Associations between motor proficiency and academic performance in mathematics and reading in year 1 school children: a cross-sectional study

This study was conducted in parallel with the Tweed Healthy Schools Project (THSP), an interprofessional clinical placement program for university health science students based in a school setting. A cross-sectional research design was employed, examining data collected at the start of the THSP. Ethics approval for the study was obtained from the Bond University Human Research Ethics Committee (Protocol number RO1836) and gatekeeper approval was granted by the State Education Research Approval Process in New South Wales, Australia (Reference number: 2014075). Parental consent was obtained in writing to confirm participation of each student involved in the study.

Students from three mainstream Year 1 classes enrolled at two public primary schools in the northern region of New South Wales, Australia, were recruited from May to July, 2014 to participate in the study. Following gatekeeper approval from the principals at both schools, information sheets and consent forms were circulated to the parents of children across the three Year 1 classes. All students enrolled in the three Year 1 classes (n = 64) were invited and eligible to participate in the study provided their parents consented and the students themselves indicated assent. The study sample consisted of 55 Year 1 children (n = 25 boys; n = 30 girls; mean age = 6.77 ± 0.40 years, range = 5.42–7.75 years).

Statistical analyses for this study were conducted using the Statistical Package for the Social Sciences (Version 26) [22]. Descriptive statistics including mean, standard deviation (SD) and range were calculated for numerical variables including age, motor proficiency and academic performance. Frequencies (%) were calculated for categorical variables including sex and ethnicity. Normality of distributions and equality of variances were assessed to determine whether assumptions for parametric statistics were met. When variables did not meet the assumptions for using parametric tests, non-parametric statistical tests were employed. When assessing sex as a potential covariate for subsequent analyses examining relationships between motor proficiency and academic performance, independent samples t-tests were performed to determine any significant differences in mean age, academic scores and motor proficiency scores between girls and boys within the participant sample. Similarly, one way analyses of variance (ANOVA) with Bonferroni post hoc tests (using an overall alpha level of .05) were performed to determine any significant differences between the three classes and three ethnic groups in academic performance scores. Pearson’s and Spearman’s correlation analyses were performed to examine relationships between both age and academic performance scores. Pearson’s correlation analyses were subsequently used to examine the relationships between motor proficiency and academic performance in mathematics and reading for the total sample. Where assumptions of normality were not met, Spearman’s correlation analyses were performed on ranked data to analyse relationships between motor proficiency and academic performance variables. To account for multiple analyses of associations, Bonferroni corrections were applied to tests of significance in the correlational analyses. To describe the strength of correlation (r) between motor proficiency and academic performance variables, the rating guide described by Evans [23] was used as follows; r = 0.00–0.19 (very weak), r = 0.20–0.39 (weak), r = 0.40–0.59 (moderate), r = 0.60–0.79 (strong), and r = 0.80–1.0 (very strong). Finally, hierarchical multiple linear regression analyses were performed to determine how much variance in mathematics or reading composite scores (dependent variables) could be explained by each of the motor proficiency scale scores (independent variables) while controlling for covariates, enabling the relative predictive contribution of the components of motor proficiency to be assessed. Sensitivity analyses were conducted by repeating regression analyses with the removal of any outliers that were identified. To determine the effect size for the proportion of unique variance in academic performance explained by each predictor variable, Cohen’s f² was calculated. According to Cohen’s [24] conventions, an effect size of .02 can be considered small, .15 can be considered medium and .35 can be considered large. A significance level of 5% (α = 0.05) was applied to all statistical tests, with Bonferroni corrections when appropriate. A statistical power analysis using G*Power 3 [25] indicated that the correlation analyses would have an 80% power to detect a correlation between two variables of 0.39 (a weak correlation) or greater if the sample numbered at least 50 participants, assuming an alpha level of 0.05.

The Australian normative scores for the pseudoword decoding subtest, reading comprehension subtest and the reading composite of the WIAT-II were only available for participants aged 5 years, 8 months and older, resulting in this data being unavailable for one participant (age = 5.42 years) and leaving data available for these subtests from 54 of the 55 participants (Fig. 1). Means, SD and ranges of performance data for the BOT-2 and WIAT-II (both by total sample and sex) are presented in Table 2.

Overall the mean total motor composite standard score for the total sample was 51.56 ± 10.65 (range 22–79), which was considered ‘average’ motor proficiency. This was consistent with the mean of the normative sample [17], falling between ±1 SD for age and sex-specific norms (i.e. mean = 50, SD = 10, range 20–80). The mean mathematics composite standard score (94.87 ± 15.60, range 64–148) and mean reading composite standard score (97.96 ± 16.70, range 66–132) for the total sample were considered ‘average’ achievement. Each was slightly below but still within ±1 SD of the mean of the Australian normative sample (i.e. mean = 100, SD = 15, range 40–160) [19]. Finally, the mean total motor composite, mathematics composite and reading composite standard scores were categorised ‘average’ for both boys and girls (Table 2).

Overall, for the total sample, significant moderate positive correlations were found between mathematics composite scores and total motor composite (r = .466, p < .001) and fine manual control composite scores (r_s = .572, p < .001) (Table 3). Significant moderate positive correlations were evident between mathematics composite scores and the fine motor precision (r_s = .449, p = .001) and fine motor integration (r_s = .525, p < .001) subtests (Table 4). Significant moderate positive correlations were also found between the maths reasoning subtest and the fine motor precision (r_s = .449, p = .001), fine motor integration (r_s = .530, p < .001) and manual dexterity (r_s = .436, p = .001) subtests (Table 4). However, significant moderate positive correlations were only found for the numerical operations subtest with the fine motor integration subtest (r_s = .461, p < .001) (Table 4).

Significant moderate positive correlations were evident between the reading composite and fine manual control composite (r_s = .476, p = .001) (Table 3) and fine motor integration subtest (r_s = .470, p < .001) (Table 4). There were significant moderate positive correlations between the pseudoword decoding and word reading subtests and the fine manual control composite (r_s = .438, p = .001 and r_s = .521, p < .001 respectively) (Table 3) and fine motor integration subtest (r_s = .459, p < .001 and r_s = .512, p < .001 respectively) (Table 4).

Following consideration of a range of possible covariates, only participant age was included in subsequent regression analyses. Correlation analyses revealed a significant negative weak correlation between age (measured in years and months) and mathematics composite scores (r = −.327, p = .015) but a non-significant relationship between age and reading composite scores (r_s = −.197, p = .154). Although no such relationship was found for reading composite scores, given that age is known to be a key factor affecting scores on academic tests, it was included in subsequent regression analyses as a covariate wherever mathematics composite or reading composite scores were the dependent variables. No significant relationships between mathematics or reading composite scores and sex or school class or ICSEA or ethnicity were detected, and thus none of these were included as covariates in subsequent analyses.

To determine how much variance in mathematics performance could be explained by the components of motor proficiency beyond that accounted for by age, a hierarchical multiple regression analysis was conducted. Findings from the simple correlation analyses (Tables 3 and 4) were used to guide the motor proficiency variables that were entered into regression models, and these included fine motor integration and fine motor precision. Preliminary analyses found no violation of the assumptions of linearity, multicollinearity and homoscedasticity [26, 27]. However, a standardised residual greater than 3SD was found for one participant and was thus identified as an outlier and subsequently considered in a sensitivity analysis. Variables were entered into the model in the following steps: Age at step 1, fine motor precision at step 2 and fine motor integration at step 3 (Table 5). In combination, at step 3, the three predictor variables explained 34.7% of the variance in mathematics performance (R² = .347, adjusted R² = .309, ΔR² = .094, F (3, 51) = 9.03, p < .001). By Cohen’s [24] convention, this was considered a large combined effect (f² = 0.53). As can be seen in Table 5, in the final regression model, only fine motor integration (β = .430, p = .009) was a significant predictor of mathematics performance.

A sensitivity analysis was performed to examine whether results from this hierarchical multiple regression analysis were influenced by the outlier in the sample (Table 6). Following removal of the outlier, in combination, at step 3, the three predictor variables explained 39.2% of the variance in mathematics performance (R² = .392, adjusted R² = .355, ΔR² = .084, F (3, 50) = 10.73, p < .001). By Cohen’s [24] convention, this was considered a large combined effect (f² = 0.64). When the outlier was removed from the sample, fine motor integration (β = .407, p = .012) and age (β = −.244, p = .036) were both significant predictors of mathematics performance (Table 6).

To determine how much variance in reading performance could be explained by the components of motor proficiency beyond that accounted for by age, a separate hierarchical multiple regression analysis was conducted. The two motor proficiency subtests most significantly correlated with reading composite scores were fine motor precision and fine motor integration (Table 4). Variables were entered into the model in the following steps: Age at step 1, fine motor precision at step 2 and fine motor integration at step 3 (Table 5). In combination, at step 3, the three predictor variables explained 25.2% of the variance in reading performance (R² = .252, adjusted R² = .207, ΔR² = .102, F (3, 50) = 5.60, p = .002). By Cohen’s [24] convention, this was considered a medium combined effect (f² = 0.34). As can be seen in Table 5, in the final regression model, fine motor integration (β = .450, p = .012) was the only significant predictor of reading performance.

Findings from the present study are consistent with other cross-sectional research examining associations between fine motor proficiency, mathematics and reading skills in Year 1 children [28, 29]. For example, Pienaar et al. [28] found that visual motor integration skills were more strongly associated with mathematics and reading performance than total motor proficiency in a large sample of socio-economically disadvantaged first grade learners from South Africa. Significant medium to strong correlations between the maths reasoning subtest of the WIAT-II and the fine motor precision (r = .597, p < .001) and fine motor integration (r = .569, p < .001) subtests of the BOT-2 have also been reported in a small sample of Year 1 children in the UK [29]. However, similar to the findings in the present study, significant correlations were only found between the word reading and fine motor integration subtests (r = .377, p = .003), but not fine motor precision (r = .198, p = .129) in the sample of Year 1 children in the UK [29].

Analyses conducted in the present study revealed that after accounting for age and fine motor integration, fine motor precision was not a significant predictor of mathematics and reading composite scores in this sample of Year 1 children. This is consistent with other studies that have specifically evaluated relationships between academic performance in mathematics and reading and the individual components of fine motor skills, including fine motor integration (or visual motor integration), fine motor precision and manual dexterity (or fine motor manipulation / coordination) [30‐32]. For example, a longitudinal study by Kim et al. [32] found that fine motor coordination and visual motor integration were related to the mathematical ability of students in Kindergarten, however, only visual motor integration was related to mathematical skills in the same sample of students when they reached Year 1. The authors suggested that children’s mastery over fine motor coordination skills may explain why they were no longer related to children’s mathematical skills in Year 1 [32].

It was beyond the scope of this study to ascertain the underlying mechanisms that may explain the observed study findings. However, one potential explanation as to why fine motor integration may be more strongly related to mathematics and reading in Year 1 children than other fine motor skills (i.e. fine motor precision and manual dexterity) has been proposed in the literature, and relates to the notion of automaticity [32, 33]. Motor and cognitive processes (such as executive functions) may share similar neural pathways in the brain with researchers conducting functional neuroimaging studies demonstrating that when tasks are novel or complex, the cerebellum and pre-frontal cortex are both activated [34]. Motor tasks appear to become more automatic with practice leading to a reduction of activity in these two regions [34, 35].

Previous studies examining relationships between gross motor composite scores and mathematical skills in children in the early years of school (e.g. pre-kindergarten to Year 2) have reported significant very weak to moderate positive associations [16]. However, few studies have previously investigated relationships between the individual components of gross motor proficiency and academic performance in mathematics in Year 1 children, like the present study [16, 36]. Overall, the components of gross motor proficiency that were most strongly related to mathematics composite scores were running speed and agility and bilateral coordination, though these relationships did not reach statistical significance after adjusting for multiple comparisons. The lack of significant findings are thus in contrast to those reported in the systematic review by Macdonald et al. [16] who found a strong level of evidence to support significant very weak to weak positive associations between speed and agility and mathematical ability in studies conducted with slightly older children aged nine to 13 years.

Significant very weak to moderate positive correlations between gross motor composite scores and reading skills in children in the early years of school have also previously been reported [16]. In the present study, upper limb coordination appeared to be the component of gross motor proficiency most strongly related with reading composite scores, particularly pre-reading skills including word reading and pseudoword decoding, but these relationships did not reach statistical significance. Again, the lack of significant findings are in contrast with the systematic review by Macdonald et al. [16] who found evidence to support significant weak positive associations between upper limb coordination and reading ability, including in Kindergarten children [37], students in Year 5 [38] and adolescents [39, 40].