You are on page 1of 11

This article was downloaded by: [McGill University Library

On: 27 December 2012, At: 22:00
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,

Laterality: Asymmetries of Body,
Brain and Cognition
Publication details, including instructions for authors
and subscription information:

Modification of the Edinburgh
Handedness Inventory: A
replication study

Sanja Milenkovic & Milan Dragovic

b c


Institute for Hygiene and Medical Ecology, University
of Belgrade, Belgrade, Serbia

North Metropolitan Area Health Service – Mental
Health, Graylands Hospital, Perth, Western Australia

Centre for Clinical Research in Neuropsychiatry,
University of Western Australia, Perth, Australia
Version of record first published: 02 Jul 2012.

To cite this article: Sanja Milenkovic & Milan Dragovic (2012): Modification of the
Edinburgh Handedness Inventory: A replication study, Laterality: Asymmetries of Body,
Brain and Cognition, DOI:10.1080/1357650X.2012.683196
To link to this article:

Full terms and conditions of use:
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sublicensing, systematic supply, or distribution in any form to anyone is expressly
The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should be
independently verified with primary sources. The publisher shall not be liable

. demand. or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. proceedings. claims. actions.Downloaded by [McGill University Library] at 22:00 27 December 2012 for any loss.

1080/1357650X. The results also suggest that the impact of modification did not affect classification of individuals into left.and right-handedness categories but did influence classification of mixedhanders with the original version classifying significantly more individuals into mixed-handedness category than the modified version.g. Edinburgh Handedness Inventory. Belgrade. Removal of a few problematic items improved internal consistency of the questionnaire and improved its validity. Australia The Edinburgh Handedness Inventory was administered to 1224 high school students (605 males and 619 females). However. Nevertheless these assumptions do not preclude the Address correspondence to: Sanja Milenkovic. Serbia. Confirmatory factor analysis was used to reexamine its psychometric properties. E-mail: sanjavecko@yahoo. Graylands Hospital. Keywords: Confirmatory factor analysis.3 1 Institute for Hygiene and Medical Ecology. iFirst. laterality assessment is often conducted to obtain secondary measures. Belgrade.psypress. Measurement models. University of Belgrade.683196 . Institute for Hygiene and Medical Ecology. The results showed that this instrument has poor measurement properties. Another rationale may lie in their plain face validity. 19 Modification of the Edinburgh Handedness Inventory: A replication study Downloaded by [McGill University Library] at 22:00 27 December 2012 Sanja Milenkovic1 and Milan Dragovic2.LATERALITY. # 2012 Psychology Press. This is perhaps the principal reason as to why handedness questionnaires are rarely subjected to rigorous investigations of their psychometric properties. Serbia 2 North Metropolitan Area Health Service  Mental Health. School of Medicine. an imprint of the Taylor & Francis Group. e. Measurement of lateralisation of hand preferences (direction and degree) by various inventories is common practice and employed across various disciplines. Western Australia 3 Centre for Clinical Research in Neuropsychiatry. an Informa business http://www.. Handedness.doi.2012. University of Belgrade. University of Western Australia. http://dx. 2012. to remove or to control for the confounding effect of left-handedness from MRI studies.

On the other hand a small sample (including a mix of mentally well relatives of schizophrenia patients and healthy controls) used in this study might raise a doubt as to whether the results would actually survive replication. writing and drawing). A quick inspection of the literature reveals that two handedness questionnaires. This study has been well received and followed by about 20 subsequent studies that used the modified version of the EHI recommended by Dragovic. 1980.g. In addition to several factor-analytic studies (Bryden. It was shown that the EHI contains items that are almost singular (e. 1977) has suggested a simplification of the EHI as a way to assess the primary handedness factor. Google scholar lists thousands of articles since 2004 containing the phrase ‘‘edinburgh handedness inventory’’. These two limitations have instigated the current study. and also items that are more a benign reflection of the environment than a genuine hand choice (e. This sample comprised 605 male students (mean age 15. Williams.9. have monopolised the laterality arena: the Edinburgh Handedness Inventory (Oldfield. METHOD Participants The complete EHI was administered individually to a large sample of 1224 students from six government high schools in Belgrade. as is assumed by their designers. McFarland & Anderson. 1976. 2004) has convincingly demonstrated that the Edinburgh Handedness Inventory (EHI) is a poor research instrument that demands modification.. 1977. is to corroborate the results of Dragovic’s (2004) attempt to provide a psychometrically enhanced version of the EHI. To our best knowledge this reasonable suggestion has never been followed up. therefore. 1971) and Annett’s Handedness Questionnaire (Annett.42) and 619 female . For example. A more recent study on a relatively small sample (Dragovic. The principal aim of this study.g. there were many more studies that continued with the unmodified versions. for better or for worse. with the former being used more frequently (Bishop. Serbia. Despite their widespread use these questionnaires have only sporadically been submitted to psychometric scrutiny. 1996). opening a box lid).. Only one factor-analytic study (Bryden. Clearly a successful replication of the original study using a large and more homogeneous sample is warranted.Downloaded by [McGill University Library] at 22:00 27 December 2012 2 MILENKOVIC AND DRAGOVIC question of whether handedness questionnaires are really valid and reliable. This clearly indicates that the impact of Dragovic’s effort (2004) is somewhat limited. We used confirmatory factor analysis (CFA) to re-investigate latent structure and measurement properties of the EHI. White & Ashton. 1986) little effort has been made to truly dissect their validity and reliability. 1970). However. SD0.

a tau-equivalent model assumes that each item is an equally accurate indicator of the composite score. and CFI. SD0. Written informed consent was obtained from all participants. As in the original study a congeneric measurement model. less strong (). . Statistical analysis Confirmatory factor analyses were conducted using LISREL (version 8. the responses were recoded into a 5-point Likert-type scale. For each item. This was necessary to decide which way of creating a composite EHI score is justified. 2004). to indifferent (/ ). Strong left-handedness was given a score of 1 and strong right-handedness a score of 5. First. As in the original study (Dragovic. In addition to the chisquare statistic.HANDEDNESS MEASUREMENT 3 students (mean age 15. measurement models differ with regard to the content of both useful and error variance in items that are indicators of the composite score. GFI. Downloaded by [McGill University Library] at 22:00 27 December 2012 Instrument The original Edinburgh Handedness Inventory comprises 10 items indexing hand preference.51). a parallel model assumes that each item carries an equal prediction of the composite score. i. Third is a congeneric model. but with a unique contribution to the composite score and an item-specific error variance. RESULTS First we compared the fit of three measurement models for the original 10item scale. SRMR. comparative fit index).8.52). The lack of statistically significant differences between the three models justifies calculation of a total handedness score by simply summing responses to each item. AGFI. tau-equivalent. Otherwise each item needs to be weighted separately to reflect their imbalanced contribution to the total score. pre-processing software. Prior to this. standardised root mean-square residual. and congeneric) were tested prior to any modification of the EHI and after modification. Three measurement models (parallel. root mean square error of approximation. in which each item is assumed to indicate the same generic true score. participants indicated their hand preference ranging from strong ().e.. goodness of fit index. polychoric coefficients and an asymptotic covariance matrix were generated in PRELIS. Again scale reliability was assessed using Fornell and Larcker’s (1981) approach to calculate the amount of variance extracted. but allows error variances to differ. Second. adjusted goodness of fit index. the fit of all subsequent scale modifications is assessed by evaluating a number of other indices (RMSEA. In brief.

Factor weights.05. However. were condensed within two items only (writing and drawing) indicating a fairly uneven contribution of individual items to the total score. suggesting that error variance outweighs variance attributable to the construct. Table 4 shows an agreement TABLE 1 Chi-square measures of fit for three measurement models for the Edinburgh Handedness Inventory Model Parallel Tau-equivalent Congeneric# x2 Chi-square. tau-equivalent. construct reliability and estimates of variance extracted were not impressive.2 53 44 35  173. * p B.90) and fairly balanced factor scores. Elimination of this item produced a slightly better. and several indices of fit for all models (baseline and subsequent modifications). All other indices of fit failed to reach conventional thresholds.3 1743. Several items had squared multiple correlations lower than 0. Second we assessed the goodness of fit of the complete 10-item scale. x2 df x2 1570. fit. factor scores (i. 1981). weights). and congeneric model were performed (Table 3). which are supposed to quantify capacity of each item to indicate the construct.74 and slightly larger proportion of the shared variance in a set of handedness items. Next we removed two items that again had the lowest contribution to the construct of measurement (opening box lid and holding broom).. Large absolute values of chi-square statistic suggest that there might be room for further scale refinement.6 1501. # Congeneric versus parallel. Figure 1 shows a scatter plot of laterality quotients from the original and shortened version.Downloaded by [McGill University Library] at 22:00 27 December 2012 4 MILENKOVIC AND DRAGOVIC each item contributes uniquely to the composite score and has an itemspecific error variance. Elimination of these two items resulted in improved indices of fit (all above 0.e. The second modification led to a slightly improved construct reliability of 0. was found to be superior (Table 1) to the other two models.1* diff Ddf  9 18 . model-based measures of reliability (Fornell & Larcker.50. The first modification we implemented was to exclude the highly collinear and thus redundant item: drawing. Table 2 contains squared multiple correlations for observed variables.3* 69. but still unacceptable. Finally the impact of the concluding modification on both laterality quotients and the categorical classification was examined and the tests of differences between parallel. Again the complete 10-item scale showed unsatisfactory indicators.

Table 4 also demonstrates that prevalence of mixed-handed individuals was significantly higher with the original version .41 .36 . # difference in chi-square between first and second modification (Dx2 49.65 .39 Variance extracted estimate 6 178.80 AGFI .28 . CFI  Comparative fit index.22  .4# 14 .32 .28 .02 .1 Degrees of 35 freedom (df) RMSEA .31  .91 Construct .13 .53 .19 .95 . x2 chi-square.42 .07 .08 .11 . factor score regressions (specific weight of each item).09 .55  .80.94 .89 .05).14 .05 .98 .07 .12 .85 .11 .98 .01 .05 .14 .34 .22 .66  .97 .17  Goodness-of-fit measures No of iterations 11 x2 1501.17 .05).04 . AGFI  Adjusted Goodness-of-Fit Index.04 .97 .3§ 27 6 128.40 .03 .48 One-factor congeneric measurement models for the EHI and the two modifications.07 .09 RMR GFI . mixed-.85 reliability .49 . between the two categorical assignments (left-.03 . including squared multiple correlations (proportion of variance explained by the construct).03 .9) is significant (p B.87 . § difference in chi-square between baseline and first modification (Dx2 1322.19 Standardised .19 .88 . pB.40 .5 HANDEDNESS MEASUREMENT TABLE 2 One-factor congeneric measurement models for the EHI and the two modifications Baseline 10-items Downloaded by [McGill University Library] at 22:00 27 December 2012 Observed variables Writing Drawing Throwing Scissors Toothbrush Knife Spoon Broom Matches Opening box-lid First modification 9-items Second modification 7-items Squared multiple correlations Factor scores Squared multiple correlations Factor score Squared multiple correlations Factor scores . RMSEA  Root Mean Square Error of Approximation.11 .47  . and goodness-of-fit measures.41 .001).69 CFI .12 .37 .57  . Standardised RMR  Standardised Root Mean-square Residual.16 .8) is significant (p B.19 . The overall agreement was significant (Kappa 0.02 .53 .21  .36 . GFI  Goodness-of-Fit Index.31 .56 . and right-handed) using the cut off score of 0960.

8) 645 (98.5%) were classified as mixed-handers according to the original version. Similar to the modification proposed by Dragovic (2004) the present study demonstrates that the modified version is statistically superior to the full scale.2) 657 37 433 754 1224 .1 209.4* 81. Removal of a few items generally improved the psychometric properties of this instrument. * p . Both studies. These two items have already been described as items that do not matter (Peters.6 MILENKOVIC AND DRAGOVIC TABLE 3 Chi-square measures of fit for three measurement models for the 7-item scale Model Downloaded by [McGill University Library] at 22:00 27 December 2012 Parallel Tau-equivalent Congeneric# x2 df Dx2 Ddf 605. # Congeneric vs Tau-equivalent.2%) and right-handers (14.3) 536 0 (0.(16.7 128. 1990) or items that do not load on to the primary TABLE 4 Agreement of categorical classifications between the original and modified version (% within the original EHI) EHI original version EHI modified version Total (n) Left Mixed Right Left n (%) Mixed n (%) Right n (%) Total (n) 31 (100) 0 (0.1) 421 (78. Approximately the same proportions of left. We suggest that the modified version should be preferred in research settings.5) 109 (20.0) 31 6 (1.0) 0 (0. and two items (holding a broom and opening a box lid) due to an unacceptably low amount of shared variance with the handedness construct. DISCUSSION Similar to the original study (Dragovic. original and present.0) 12 (1.2*  6 6 x2 Satorra-Bentler Scaled Chi-square.05. 2004) confirmatory factor analysis using a large and homogeneous sample of students clearly demonstrates that the EHI should be used in a modified form. One item (drawing) was problematic due to its excessively high correlation with writing. (43.4%). primarily on the basis of two problematic items (‘‘holding broom’’ and ‘‘opening box lid’’).4 22 20 14  395.8%) than with the modified version (35. encountered the same problematic items.

tauequivalent. There are two discrepancies with the original study. should not be used in any handedness questionnaire. and Bender (2010) also observed that elimination of those two items improves measurement quality of the EHI. due to their marginal contribution to the measurement construct.Downloaded by [McGill University Library] at 22:00 27 December 2012 HANDEDNESS MEASUREMENT 7 Figure 1. and congeneric). some items appear to be better indicators of the latent . i. Hagemann. 1994). namely the lack of significant differences between the three measurement models (parallel. in the second modification we failed to replicate an important result. Similarity of these models warrants the calculation of a simple sum of scores to obtain the EHI total score. The present results demonstrate that contribution of each item to the total score is slightly different. Measurement of these items is also notoriously unreliable (Ransil & Schacter. Recently. A jitter function was used to improve the look of scatter plot by avoiding extensive overlap of data points. Bu¨sch. handedness factor (Bryden.e. However.. First. using a Mixed-Rasch model. our study suggests this may not be true. and which. Scatterplot of laterality quotients with regression line and R2 from the original and shortened version of the Edinburgh Handedness Inventory. 1977).

(1977). Manuscript received 21 December 2011 Manuscript accepted 21 March 2012 First published online 3 July 2012 REFERENCES Annett. The measurement of hand preference: A validation study comparing three groups of right-handers. C. Hagemann. McCartney & Hepper. Hepper. M. 411419. 1999). M. D. McCartney. hand preferences appear to be well established before birth (e. & Larcker.. 18. 531534. P. 87. D. The dimensionality of the Edinburgh Handedness Inventory: An analysis with models of the item response theory. Practical reason advocates that the seven remaining items have equal weighting for calculation of the total score. Bryden. (2004). 15. significant age differences between the two samples (original and present) may hint that the younger sample exhibits developmental differences related to prolonged maturation of hand dominance. Laterality. Hepper. (1998). N.. Lateralised behaviour in first trimester human foetuses. M. McCartney.Downloaded by [McGill University Library] at 22:00 27 December 2012 8 MILENKOVIC AND DRAGOVIC handedness construct than others. S. Bu¨ sch. G. Fornell.. 15. un-established clear hand preferences. .. (2003). Wells. 617624. 303321. is often under scrutiny for its associations with some clinical and psychiatric conditions.. Evaluating structural equation models with unobservable variables and measurement error.. Journal of Marketing Research. & Lynch 2005.e. McManus et al. M. P. A. Measuring handedness with questionnaires. 36. Neuropsychologia.. Dragovic. However. & Bryden. i. Hepper. Laterality. (2010). 9.. the modified version prevents an unjustifiable categorisation of individuals with otherwise clear right hand preferences into the mixed category. (1996). Second. (1981). & Bender. Moreover. 1988). & Shanon. or to calculate the laterality quotient. Neuropsychologia.g. Development of handedness: Comparison of questionnaire and performance-based measures of preference. although weights in the present study appear fairly balanced. We have demonstrated that modification of the EHI still provides an excellent agreement when classifying individuals with clear hand preferences. D. Importantly. N. R. However. Towards an improved measure of the Edinburgh Handedness Inventory: A one-factor congeneric measurement model using confirmatory factor analysis. E. British Journal of Psychology. 269285. This is important because the middle handedness category. 61. 610628. G. Bishop. they cannot be used to calculate total scores in samples from different populations. V. Cavill. A classification of hand preference by association analysis. Brain and Cognition. & Shanon 1998. hand preference development appears to be strongly established by middle childhood (Cavill & Bryden. we conclude that our replication study has additionally strengthened the need for using the modified EHI version instead of the psychometrically flawed original version. P. 149151. British Journal of Psychology. 2003. In summary. (1970). 3950. 53..

S. McFarland. P. 257273. and sex.. & Schachter. In S. Prenatal thumb sucking is related to postnatal handedness. 71. 261264. G.. 43. R. 6. (1988). Phenotype in normal left-handers: An understanding of phenotype is the basis for understanding mechanism and inheritance of handedness. B. M. G. 167192). & Anderson.. Wells. 313315. 325326. 14. 79. 22. D. C. (1999). J. & Hepper. R. (1976). Developmental Medicine and Child Neurology. J.. D. Neuropsychologia. 13551372. C. (1971). A. Ransil. 135142. (2005). . K. Cortex.Downloaded by [McGill University Library] at 22:00 27 December 2012 HANDEDNESS MEASUREMENT 9 Hepper. Neuropsychologia. S.. Lefthandedness: Behavioural implications and anomalies (pp. McManus. Wong. Mellon. & Lynch. & Kloss. Peters. P. North-Holland: Elsevier Science Publishers BV. British Journal of Developmental Psychology.. Neuropsychologia. C. Coren (Ed. Perceptual and Motor Skills.. Factor stability of the Edinburgh Handedness Inventory as a function of testretest performance. I. McCartney. R. Sik. The development of handedness in children. Handedness assessment inventory. British Journal of Psychology. 97113. age. K. 8386.. J. (1986). Williams. L.. The assessment and analysis of handedness: The Edinburgh Inventory. F. (1990). Cole. White. (1980). M. Testretest reliability of the Edinburgh Handedness Inventory and global handedness preference measurement. J. Development of lateralized behaviour in the human fetus from 12 to 27 weeks’ gestation.). 9.. and their correlation. 41. G.. C (1994). Oldfield. & Ashton. Factor analysis of the Edinburgh Handedness Inventory.