‘Sinister psychotics’ is the title of an article by Taylor & Amir Reference Taylor and Amir1 and it refers to the idea that there is a relatively higher percentage of non-right-handers in schizophrenia compared with healthy individuals. This excess of non-right-handedness is a cornerstone for the assumption that schizophrenia is characterised by - some argue even caused by Reference Crow2 - a lack of lateralisation. Indeed, a number of studies have found non-right-handers to be overrepresented in schizophrenia. Reference Collinson, Phillips, James, Quested and Crow3,Reference Cannon, Byrne, Cassidy, Larkin, Horgan and Sheppard4-Reference Gur6 However, there are also null findings Reference Wahl7,Reference Oddy and Lobstein8 and even reports of fewer non-right-handers in schizophrenia samples. Reference Taylor, Dalton and Fleminger9,Reference Oyebode and Davison10 A meta-analysis by Sommer et al Reference Sommer, Aleman, Ramsey, Bouma and Kahn11 revealed significant odds ratios (ORs) in the range of OR = 1.48-1.61, indicating that the odds of being non-right-handed were approximately 1.5 times higher in schizophrenia than in healthy individuals or other psychiatric control patients. This was confirmed in another meta-analysis by Dragovic & Hammond, Reference Dragovic and Hammond12 in which left- and mixed-handers (i.e. individuals with no consistent hand preference) were overrepresented in schizophrenia (OR = 1.5-1.8). These systematic reviews suggest that patients with schizophrenia are indeed more ‘sinister’ than controls, corroborating Crow’s theory. Reference Crow2,Reference Oertel-Knöchel and Linden13 However, critics have argued that the higher prevalence of non-right-handedness would disappear if gender and/or biases in self-report handedness questionnaires are controlled for.
A gender artefact?
Schizophrenia research samples typically consist of more male than female patients and in the general population there are more non-right-handed males (about 12%) than females (about 10%) (for review see Papadatou-Pastou et al Reference Papadatou-Pastou, Martin, Munafo and Jones14 ). Thus, unless control groups are matched for gender (or gender is taken into account statistically), most likely one will find a higher proportion of non-right-handers in the schizophrenia compared with the control sample. Indeed, Buijsrogge et al Reference Buijsrogge, Benali, Brazo and Dollfus15 reported that the excess of non-right-handedness in schizophrenia disappeared when gender was taken into account, but the sample was relatively small (73 patients and 81 controls). More recently, Deep-Soboslay et al Reference Deep-Soboslay, Hyde, Callicott, Lener, Verchinski and Apud16 replicated this finding with a total of 1445 participants (375 patients with schizophrenia) and concluded that ‘theories of schizophrenia based on anomalous lateralization related to handedness are unsupported by our data’ (p. 3120). On the other hand, Cannon et al Reference Cannon, Byrne, Cassidy, Larkin, Horgan and Sheppard4 found increased rates of non-right-handedness, although gender was statistically controlled for, and other studies with gender-matched control groups also reported an excess of non-right-handers. Reference Collinson, Phillips, James, Quested and Crow3,Reference Dvirskii5,Reference Gur6 However, there are also examples of studies with gender-matched samples which did not find an increased rate of non-right-handers. Reference Gureje17,Reference Kameyama, Niwa, Hiramatsu, Saitoh, Flor-Henry and Gruzelier18 To add further confusion, some studies report an excess of non-right-handedness in either only female Reference Hauser, Hudson, Hood and Seeman19,Reference Orr, Cannon, Gilvarry, Jones and Murray20 or male patients with schizophrenia. Reference Sperling, Martus and Barocka21,Reference Ortuno, Lopez, Landecho and Bonelli22 To summarise, the literature is highly inconsistent and it cannot be ruled out that the excess of non-right-handers in schizophrenia is a gender artefact. Moreover, it is unclear whether there is, in fact, a gender difference with respect to non-right-handedness. That is, whether both males and females with schizophrenia are more likely to be non-right-handed. The two meta-analyses Reference Sommer, Aleman, Ramsey, Bouma and Kahn11,Reference Dragovic and Hammond12 did not resolve this matter, because they did not control for gender as a mediating factor.
Biases in handedness questionnaires?
Dragovic & Hammond Reference Dragovic and Hammond12 did not find an excess of non-right-handedness when Annett’s hand preference questionnaire Reference Annett23 was used (OR = 1.0-1.2). Conversely, studies using the Edinburgh Handedness Inventory (EHI) Reference Oldfield24 revealed that the odds of being non-right-handed were about three times higher in patients with schizophrenia than in controls. Although both questionnaires consist of similar items, it has been demonstrated in healthy controls that the EHI produces significantly more ‘either hand’ responses than Annett’s questionnaire. Reference McMeekan and Lishman25,Reference Williams26 This discrepancy might be rooted in the instructions of the EHI, Reference Williams26 which encourage participants to indicate ‘either hand’ responses, whereas such a statement is missing in the instructions for Annett’s questionnaire. If patients with schizophrenia have a higher propensity to follow this instruction, this would lead to an excess of non-right-handedness, especially mixed-handedness. We could only find one study that assessed handedness with both Annett’s questionnaire and the EHI. Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27 Here, however, participants did not fill in the questionnaires. They either demonstrated the questionnaire items (i.e. performing the actions as described in the items) or they gave verbal responses, which makes comparison with self-report questionnaires difficult. Moreover, verbal responses on Annett’s questionnaire items were not consistently recorded and data were missing for a few participants. Byrne et al Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27 found that there were more mixed-handed controls when handedness was based on verbal reports than when it was based on demonstration. In other words, controls rated themselves less left- or right-handed than the behavioural assessment implied. No such effect was found in schizophrenia, but there were only 17 patients, as the main focus was on participants with a high risk of developing schizophrenia (n = 115). Interestingly, the high-risk participants also rated themselves more mixed-handed than the behavioural assessment implied.
The findings of Byrne et al Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27 do not resolve the discrepancy between the EHI and Annett’s questionnaire, Reference Dragovic and Hammond12 but they point to yet another problem: participants may not always report their handedness accurately. This applies especially to patients with schizophrenia, whose completion of handedness questionnaires might be affected by cognitive deficits and, perhaps, a higher propensity for mixing up ‘left’ with ‘right’. Reference Curcio28 There are, in fact, reports in the literature that support such a notion. For example, Wahl Reference Wahl7 found that the significant excess of non-right-handedness disappeared on exclusion of all participants who showed a mismatch between handedness based on self-report and based on observation. In another article, the authors Reference Taylor, Dalton and Fleminger9 noted: ‘Two females and eight males, all right-handed writers, had right-left confusion and consistently referred to their right hand as their left hand’ (p. 376). We do not go so far as arguing that the majority of patients cannot correctly self-report their handedness. The two examples, however, highlight that handedness determined with self-report questionnaires, particularly in patients with schizophrenia, might indeed lack validity.
In summary, there are indications that the higher prevalence of non-right-handedness in schizophrenia might be the result of biases and/or invalid responses in self-report handedness questionnaires. It would be interesting to further investigate why the excess of non-right-handedness in schizophrenia only emerges for the EHI and not Annett’s questionnaire. Reference Dragovic and Hammond12 However, given the findings of Byrne et al, Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27 behavioural assessment seems a far more reliable method to determine handedness. So far it has not been investigated in a systematic, quantitative way whether studies that assessed handedness behaviourally also show an excess of non-right-handedness in schizophrenia.
New meta-analyses
There are reasons to question the excess of non-right-handedness in schizophrenia. Reviews Reference Oertel-Knöchel and Linden13,Reference Satz and Green29 and meta-analyses Reference Sommer, Aleman, Ramsey, Bouma and Kahn11,Reference Dragovic and Hammond12 propagating this excess might be subject to a gender artefact and/or biases and validity problems related to self-reported handedness. This would have far-reaching consequences, since the excess of non-right-handedness is a key argument for the theory that schizophrenia is characterised or even caused by anomalous lateralisation. As a consequence, we carried out three separate meta-analyses. These analyses addressed whether the excess of non-right-handedness in schizophrenia persists when (a) gender is taken into account, (b) handedness is assessed behaviourally, and (c) both factors are taken into account simultaneously. In addition, the present study provides an update of previous meta-analyses Reference Sommer, Aleman, Ramsey, Bouma and Kahn11,Reference Dragovic and Hammond12 by including more recent studies.
Method
Literature search and study selection
The platforms PubMed, ISI Web of Knowledge and PsycINFO were searched with the key words “schizophrenia”, “handedness”, “sex” and “gender” between January and March 2013 (for the full electronic search strategy, see the online supplement to this paper). Moreover, we examined the studies included in Dragovic & Hammond Reference Dragovic and Hammond12 and Sommer et al Reference Sommer, Aleman, Ramsey, Bouma and Kahn11 and contacted authors of recent articles with substantial sample sizes to obtain additional information on the handedness of their female and male participants. In principle, we had no restrictions regarding study language but the meta-analysis contains only English articles, presumably as a result of conducting most of the literature search via English platforms. We applied the following inclusion criteria.
Diagnosis of schizophrenia
We did not restrict the meta-analysis to studies that used DSM or ICD criteria, because this would exclude older studies that investigated handedness in schizophrenia before DSM was introduced and because the core symptoms of schizophrenia (hallucinations, delusions, negative symptoms) have changed relatively little over the last decades. We did not include psychiatric disorders other than schizophrenia and, if possible, only included ‘pure’ schizophrenia cases and not schizophreniform or schizoaffective cases. However, if handedness was only available for patients with schizophrenia, schizophreniform and schizoaffective disorder combined, then the combined handedness data were included.
Number of patients and controls that were right- and non-right-handed
In cases where percentages were reported (e.g. 5% non-right-handers), these were transformed into integers. For the meta-analysis that specifically examined gender effects, only studies that reported the exact number (or percentages) of female and male participants were included.
Control group
Dragovic & Hammond Reference Dragovic and Hammond12 included studies in which the schizophrenia sample was matched with a control sample obtained from other studies to increase the number of available studies. We took a more restrictive approach (see also Sommer et al Reference Sommer, Aleman, Ramsey, Bouma and Kahn11 ), because we wanted to make sure not only that the same handedness test was used, but that it was also applied under similar circumstances by the same experimenters. If there were two control groups, one comprising healthy siblings and the other random healthy controls, the two groups were combined. Studies including patients with psychiatric disorders as controls were excluded.
Patient and control samples were not matched for handedness
Studies were excluded if control samples were described as ‘handedness-matched’ or ‘matched for handedness’, because this does not represent a natural proportion of right- and non-right-handers. We did not exclude studies that specifically stated they were gender-, age-, education- or ethnicity-matched but did not include handedness in the list of matching factors. We also included four studies that described their controls as ‘similar’ or ‘comparable’ to patients with schizophrenia in handedness, Reference Bonner-Jackson, Haut, Csernansky and Barch30-Reference Seidman, Lanca, Kremen, Faraone and Tsuang33 as participants were apparently not actively matched for handedness.
Patient and control samples did not solely consist of right-handers
Several studies, particularly neuroimaging studies, tested only right-handed participants. We excluded those studies because it is unclear whether only right-handed participants were recruited or whether participants were randomly selected and just happened to be all right-handed.
Reporting bias
The final two criteria were implemented to reduce the risk of selection bias (i.e. systematic differences between baseline characteristics of participants) and were checked in every eligible study. Moreover, we checked every eligible study for duplicates and sample overlaps to reduce the risk of reporting bias. Handedness assessment is very basic and the questionnaires and behavioural tasks used to assess it leave examiners very little room for subjective decisions. Therefore ‘masking’ participants or examiners to prevent a detection or performance bias did not seem vital. Finally, to investigate the robustness of effects and to address the reporting bias across all studies we computed Rosenthal’s fail-safe n. Reference Rosenthal34 The data extraction and statistical analysis were carried out by M.H.
Handedness
In the literature, the by far most common and prolific comparison is between right- and non-right-handers (i.e. left-plus mixed-handers, see also Sommer et al Reference Sommer, Aleman, Ramsey, Bouma and Kahn11 ). However, there is disagreement whether schizophrenia is marked by a general excess of non-right-handedness that includes left- and mixed-handedness Reference Dragovic and Hammond12 or a specific excess of only mixed-handedness. Reference Orr, Cannon, Gilvarry, Jones and Murray20,Reference Satz and Green29,Reference Crow, Done and Sacker35 As a result, we provide separate meta-analyses for left- and mixed-handedness in the online supplement.
Statistical analyses
For each study, ORs were computed so that a value of 1 indicates equal handedness proportions in patients with schizophrenia and controls, OR <1 indicates an excess of non-right-handedness in controls, and OR >1 indicates an excess of non-right-handedness in patients.
In the first analysis we only included studies that provided handedness information for female and male participants (regardless of how handedness was assessed) to investigate whether the excess of non-right-handedness in schizophrenia is a gender artefact. This also enabled us to test whether there is a gender difference with respect to handedness in schizophrenia. In the second analysis, we examined whether behavioural assessment yields an excess of non-right-handedness and whether this excess is greater or smaller than in other forms of handedness assessment (regardless of gender). Finally, we combined gender and behavioural assessment: that is, we examined whether those studies that assessed handedness behaviourally and provided handedness information for male and female participants also revealed an excess of non-right-handedness.
The criteria for handedness categorisation vary substantially across studies. In addition, different diagnostic systems were used, samples were collected in several different countries, and the studies cover a time period of almost 40 years. As a result we chose a random effects model to compute the summary effect within a group (e.g. females) assuming the true effect is not the same in each study. By contrast, when we combined the summary effects to compute an overall effect across groups we used a fixed-effects model, since effect sizes can only fall in one of two categories (female or male, behavioural or other forms of assessment). This is called a mixed-effects model and follows the recommendations of Borenstein et al. Reference Borenstein, Hedges, Higgins and Rothstein36 The variance was assumed to be the same for all subgroups (i.e. Cochran’s Q and I 2 as indicators for heterogeneity were computed within subgroups and then pooled across groups). The meta-analysis was carried out with Comprehensive Meta-Analysis software for Windows version 2.0. Reference Borenstein, Hedges, Higgins and Rothstein37
Results
In total, 540 references were screened. The 50 studies that were included in the meta-analysis Reference Cannon, Byrne, Cassidy, Larkin, Horgan and Sheppard4,Reference Wahl7,Reference Deep-Soboslay, Hyde, Callicott, Lener, Verchinski and Apud16,Reference Kameyama, Niwa, Hiramatsu, Saitoh, Flor-Henry and Gruzelier18,Reference Hauser, Hudson, Hood and Seeman19,Reference Sperling, Martus and Barocka21,Reference Ortuno, Lopez, Landecho and Bonelli22,Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27,Reference Bonner-Jackson, Haut, Csernansky and Barch30,Reference Hinsberger, Williamson, Carr, Stanley, Drost and Densmore31,Reference Crow, Done and Sacker35,Reference Baare, van Oel, Hulshoff Pol, Schnack, Durston and Sitskoorn38-Reference Yan, Flor-Henry, Chen, Li, Qi and Ma74 are provided in the online supplement (Table DS1) and a comprehensive overview of the study selection process is provided in Fig. 1.
Controlling for gender
There were 33 effect sizes covering data from 1834 patients with schizophrenia and 7547 controls, 16 effect sizes for females (patients: n = 621, controls: n = 3747) and 17 effect sizes for males (patients: n = 1213, controls: n = 3800). In females, an OR of 1.63 with a 95% confidence interval (CI) of 1.16-2.30 emerged that deviated significantly from 1 (Z = 2.80, P = 0.005), suggesting an excess of non-right-handedness in schizophrenia. The same pattern was found in males (OR = 1.50, 95% CI 1.14-1.99, Z = 2.84, P = 0.004) and the difference between females and males did not become significant (Q(1) = 0.13, P = 0.722). The combined, overall effect also became significant (OR = 1.55, 95% CI 1.25-1.93, Z = 3.97, P<0.0001) demonstrating that, in general, the odds of being non-right-handed are 1.5 times higher in schizophrenia than in the control group when gender is accounted for (Fig. 2). The overall variance across studies I 2 (pooled across subgroups) was 1.78 (Q(32) = 32.58). Rosenthal’s fail-safe n for the overall effect is 216; thus 216 studies with a null effect (OR = 1) would be required to offset the overall effect.
Behavioural assessment of handedness
Handedness was behaviourally assessed in 1308 patients and 6280 controls (11 effect sizes). Other forms of handedness assessment were available for 2714 patients and 5711 controls (39 effect sizes); thus we had data for 4022 patients and 11 991 controls in total. Both behavioural assessment (OR = 1.90, 95% CI 1.42-2.53, Z = 4.35, P<0.001) as well as other forms of handedness assessment (OR = 1.39, 95% CI 1.15-1.69, Z = 3.39, P = 0.001) revealed significantly greater non-right-handedness in schizophrenia patients than in healthy controls (online Fig. DS1). As expected, the overall effect was also clearly significant (OR = 1.53, 95% CI 1.31-1.80, Z = 5.23, P<0.001, I 2<0.01, Q(49) = 46.36) resulting in a fail-safe n = 481. There was a trend towards greater non-right-handedness in schizophrenia when handedness was assessed behaviourally (Q(1) = 3.05, P = 0.081).
Fig. 1 Selection procedure for the present meta-analysis on handedness in schizophrenia.
Gender effects and behavioural assessment of handedness combined
Eight studies reported the exact number of right- and non-right-handed females (patients: n = 428, controls: n = 3468) and males (patients: n = 819, controls: n = 3458) and assessed handedness behaviourally. As shown in Fig. 3, females (OR = 1.98, 95% CI 1.25-3.14) differed significantly from 1 (Z = 2.89, P = 0.004) and in males a trend towards greater non-right-handedness in schizophrenia emerged (OR = 1.45, 95% CI 0.98-2.15, Z = 1.85, P = 0.065). The difference between the two was not significant (Q(1) = 1.01, P = 0.315) and the combined, overall effect confirmed that schizophrenia patients were more likely to be non-right-handed than controls (OR = 1.65, 95% CI 1.22-2.23, Z = 3.28, P = 0.001, I 2<0.01, Q(15) = 14.34). To offset the overall effect 78 studies with a null effect would be necessary.
Discussion
The present study investigated in a meta-analytical approach whether the excess of non-right-handedness in patients with schizophrenia is confounded by gender Reference Buijsrogge, Benali, Brazo and Dollfus15,Reference Deep-Soboslay, Hyde, Callicott, Lener, Verchinski and Apud16 and/or biases in self-report handedness Reference Byrne, Clafferty, Cosway, Grant, Hodges and Lawrie27 and the results were unequivocal: patients are indeed more likely to be non-right-handed than controls and this is neither the result of a gender artefact nor of biases in handedness questionnaires.
No gender artefact or handedness assessment bias
It is tempting to assume that the excess of non-right-handedness in schizophrenia is a gender artefact when one considers that males are more often non-right-handed than females in the general population, and that in schizophrenia samples there are often more males than females. Indeed, the present study also found 66% males in the schizophrenia and 50% males in the control sample. However, this imbalance could not account for the excess of non-right-handedness: both female and male patients with schizophrenia had significant ORs of about 1.6. Moreover, we did not find significant gender differences. This argues against claims that the excess of non-right-handedness in schizophrenia only exists in either females Reference Hauser, Hudson, Hood and Seeman19,Reference Orr, Cannon, Gilvarry, Jones and Murray20 or males. Reference Sperling, Martus and Barocka21,Reference Ortuno, Lopez, Landecho and Bonelli22 The present study also clearly demonstrated that there is a surplus of non-right-handers in schizophrenia, if handedness is assessed behaviourally. In fact, there was a trend for higher ORs when handedness was assessed behaviourally than compared with other forms of handedness, of which the vast majority were self-report handedness questionnaires. Thus, if anything, the excess of non-right-handedness is even bigger than self-report questionnaires suggest. Finally, even when possible confounding gender and handedness assessment effects were controlled for simultaneously, we observed a significant excess of non-right-handedness in females, a trend in males, and a highly significant combined effect.
Converging evidence
The ORs obtained in the present study were in the range of 1.4-2.0. This very well matches the findings from two previous meta-analyses. Reference Sommer, Aleman, Ramsey, Bouma and Kahn11,Reference Dragovic and Hammond12 One might wonder whether this is simply the consequence of including similar studies. However, although there is naturally some overlap across the three meta-analyses, the difference in study selection is considerable. This is due to different inclusion criteria, new studies (at least 15 studies between 2005 and 2013 were added), and a different research focus (gender differences). Thus, there is compelling converging evidence that the excess of non-right-handedness in schizophrenia truly exists and that it has a magnitude centring around OR = 1.6. That is, if 10% of the healthy population is non-right-handed, then approximately 15% of patients with schizophrenia are non-right-handed. An OR = 1.6 is a small effect (Cohen’s d = 0.26) but Rosenthal’s fail-safe n suggest that it is fairly robust. For instance, it would require 78 studies with null findings to abolish the excess of non-right-handedness when the participants’ gender is included and handedness assessed behaviourally. The number is even higher for gender (n = 216) and behavioural handedness assessment alone (n = 481).
Fig. 2 Forest plot for non-right-handedness in males v. females.
Fig. 3 Forest plot for non-right-handedness in males v. females when handedness was assessed behaviourally.
No excess in consistent left-handers
The main focus of the present study was on non-right-handedness, but the results for left- and mixed-handedness revealed similar results (see online supplement). Interestingly, however, there was no excess of patients with schizophrenia when strong left- v. strong right-handers was compared (with all mixed-handers excluded). This would suggest that the excess in left-handedness is mainly driven by left-handers with a weak hand preference and that degree not direction of handedness is linked to schizophrenia, supporting the notion that schizophrenia is mainly characterised by an ‘absence’ or ‘lack’ of hemispheric specialisation. Reference Crow2 On the other hand, statistical power to detect an excess was much lower in the strong left- v. strong right-handers as compared with the general left- v. right-handers analysis. Thus we are a bit cautious with claiming that the excess of non-right-handedness in schizophrenia is only present in mixed- and not left-handers.
Genetic link
It has been suggested that the excess of non-right-handedness in schizophrenia is the result of subtle brain damage, Reference Satz and Green29 but there is little support for this notion. Reference Dragovic and Hammond12 As McManus Reference McManus75 pointed out, it is generally questionable whether pathological left-handedness exists. The genes and exact mechanism that underlie handedness have not been identified, but the empirical evidence clearly demonstrates that hand preference is under strong genetic influence. Reference Annett76,Reference McManus, Davison and Armour77 The most plausible explanation is thus that the genetic origin of non-right-handedness is linked to schizophrenia as suggested by Crow. Reference Crow2,Reference Crow78
As with all meta-analyses, the findings of the present study are dependent on the studies that were included. We used quite restrictive selection criteria similar to Sommer et al, Reference Sommer, Aleman, Ramsey, Bouma and Kahn11 but the fact that Dragovic & Hammond Reference Dragovic and Hammond12 found similar results with more liberal criteria implies that the excess of non-right-handedness is a reliable finding. We did not include studies with psychiatric control participants, because we wanted to specifically investigate whether non-right-handedness is overrepresented in schizophrenia as compared with healthy controls. It would be interesting to examine, however, whether this excess is specific to schizophrenia or also emerges in other psychiatric patients. Dragovic & Hammond Reference Dragovic and Hammond12 included psychiatric samples but did not specifically compare them with patients with schizophrenia. A recent study Reference Webb, Schroeder, Chee, Dial, Hana and Jefee79 reported that the prevalence of left-handedness in schizophrenia is 40% compared with 11% for patients with mood disorder (OR = 7.9), rather implying the excess of non-right-handedness is typical for schizophrenia. One may argue that due to the large variability in handedness assessment and categorisation, it is questionable whether those studies can be compared and aggregated. We tried to resolve this issue statistically by assuming a random-effects model, in which the true underlying effect can differ from study to study. Finally, it should be pointed out that handedness is only one marker of lateralisation and the finding of an excess of non-right-handedness in schizophrenia - however robust it is - only suggests that brain lateralisation and schizophrenia are linked. The literature on neuroscientific evidence for abnormalities in structural or functional hemispheric asymmetries is less rich and consistent than the excess of non-right-handedness. Recently, however, Crow Reference Crow80 provided an overview of neuropsychological and neuroimaging findings that, taken together, suggest anomalous lateralisation in schizophrenia also manifests in the brain.
We took great care to minimise the risk for the selection and reporting bias. On the individual study level, we checked whether samples had been pre-selected for handedness. Studies were excluded when samples consisted solely of right- or left-handers and when studies stated they had handedness-matched samples without specifying whether the matching was intended or coincidence. The vast majority of studies with only right-handers (or left-handers) were neuroimaging studies and from our own experience in this field we know that often only right-handers are recruited to keep the study design simple and to avoid large variance in the data. Moreover, since handedness is believed to be associated with schizophrenia, many researchers make sure that their studies have handedness-matched samples. We thus believe we had good reasons to exclude those studies to prevent a selection bias, but it may be argued that excluding them represents a selection bias itself, and the OR would be lower, if they were admitted to the meta-analysis. One may also argue that there is a risk of a reporting bias since publishing studies with significant handedness effects may be easier. We aimed to control for that statistically across all studies by calculating Rosenthal’s fail-safe n and indeed find it hard to believe that up to 481 studies with null findings have been gathering dust in file drawers.
In summary, the present study clearly demonstrated that patients with schizophrenia are more often non-right-handed and that this cannot be attributed to a gender artefact or hidden biases in handedness questionnaires. With both points of criticism disproven, there is little doubt left that the excess of non-right-handedness truly exists, which strongly supports the view that there is a genetic link between handedness, brain lateralisation and schizophrenia. Reference Crow2 In addition, the present data indicate that the excess of non-right-handedness is mostly the consequence of a higher proportion of mixed-rather than left-handers, suggesting that strength of handedness rather than direction is linked to schizophrenia.
Acknowledgements
We thank Dr Tsutomo Takahashi, Professor Richard Lewine and Amy Deep-Soboslay for sharing their data, and Alex Craven for his helpful comments and suggestions.
eLetters
No eLetters have been published for this article.