Lateral preferences and their assessment in school-aged children

A behaviour-based lateral preference instrument (the Lateral Preferences Questionnaire; LPQ; Van Strien, 1992, 2001) was administered to a large sample of school-aged children. The aims of the present study were twofold: (i) to evaluate the factor structure and the psychometric properties of the LPQ, and (ii) to evaluate the effects of age, gender, and mean level of parental education on lateral preferences and lateral consistency. Two factor models had an excellent fit with the data. In the first model the LPQ items were considered to be indicators of four different lateral preference factors (the hand, foot, eye, and ear preference factors). In the second model the LPQ items were considered to be indicators of four lateral preference factors, which were in turn expected to load on a single underlying general lateral preference factor. The psychometric properties of the derived hand and eye preference scales of the LPQ were good to excellent, and the psychometric properties of the foot and ear preference scales were acceptable. Lateral preferences and lateral consistency were not affected by age, gender, or mean level of parental education.

Keywords: Hand preference; Foot preference; Eye preference; Ear preference; Assessment; School-aged children

In neuropsychology and related scientific disciplines, hand preference is often used as an indirect indicator of brain lateralisation. Many studies have evaluated the associations between hand preference and various problems and conditions—including a shortened lifespan, developmental and psychiatric problems, gender identity disorders, and cognitive deficits—but few consistent results have emerged from these studies (Bishop, [4]; Coren & Halpern, [13]; Farwell & Temkin, [18]; Van der Elst, Van Boxtel, Van Breukelen, & Jolles, [35], [36]; Zucker, Beaulieu, Bradley, Grimshaw, & Wilcox, [43]). A possible source for the inconsistencies in the results of studies that involve hand preference is the lack of an agreement on how lateral preferences should be measured (Basso, [3]; Bishop, Ross, Daniels, & Bright, [5]; Dean & Reynolds, [16]; Peters, [28]).

The Edinburgh Handedness Inventory (EHI; Oldfield, [27]) and the Annett Hand Preference Questionnaire (AHPQ; Annett, [1]) are the most frequently used instruments to assess lateral preferences, but these instruments can be improved in at least two ways. First, the EHI, AHPQ, and similar questionnaires focus on hand preference, but hand preference is only one modality of lateral preference. It would be advantageous to have a lateral preference instrument available that assesses the full spectrum of lateral preference modalities (i.e., hand, foot, eye, and ear preferences) (Searleman, Porac, & Coren, [33]). Second, many hand preference instruments rely on self-reported information rather than on direct behavioural observation. Discrepancies between self-reported and behaviour-based indices of lateral preference have been reported (Bishop et al., [5]; Dean & Hua, [15]; Dean & Reynolds, [16]; Doyen & Carlier, [17]; but see also Raczkowski, Kalat, & Nebes, [31]). This finding indicates that the validity of lateral preference instruments can be improved by relying on behaviour-based information rather than on self-reported information. The Lateral Preferences Questionnaire (LPQ; Van Strien, [38], [39]; Vieijra, König, & van Schaik, [40]) fulfils both requirements. In the LPQ the participants are asked to show (using an "imaginary object")—rather than verbally report—how they would perform 23 activities that include writing (hand preference), kicking a ball (foot preference), looking through a small opening (eye preference), and listening to the ticking of a watch (ear preference). In this study the LPQ was administered to a large sample of school-aged children who were aged between 6.6 and 15.9 years. The aims of the present study were twofold.

The first aim was to evaluate the factor structure and the psychometric properties of the LPQ. The LPQ was constructed to be a multifactor instrument that assesses four distinct lateral preference factors (hand, foot, eye, and ear preferences, see Vieijra et al., [40]), but until now its factor structure has not yet been thoroughly evaluated. We evaluated three a priori measurement models. The first model was a unidimensional model in which the LPQ items were considered to be indicators of a single underlying general lateral preference factor. The second model was a multidimensional model in which the LPQ items were considered to be indicators of four different first-order factors (the hand, foot, eye, and ear preference factors). The third model was a combination of models 1 and 2, in which the LPQ items were considered to be indicators of four first-order lateral preference factors (the hand, foot, eye, and ear preference factors), which were in turn considered to be indicators of a single higher-order general lateral preference factor. Model 3 was thus multidimensional at one level of abstraction (i.e., four lateral preference factors were involved), and unidimensional at another level of abstraction (i.e., a single general lateral preference factor was also involved). The fit of these three models was compared to determine the best measurement model for the LPQ.

The second aim of the present study was to evaluate the impact of demographical variables on the hand, foot, eye, and ear preferences of school-aged children. Previous studies have suggested that lateral preferences were affected by demographical variables (especially gender), but the results of these studies were not always conclusive and generally limited to hand preference measures only (Annett, [2]; Brito, Lins, Paumgartten, & Brito, [7]; Cavill & Bryden, [10]; De Agostini & Dellatolas, [14]; Gabbard, [20]; Gentry & Gabbard, [22]; Gudmundsson, [23]; Peters, Reimers, & Manning, [29]; Singh, Manjary, & Dellatolas, [34]). We further evaluated the impact of age, gender, and mean level of parental education on the full spectrum of lateral preference modalities and on overall lateral consistency (i.e., the left/right concordance for the different lateral preference modalities).

Data were derived from a large-scale longitudinal cognitive development study in school-aged children, the COOS (Cognitief Ontwikkelings Onderzoek bij Schoolgaande kinderen [cognitive developmental study in school-aged children]; Wassenberg et al., [41]). Participants were recruited from 29 regular schools in the city of Maastricht and surroundings (The Netherlands). The parents (or caregivers) of the children who attended kindergarten, grades 2, 4, and 6 (primary schools), and grades 7 and 8 (secondary schools) received an information package via the school that included a brief description of the purpose of the COOS study and a form to give consent for the child to participate. Of the N = 1086 parents who replied, n = 892 parents gave consent for their child to participate. Children who (1) had not repeated or skipped a grade, (2) were native Dutch speakers, and (3) did not use medication known to affect cognitive performance—such as Ritalin (methylphenidate)—were eligible for study inclusion. A total of n = 431 eligible children were randomly selected for study participation (with the restriction that the number of selected children per Gender by Grade cell was approximately equal to 35). All children completed a battery of neuropsychological tests including verbal IQ, memory, language comprehension, and time estimation (for details see Meijs, [25]).

The first follow-up measurement of the COOS took place approximately 1 year after the baseline measurement. The parents of the children who still attended one of the participating schools (n = 425) were contacted again to request participation of their child. Of the n = 336 parents who replied, n = 331 parents gave consent for their child to participate. Among these 331 children, n = 8 had repeated or skipped a grade and n = 13 children were diagnosed with a condition known to affect cognitive functioning (or took medication known to affect cognitive performance) between the first and second wave of the COOS, and were thus excluded. The data of n = 14 children could not be collected due to the child's refusal to participate, or because the child was not present at school at the time of testing (due to, e.g., illness or other school activities). Thus a total of n = 296 children were eligible for study participation in the second wave of the COOS. The mean age of these children equalled 11.08 years (SD 2.82, age range 6.56–15.85 years), and 53.7% of the sample were girls. The mean level of parental education (MLPE) was categorised into low (at most primary education; 6.8% of the sample), medium (at most junior vocational education; 56.5% of the sample), and high (at most senior vocational or academic training; 36.7% of the sample) groups, respectively. The COOS test protocol at the second wave consisted of various cognitive tests that were intermixed with the hand, foot, eye, and ear preference parts of the LPQ (in fixed order for all participants). The hand and foot preference parts of the LPQ were administered at the middle of the COOS test protocol, and the ear and eye preference parts were administered towards the end of the test protocol. Not all children could be administered all four parts of the LPQ, because the available testing time per child was limited to a maximum of 1.5 hours (as testing took part during school hours). The hand, foot, eye, and ear preference parts of the LPQ were administered to n = 296, n = 295, n = 182, and n = 177 children, respectively. These data were analysed in the present study. The Ethics Committee of the Faculty of Psychology of Maastricht University approved the study protocol.

Each child was individually tested at school by highly trained test assistants. Lateral preferences were assessed in the four parts of the LPQ. During each part the children were asked to show (using an "imaginary object")—rather than verbally report—how they would perform several activities. The first part consisted of 11 items regarding hand preference for writing, drawing, using a toothbrush, using a bottle-opener, throwing a ball, using a hammer, holding a tennis racquet, cutting a rope with a knife, using a spoon, using an eraser, and striking a match (items 1–11). The second part consisted of four items regarding foot preference for kicking a ball, trampling down a plastic cup, stepping on a chair, and putting the first of a pair of shoes on (items 12–15). The third part consisted of four items regarding eye preference for looking through a small gap, looking in a jar to see how full it is, looking through a microscope, and looking through a magnifier (items 16–19). The fourth part consisted of four items regarding ear preference for listening to people from behind a closed door, listening to the ticking of a wrist watch, listening to someone who whispers, and listening whether an elevator is arriving (items 20–23). For the 23 lateral preference items, possible responses were "left" (score 0), "both" (score 1) or "right" (score 2).

Confirmatory Factor Analysis (CFA) was used to evaluate the factor structure of the LPQ. Three a priori models were specified. In the first model the LPQ items were considered to be indicators of a single general lateral preference factor. In the second model the LPQ items were considered to be indicators of four lateral preference factors (i.e., the hand, foot, eye, or ear lateral preference factors). In the third model the LPQ items were considered to be indicators of four lateral preference factors (i.e., the hand, foot, eye, or ear lateral preference factors), which were in turn considered to be indicators of a single underlying higher-order general lateral preference factor. Models 1, 2, and 3 are referred to as the unidimensional model, the first-order four-factor model, and the higher-order four-factor model, respectively. The LPQ item responses were categorical (each item had only three possible response options, i.e., left preference, right preference, or no preference) and non-normally distributed in the sample (with the majority of children having right lateral preferences). CFA techniques that rely on normal theory estimation using Pearson correlations are not appropriate for this type of data and often lead to biased model fit statistics, biased parameter estimates, and inflated error variances (West, Finch, & Curran, [42]). Instead, the Diagonally Weighted Least Squares method for polychoric correlation matrices should be used for this type of data. In this method the weight matrix is defined as the asymptotic covariance matrix among all polychoric correlations between items (Muthén, [26]). Thus polychoric coefficients and an asymptotic covariance matrix were generated in PRELIS for subsequent analysis in LISREL. Prior to conducting the CFAs, pilot testing of the items to be factor-analysed was conducted to ensure that the items that were designed to measure a common construct correlated at least moderately with each other (i.e., correlations of at least.20; see Floyd & Widaman, [19]). In addition, items that correlated above.90 with at least one of the other items that were designed to measure a common construct were eliminated from the analysis to avoid singularity in the input matrix. Kendall Tau-b correlation coefficients were calculated rather than Pearson correlation coefficients in view of the ordinal measurement level of the items of the LPQ.

Model fit of the different models was evaluated with the Satorra-Bentler χ2 (SB χ2; with p-values>.01 indicating a good model fit) and the Root Mean Square Error of Approximation (RMSEA;<.08 acceptable,<.05 excellent; Browne & Cudeck, [8]). The fit of the three alternative CFA models was compared by means of their RMSEA values. The fit of the nested models (i.e., the first-order and higher-order four-factor models) was further compared by means of a χ2 difference test. In this procedure the χ2 value of the less parsimonious model (i.e., the first-order four-factor model) was subtracted from the χ2 value of the more parsimonious model (i.e., the higher-order four-factor model). This difference value is itself approximately χ2 distributed, with a number of degrees of freedom that equals the difference in degrees of freedom between the two nested models (Reis & Judd, [32]).

The internal consistency (which can be interpreted as a lower bound of scale reliability) of the items of the established LPQ scales was estimated with Cronbach's alpha coefficients (≥ 0.6–0.7: acceptable; ≥ 0.8: high). In addition to Cronbach's alpha, the mean inter-item Kendall Tau-b correlation coefficients of the items of each scale were calculated (which should be about.40 or higher; Clark & Watson, [11]). If 80% or more of the items that loaded on a common lateral preference factor (as identified with the CFA models) were consistently answered with "left" or "right", the child was classified as having a left or right lateral preference, respectively (Van der Elst et al., [36]; Van Strien, [38], [39]). Otherwise, the child was classified as having a mixed lateral preference. Children with a mixed lateral preference were included in the left lateral preference group, in agreement with the literature (Bryden, McManus, & Bulman-Fleming, [9]; Coren, [12]; Porac & Friesen, [30]; Van der Elst et al., [36]). Logistic regression analyses were used to evaluate the effects of age, gender, MLPE, and all two-way interactions between these predictors on the dichotomised lateral preference scale scores (coded as 0 = left preference and 1 = right preference). Gender was coded as 1 = male and 0 = female. The full logistic regression models (including all predictors) were reduced in a stepwise fashion by eliminating the non-significant predictors (p>.01) from the models.

The percentages of children with left/right concordant hand-foot, hand-eye, hand-ear, foot-eye, foot-ear, and eye-ear preferences were calculated, and the effect of demographical variables on overall lateral consistency was evaluated. The overall lateral consistency measure was coded as 4, 3, and 2 for children who had a left/right concordance for all four, three, or two of the four lateral preference measures, respectively.

All analyses were conducted with SPSS 15.0, PRELIS, and LISREL 8.8 for Windows, and R 2.9.2 for Linux. An alpha level of.01 was used in all analyses.

The CFA was conducted on the data of the n = 176 children who had complete data for all items of the LPQ. The mean age (SD) of these children equalled 11.92 years (2.58), 52.1% were girls, and the MLPE was low, mean, and high for 5.8%, 61.0%, and 33.2% of these children, respectively. Prior to conducting the CFAs, pilot testing of the items to be factor-analysed was conducted. The Kendall Tau-b correlation coefficients between two pairs of items that were designed to measure hand preference were above.90—i.e., the correlation between item 1 (writing) and item 2 (drawing) equalled.93 and the correlation between item 2 (drawing) and item 6 (using a hammer) equalled.92. After exclusion of item 2 (drawing) from the analyses, all inter-item Kendall Tau-b correlation coefficients between the items that were designed to measure hand preference were between.30 and.85. The correlations between item 15 (putting the first of a pair of shoes on) and all other items that were designed to measure foot preference were below.20—i.e., the correlations between item 12 (kick a ball) and item 15, item 13 (trample down a plastic cup) and item 15, and item 14 (step on a chair) and item 15 equalled.13,.11, and.17, respectively. After exclusion of item 15, all Kendall Tau-b inter-correlations of the items that were designed to measure foot preference were between.27 and.44. All Kendall Tau-b inter-correlation coefficients between the items that were designed to measure eye and ear preferences were between.49 and.63 and between.27 and.32, respectively (see Table 1).

TABLE 1. Kendall Tau-b correlation coefficients between the items of the Lateral Preferences Questionnaire (LPQ)

Three CFA models were fitted to the data. The unidimensional model had a relatively poor fit with the data (SB χ2=514.07, p<.01; RMSEA=.07). Standardised factor loadings ranged between.46 and.99 (see Figure 1). The Cronbach's alpha and mean inter-item Kendall Tau-b correlation coefficient values for the derived general lateral preference scale (which contained the 21 LPQ items) equalled.91 and.37, respectively. The Cronbach's alpha value thus suggested a high level of internal consistency, while the mean inter-item Kendall Tau-b correlation coefficient suggested a lower level of internal consistency.[1]

Graph: Figure 1. The unidimensional model. Values associated with each path are standardised factor loadings, and values between brackets are error variances.

The first-order four-factor model fitted the data very well (SB χ2=226.37, p>.01; RMSEA<.001). Standardised factor loadings ranged between.67 and.99 for the items that loaded on the hand preference factor, between.55 and.91 for the items that loaded on the foot preference factor, between.86 and.95 for the items that loaded on the eye preference factor, and between.55 and.77 for the items that loaded on the ear preference factor (see Figure 2). The inter-factor correlations ranged between.46 and.84. Internal consistency as estimated with Cronbach's alpha was high for the items of the hand preference and eye preference scales (all values ≥.85), and acceptable for the items of the foot preference and ear preference scales (all values ≥.64). Similarly, the average inter-item Kendall Tau-b correlation coefficients suggested a high internal consistency for the items of the hand and eye preference scales (all correlations≥.55), and a lower internal consistency for the items of the foot and ear preference scales (all correlations≤.35).

Graph: Figure 2. The first-order four-factor model. Values associated with each path are standardised factor loadings, and values between brackets are error variances.

The higher-order four-factor model had a good fit with the data (SB χ2=206.67, p>.01; RMSEA<.001). Standardised factor loadings ranged between.67 and.99 for the items that loaded on the hand preference factor, between.55 and.91 for the items that loaded on the foot preference factor, between.86 and.96 for the items that loaded on the eye preference factor, and between.55 and.78 for the items that loaded on the ear preference factor (see Figure 3). Cronbach's alpha and mean inter-item Kendall Tau-b correlation coefficients were identical to the values for the first order four-factor model described above. The first-order-lateral preference factors had significant loadings on the second-order general lateral preference factor, with standardised factor loadings ranging between.61 and.96.

Graph: Figure 3. The higher-order four-factor model. Values associated with each path are standardised factor loadings, and values between brackets are error variances.

The RMSEA values of the first-order and higher-order four-factor models were similar (i.e.,<.001) and were much lower as compared to the RMSEA value of the unidimensional model (which equalled.07). The fit of the first-order four-factor model and the higher-order four-factor model was further compared by means of a χ2 difference test, which showed that there were no significant differences in the fit of both models (χ2 difference value = 4.48, df=2, p=.11).

Figure 4 shows the distribution of the raw hand, foot, eye, and ear preference scale scores (based on the data of n = 296, n = 295, n = 182, and n = 177 children, respectively). Right hand, foot, eye, and ear preferences were observed for 83.6%, 80.9%, 69.1%, and 57.9% of the children, respectively. Logistic regression analyses showed that none of the demographical variables significantly affected the dichotomised lateral preference measures.

Graph: Figure 4. Distribution of the raw hand (a), foot (b), eye (c), and ear (d) preference scale scores. The figures on the left and right present percentages and cumulative percentages, respectively.

Graph

Concordant left/right hand-foot, hand-eye, hand-ear, foot-eye, foot-ear, and eye-ear preferences occurred in 85.7%, 72.4%, 68.8%, 71.3%, 67.0%, and 62.3% of the children in the sample, respectively. The overall lateral consistency measure equalled 4, 3, and 2 for 17.0%, 35.8%, and 47.2% of the children, respectively. Ordinal regression analyses showed that the overall lateral consistency measure was not significantly affected by age, gender or MLPE.

The most frequently used lateral preference instruments focus on hand preference and rely on self-reported information. It would be advantageous to have a lateral preference instrument available that assesses foot, eye, and ear preferences in addition to hand preference, and which relies on direct behavioural observation rather than on self-report. The LPQ (Van Strien, [38], [39]; Vieijra et al., [40]) is such an instrument. The main aims of the present study were (i) to evaluate the factor structure and the psychometric properties of the LPQ, and (ii) to evaluate the effects of age, gender and MLPE on lateral preferences and their consistency.

Three a priori factor models were evaluated. The unidimensional model had a relatively poor fit with the data and can be rejected as an appropriate LPQ measurement model. The first-order and higher-order four-factor models had an excellent fit with the data. Both models fitted the data equally well (as evidenced by their RMSEA values and the non-significant chi-squared difference test), so statistical criteria cannot be used to determine the "best" measurement model. Non-statistical criteria, such as parsimony, may be used to tentatively guide the choice for the most appropriate measurement model (which would then be the higher-order four-factor model), but ultimately theoretical considerations regarding the "true nature" of the lateral preference construct should be decisive in the choice for the most appropriate measurement model. At present, there is no agreement regarding the "true nature" of the lateral preference construct (Dean & Reynolds, [16]). Future studies in which multiple theoretical lateral preference models are fitted on large data sets that are collected by means of a variety of methods (e.g., direct behavioural observations of lateral preferences corroborated by neuroimaging data) in various populations (e.g., in children, adults, and elderly) may shed more light on this issue. Note that from an applied psychometric viewpoint, both the first-order and the higher-order four-factor models justify the clustering of the LPQ items in four lateral preference scales. The main difference between both models is that the higher-order four-factor model also explicitly justifies the construction of a single higher-order general lateral preference measure; i.e., a single composite measure that combines the hand, foot, eye, and ear preference scale scores of an individual. There are many different ways to construct such a composite measure. For example, a procedure may be used in which the standardised lateral preference scale scores are multiplied by their relevant standardised factor loadings and aggregated. The aggregated values are then divided by the SD of the aggregated measure in a normative sample (for details see Van der Elst, Van Boxtel, Van Breukelen, & Jolles, [37]). Such a procedure yields statistically sound summary measures that are easily interpretable (as Z-scores), but it has the disadvantage that it requires quite some calculation. Alternatively, a simple higher-order general lateral preference measure can be constructed by summing up the four dichotomised lateral preference scale scores (coded as 0 = left preference and 1 = right preference). This simple higher-order general lateral preference measure has a score range between 0 (=a left lateral preference for all four lateral preference measures) and 4 (= a right lateral preference for all four lateral preference measures). Figure 5 presents the distribution of the higher order general lateral preference measure in our sample.

Graph: Figure 5. Distribution of the higher-order general lateral preference measure. Percentages are presented in the left figure, cumulative percentages are presented in the right figure.

The majority of the children in our sample had right lateral preferences (i.e., 83.6%, 80.9%, 69.1%, and 57.9% of the children had right hand, foot, eye, and ear lateral preferences, respectively). Lateral preferences were not affected by age or MLPE, in agreement with most previous studies (Annett, [2]; Cavill & Bryden, [10]; Gabbard, [20]; Longoni & Orsini, [24]). Gender was also not significantly associated with lateral preferences (i.e., left hand, foot, eye, and ear preferences were seen in 17.9%, 19.9%, 28.6%, and 40.0% of the girls, respectively; and in 14.5%, 18.1%, 33.7%, and 44.4% of the boys, respectively). However, Peters and colleagues (2006) did find significant effects of gender on hand preference in an Internet-based study that included more than a quarter of a million people. The discrepancy between our results and the results of Peters and colleagues may be attributable to multiple factors, including differences in how the lateral preferences were measured in both studies (Peters and colleagues only measured hand preference for writing), differences in the population characteristics of both studies (the Peters et al. study mainly included adults and elderly), or differences in sample sizes and statistical power (the Peters et al. study included more participants and had more power to detect gender effects).

Several remarks regarding the design and the limitations of the present study can be made. First, the semi-continuous lateral preference scale scores were dichotomised in agreement with the literature (e.g., Bryden et al., [9]; Coren, [12]; Porac & Friesen, [30]; Van der Elst et al., [36]), but not all researchers may agree with this classification scheme. For example, it has been suggested that lateral preferences are better represented as semi-continuous variables (Annett, [2]; Dean & Hua, [15]). A major problem with using semi-continuous lateral preference measures rather than categorised lateral preference measures is that the statistical methods that are commonly used to analyse semi-continuous dependent variables assume a normal distribution of the residuals and a homogeneous distribution of the residuals over the entire range of predicted values. Both assumptions are almost certainly violated when semi-continuous lateral preference data are analysed because of their extreme non-normal distribution (see Figure 4). As a consequence the conclusions that are drawn from such analyses may be biased.

Second, about 40% of the sample could not be administered all four parts of the LPQ within the available testing time. The CFAs were conducted on the data of the n = 176 children who completed all four parts of the LPQ within the available test time (i.e., 60% of the data of the total sample). It has been argued that conducting a CFA requires a sample of at least N = 200 participants (Boomsma, [6]), but there is little empirical evidence for such a claim. Gagné and Hancock ([21]) argued that CFA sample size considerations should rely on the quality of the measurement model (e.g., the magnitude of the factor loadings and the number of indicator variables per construct), rather than on such a general "rule of thumb". For example, CFA models with high average standardised factor loadings (e.g., mean factor loadings that are above.60 or.80) typically require samples sizes as small as N = 100 people to achieve proper model convergence and accurate parameter estimates. On the other hand, models with lower standardised factor loadings (e.g., mean factor loadings of about.40) may not converge with sample sizes as large as N = 1000. The mean standardised factor loadings in our models were high, which suggests that our sample size of n = 176 was sufficient to obtain accurate parameter estimates.

On a related note, logistic regression analyses were conducted to evaluate the effect of demographical variables on the ability of a child to finish the four parts of the LPQ within the available test time. Age was found to have a significantly positive effect on the ability of a child to finish the four parts of the LPQ (data not shown). The mean (SD) age of the subsample of the children who could, and of the children who could not, be administered all four parts of the LPQ within the available test time equalled 11.92 (2.58) and 9.81 (2.68) years, respectively. Thus the results of the CFAs on the LPQ were mainly based on the data of the older children in our sample. It is unknown whether the results of the CFAs as based on a sample of younger versus older children would be equivalent. This question cannot be reliably addressed with the present dataset, because dividing the sample in half (after a median split of age) would result in two samples of n = 88 (=176/2) children. These samples are likely to be too small to yield reliable parameter estimates, so more data need to be collected and additional analyses are required to further evaluate this issue.

Third, we presented the hand, foot, eye, and ear preference parts of the LPQ in fixed order. It is possible that (potential) carry-over effects between the four LPQ test parts (and hence lateral consistency indices) vary as a function of the order of presentation of the LPQ test parts. This possibility could not be evaluated in the present study because the order of presentation of the different LPQ parts was fixed. Future studies should consider this possibility and estimate the magnitude of (potential) carry-over effects.