightly differs in the number of subjects from the data available  Witelson, 1989) Notes. n.a. = not available, a) handedness group classification criterion: "qualitative" indicates that it was based on the answer per individual item not on the LQ score respectively; b) where available %female is given by group (cRH/NcRH); c) for whole sample  2. Table S5. Studies excluded from meta-analysis and reason for exclusion Table S5 Studies not included in the meta-analysis but which were considered eligible # Study Approach and narrative findings Reason for exclusion 1 Bruke & Yeo (1994) In men, posterior corpus callosum was correlated positively with increasing right-handedness (LQ score). In women, greater right-handedness was associated with smaller callosal areas in anterior and posterior corpus callosum.
Does not report group means, so that the study could not be included in the statistical analysis.
2 Cherbuin et al. (2013) Analysis the relationship of brain asymmetries and corpus callosum thickness including a group of non-right handers. Handedness effects on callosal measures are, however, not explicitly reported.
Sample overlap with Luders et al. (2010); not included in the metaanalysis (see below) 3 Clarke et al. (1993) Focusses on the relationship of corpus callosum morphology and dichotic listening performance. Differences between left-and right-handers in total corpus callosum area are reported as nonsignificant.
More complete report of results in Clarke and Zaidel (1994), so that data reported in Clarke et al. (1993) was considered redundant 4 Cowell et al. (1993) Provides a more detailed analysis of the relationship of consistency and direction of hand preference and the corpus callosum, as well as an analysis of "throwing hand" as important factor.
Same sample as reported in other included studies (Denenberg et al., 1991;Kertesz et al., 1987); the data of Cowell et al. (1993) was thus redundant to the previous publications 5 Gurd et al. (2013) Comparison between right-and lefthanded twins non-significant except for the region W22-39 (approx. genu) with a larger average thickness in the LH twins.
The results of the extended sample reported in Cowell and Gurd (2018) was included in the meta-analyses; Gurd et al. (2013) was not included.
6 Habib et al., (1991) Report larger area of the total corpus callosum, and various subregions in NcRH than in cRH after correction for brain size.
The authors do not provide callosal data that is uncorrected for brain size. The data was not included n metaanalysis.
7 Hopper et al. (1994) Report larger area of the anterior body of the corpus callosum in cRH compared to NcRH after correction for brain size. Comparison without correction was not significant.
The authors do not provide any measure of dispersion or test statistics so that the data cannot be included in the meta-analysis.
8 Josse et al. (2008) In the supplement to the publication the authors report a non-significant association of the EHI test score with total corpus callosum area in multiple regressions analysis including sex, age, and hemispheric white matter volume. The association was significant in a subsection covering genu and anterior body of the corpus callosum, but in no other subsection.
The data was not included as the mean raw data was not available for handedness groups.
9 Luders et al. (2010) The authors report negative correlations between callosal thickness and absolute Annett handedness scores in anterior and posterior midbody. Group comparisons between cRH and other handedness groups not significant. All analyses uses brain size as covariate.
Not included as uncorrected mean data for groups could not be obtained. 10 Mitchell et al. (2003) Reports normative corpus callosum data. No significant correlations were noted between handedness and callosal area or the area:cerebral volume ratio.
Not included in meta-analysis as mean data not reported.
11 O'Kusky et al (1988) Analyses focusses on differences between lateralisation groups, not handedness. Association of corpus callosum and handedness measures only reported for patients with epilepsy not for the control sample.
Not included in meta-analysis as data of healthy sample to presented. .
12 Preuss et al. (2002) No difference in total or subsection area of the corpus callosum between consistent and non-consistent right handers.
All participants are nominal righthanders (dRH); classification is based on hand-skill/performance data, and the study was consequently not included in the meta-analysis.
13 Reinarz et al. (1988) No differences between handedness groups (dRH vs dLH) in callosal area relative to brain size or in subsectional area relative to total corpus callosum size.
Only the subsection area relative to total corpus callosum area available in the article. Thus, the data cannot be included in the meta-analysis.
Data presented in graph and cannot be extracted.
15 Steinmetz et al. (1995) No difference between handedness groups in total corpus callosum area before or after adjustment for brain volume.
Handedness was defined based on hand performance, and is thus not included in the meta-analysis. However, a reanalysis of the data based on hand preference is reported by Jäncke et al (1997), which is included in the metaanalysis.
16 Welcome et al. (2009) Report no differences between consistent handers (combined cRH/cLH) and participants of inconsistent preference after correcting for brain size.
Data provided was not sufficient for inclusion in the meta analysis. However, the data was presented after re-analysis by McDowell et al. (2016), and raw data of this study was included in the present meta-analysis. 17 Westerhausen et al. (2006) Analysis based on language lateralisaton with handedness as covariate.
Sample is an extension of the Westerhausen et al. (2004) report and raw data of this study was included in the present meta-analysis.

Extraction of data available only through author contact
This section reports the data that became available after contacting the authors and explains how the mean and standard deviations included in the meta-analyses of the total corpus callosum size were determined. The means and standard deviations for the subregions can be found in the data tables provided on the accompanying OSF platform (https://osf.io/sw6ev/).
a. Westerhausen et al. (2004Westerhausen et al. ( , 2006 sample The original publication only published mean corpus callosum area data of a subsample: (a) Westerhausen et al. (2004) included 67 consistent left-and right-handed individuals for which both DTI and morphological corpus callosum measures were available; (b) Westerhausen et al. (2006) included 22 additional participants, which did not show consistent hand preference. The present analysis extends these previous two reports by also including participants for which no DTI data was available, so that the total N = 147.
Hand preference was assessed using a modified German version of the Edinburgh Handedness Inventory (EHI) including 9 items answered on a 5-point scale. Answers were scored from −2 for "always left" to +2 for "always right," yielding a total score ranging from -18 to 18.
For meta-analysis 1, a total score of 0 was used as cut-off to split the sample into dominant left-(dLH) and right-handed (dRH) individuals. Considering meta analyses 2 to 4, a total score of -14.4 or smaller and 14.4 or larger (representing 80% of the maximum absolute score) was applied to define consistent left-(cLH) and right-handed (cRH) individuals, respectively. This cut-off follows the strategy reported by Habib et al. (1991), one of the few studies which reported significant group differences. Individuals with a total score between these values were considered mixed-handed (MH). The group of non-consistent right-handers (NcRH) was formed by joining MH and cLH participants.
The resulting number of subjects (N), mean callosal areas (cm 2 ), and standard deviation (sd) for the groups were: The correlation of the hand preference score with total callosal area was r = .166, p = .045. Using the absolute score the correlation was r = .075, p = .368. Hand preference for this sample was assessed using the EHI questionnaire and the ariable coding hand prereference ranged range from -3 (all activities done with the left hand) to 3 (all activities done with the right hand).
Regarding meta-analysis 1, a total score of 0 was used to form a dLH and a dRH group. For meta-analyses 2 to 4, a total score of -2.4 or smaller and 2.4 or larger (representing 80% of the maximum absolute test score) was applied to cLH and cRH individuals, respectively. A total score between these values led to a classification of the indivuals as mixed-handed (MH). MH and cLH participants were joined to form the NcRH group.
Applying the above criteria the following number of subjects (N), mean callosal areas (cm 2 ), and standard deviation (sd) were obtained: Hand preference score correlated with callosal area at r = .003, p = .975. The correlation with absolute r = .020, p = .846.
c. Häberling et al., (2011Häberling et al., ( , 2012 sample The first author provided the mean values for the two publications (Haberling et al., 2011(Haberling et al., , 2012 by personal communication on February 12, 2021. Handedness was assessed with the Annett inventory (range -100% to 100%). A cut-off of 0 points was used to form the dRH and dLH samples and an index of larger 80 and smaller -80% was used to define cRH and cLH groups, respectively. MH included all participants between these values.
The following number of subjects (N), mean callosal areas (cm 2 ), and standard deviation (sd) was obtained for the 2011 publication: The values for the 2012 publication were: For the 2011 study, hand preference score correlated with callosal area at r = -.055, p = .679. Handedness was assessed using the 5-item Bryden questionnaire, and the laterality score reached from -1 (consistent left) to 1 (consistent right hander). The original publication compared all consistent individuals (i.e. scoring 1 or -1) with all others individuals.
Re-analysing the data for the meta-analyses, a total score of 0 was used as a cut-off value separating the dLH and dRH groups. A total score of -0.80 or smaller and 0.80 or larger (i.e. 80% of the maximum test score) was used to define cLH and cRH handers, respectively. A score between these values led to a classification of individuals as mixed-handed (MH). MH and cLH groups taken together formed the NcRH group.
Of note, the dependent variable was total corpus callosum volume, obtained using Freesurfer segmentation so that mean corpus callosum is expressed as volume (cm 3 ): Hand preference score and callosal volume correlated at r = -.008, p = .915. The correlation with the absolute value of the hand preference score was r = -.018, p = .816.

e. Labache et al. (2020) sample
The sample includes N=287 datasets as published in Labache et al. (2020), whereby the corpus callosum data has previously not been analyzed regarding hand preference. Hand preference was assessed using the EHI and the total score ranged from LI = -100 (all activities done with the left hand) to 100 (all activities done with the right hand).
For meta-analysis 1 a total score of 0 was used as a cut-off value separating the dLH and dRH groups, whereby 8 participants with the total score of 0 were excluded. For meta-analyses 2 to 4, a total score of -80 or smaller and 80 or larger (i.e. 80% of the maximum absolute test score) was used to define cLH and c>RH individuals, respectively. A score between these values led to a classification of the individuals as mixed-handed (MH). MH and cLH groups taken together formed the NcRH group.
Following the above classification criteria the following number of subjects (N), mean corpus callosum volume (cm 3 ), and standard deviation (sd) were obtained: Correlating the LI with total corpus callosum volume yielded an r = -.022, p = .706. Considering absolute LI values, the correlation was r = .200, p = .001.

f. Human Connectome Project (Van Essen et al., 2013)
The total sample includes N = 1113 datasets, which are all participants published in the 1200 Subjects Data Release of the HCP consortium (Van Essen et al., 2013) for which MRI data is available for download in tabulated form. The corpus callosum data represent data-extraction using Freesurfer software (Glasser et al., 2013). Hand preference was assessed using the EHI, and the LQ ranged from -100 (all activities done with the left hand) to 100 (all activities done with the right hand).
Mean and standard deviation of corpus callosum size was determined as follows: an LQ of 0 was used as a cut-off value for separating the dLH and dRH groups (note here 4 participants were excluded as the LQ was 0). An LQ of -80 or smaller and 80 or larger (i.e., 80% of the maximum test score) was used to define cLH and cRH handers. A score between these values led to the classification of the individuals as mixed-handed (MH). The NcRH group was formed by combining MH and cLH groups.
The following number of subjects (N), mean corpus callosum volume (cm 3 ), and standard deviation (sd) were obtained: The correlation of the LQ with total corpus callosum volume was r = .034, p = .261; and for absolute LQ the correlation was r = -.011, p = .707.

Risk-of-bias analysis within studies
The nature of the topic of the present meta-analysis reduces the risk-of-bias assessment to the "Bias in selection of the reported results" criterion as all other potential criteria are conceptualized for evaluating randomized clinical intervention studies (see Higgins, Savović, Page, Elbers, & Sterne, 2021). All included studies and datasets were scanned for selective reporting, that is, whether potentially available corpus-callosum data was also reported in the results section (evaluator R.W.). This was evaluated separately concerning total and subsection data. The results are presented in the table below. Table S6. Risk-of-bias assessment within studies concerning the data of the total corpus callosum (cc) and the subsection measures # Study Total cc Subsection comment 1 Clarke and Zaidel (1994) Low risk of bias Low risk of bias All data reported 2 Cowell and Gurd (2018) Low risk of bias Low risk of bias All data reported 3 Denenberg et al. (1991) Low risk of bias Low risk of bias All data reported 4 Haberling et al.
Low risk of bias not analysed All data available after author contact 5 Haberling et al.
Low risk of bias not analysed All data available after author contact 6 HCP 2017 Low risk of bias Low risk of bias All data available 7 Hines et al. (1992) High risk of bias High risk of bias Data only reported for sign. 15 Morton and Rafto (2006) Low risk of bias not analysed All data reported 16 Nasrallah et al. (1986) High risk of bias not analysed Only data of male sample reported 17 Ozdikici (2020) Low risk of bias not analysed All data reported 18 Steinmetz et al. (1992) Low risk of bias Some concerns Data of anterior regions not reported for theoretical reasons 19 Tuncer et al. (2005) Low risk of bias Low risk of bias All data available 22 Witelson (1985) Low risk of bias Low risk of bias All data reported 23 Witelson (1989) Low risk of bias Low risk of bias All data reported 24 Witelson and Goldsmith (1991) Low risk of bias Some concerns Data only of isthmus reported for theoretical reasons 6. Supplement Figure S2: Funnel plots of main analyses Fig. S2. Funnel plots of the four meta-analyses relating hand preference to total corpus callosum size (for details see text). Note that the shaded areas reflect the level of significance as indicated in the legend of each plot 20 7. Supplement Figure S3: moderator analysis by sex Fig. S3. Forest plots of the moderator analysis by sex. The results for the total corpus callosum are presented in (a) and for the isthmus in (b). More details regarding the main analysis can be found in the main text. 21 8. Supplement Figure S4: overview meta-analysis by subsection for cRH vs. cLH and cRH vs. MH comparisons. Fig. S4. Presents the effect size (g) and standard error of the effect size (se(g)) for each study included in the comparison of dRH and dLH (top), cRH and cLH (middle), and cRH and MH (bottom), respectively. Negative values indicate the subsection to be larger in the non-right-handed group (dLH, cLH, MH), positive values indicate the dRH/cRH group to have a larger callosal subsection. A meta-analytic average was only calculated were a sufficient amount of studies was available (as determined by power analysis, see main text), and otherwise not determined (n.d.). The provided mean effect is determined within a random-effects model. The values in brackets are the 95% confidence interval. Color coding was based on the Cohen's effect-size heuristics (Cohen, 1992) as indicated in the figure legend. Note, for some studies data was not available (n.a.) for some of the subsection.

Supplementary analysis: meta-regression differential sex distribution
As sex is associated with differences in the absolute size of the corpus callosum (Bishop & Wahlsten, 1997;Smith, 2005), a difference in the distribution of males and females in the compared handedness groups might potentially confound the comparison of the effect sizes.
To test for this possibility, we conducted a meta-regression analysis for the comparisons dRH vs. dLH and cRH vs. NcRH using the difference in the proportion of females as covariate. That is, a variable deltaF was determined as the percentage of female in the right-handed group minus the percentage of females in the non-right handed group (see Tables S1 and S2 for the data basis). Thus, positive values indicated a larger proportion of females in the right-handed sample and negative values a larger proportion of females in the non-right handed sample. Given that males can be expected to have the larger absolute corpus callosum, a larger proportion of females should affect the mean corpus callosum size in the respective group negatively. Consequently, a negative association of deltaF with the effect size would be indicative of the suspected confounding effect. Note, the sex distribution by sample was not available for all studies included in the meta-analysis so that the number of the studies included in the meta-regression analysis is reduced compared with the respective meta-analysis. The analysis was conducted using the metareg function of the metafor R library (Viechtbauer, 2020).
The meta-regression coefficient was β = 0.218 with a t = 0.11, p = 0.91 for the dRH vs. dLH analysis and including k = 11 studies. The coefficient was β = -0.87 with t = -1.17, p = 0.28 for the cRH vs. NcRH effect sizes (k = 10 studies). In both cases the percentage explained variance by deltaF was below 0.01%.

Supplementary analysis: absolute LQ
Spearman correlation (rsp) for the association of |LQ| and total corpus callosum area or volume. Rank correlation were considered appropriate as the distribution of |LQ| is nonnormal. All correlations were calculated from the raw data available (see Supplement Section 3). Note: One other study, Habib et al. (1991), report the association |LQ| and total corpus callosum area being significant (N = 53, rsp = 0.297, p = 0.03) but used callosal measures corrected for brain size (see Table 2, p. 49, of this publication). Luders et al. (2010), conducting their analysis based on thickness segments, find negative correlations between |LQ| and callosal thickness in anterior and posterior midbody but do not report the magnitude of the correlation. This study used brain size as covariate.

Supplementary analysis: removing HCP data from subsection meta-analysis cRH vs NcRH
In the all four subsection analyses comparing cRH vs NcRH samples, the HCP dominated the total sample, so that we re-calculated the mean effect after removing the HCP data to exclude that the found effects are mainly driven by one sample. While the mean estimates were comparable to those of main analyses, the confidence intervals were naturally wider Anterior third: HCP sample had an original weight of 38.9%. Re-calculating the metaanalysis without the HCP data, yielded a mean effect size of g = -0.03 (CI95%: -0.29 to 0.22; t = -0.34, p = 0.75; τ² = 0.03).