Abstract
The aim of the current study was to explore relative age’s influence on physical and motor tests among fourth grade children (9 to 10 years) from Germany. Data from 1218 children (49% female) who had performed the German Motor Ability Test (Bös et al., 2009) were analysed. The test battery, which was comprised of physical and motor tests, included 20 m sprint, balance backwards, jumping sideways, stand and reach, push-ups, sit-ups, standing broad jump, and six-minute run. Analyses of variance only revealed statistically significant effects for height, weight, and 20 m sprint time (p < .01) among boys, with relatively older boys performing better than relatively younger boys. For the girls, the only significant difference between quartiles was for height (p < .01), with the oldest quartiles being taller than the younger quartiles. These results may have implications for statistical vs. practical significance, sampling, and how youth are evaluated in physical education classes.
Introduction
Annual age grouping, a ubiquitous policy in sport and education, has been found to probabilistically advantage youth with birth dates at the beginning of a cohort selection year (see Bedard and Dhuey, 2006; Musch and Grondin, 2001). For example, in a system that utilises a selection date of 31 December to determine eligibility for school entry or sport participation (i.e., with an annual age group defined by birth dates from January to December), youth born in January will be approximately 11 months older than their peers born in December of the same selection year. This difference in age within an annual age group has been termed relative age (Wattie, Cobley and Baker, 2008). The influence of relative age on developmental outcomes is typically described as relative age effects (RAEs: Barnsley, Thompson and Barnsley, 1985).
The most notable manifestation of RAEs in sport has been the increased likelihood of relatively older youth being selected for sports teams (see Cobley, et al., 2009). For example, in one of the first studies on relative age in sport Barnsley et al. (1985) observed that approximately 40% of youth participating in the major Canadian amateur developmental ice hockey leagues (the Western and Ontario Hockey League) were born in the first three months (January-March) of the annual age group, while only 10% of participant were born in the last three months (October-December), compared to an expected 25%. These trends have been replicated in Canadian youth ice hockey (Sherar et al., 2007) and in youth soccer worldwide (Barnsley, Thompson and Legault, 1992; Helsen, Van Winckel and Williams, 2005), and have been found to persist from youth levels of participation into elite (e.g., professional) levels of sport (e.g., Wattie et al., 2007; for an exception see Schorer et al., 2009). Moreover, RAEs have been observed in tennis (Baxter-Jones, 1995), rugby (Till et al., 2010) and a number of other sports (see Cobley et al., 2009 for a review). Within the educational domain, researchers have found that relatively older youth are more likely to receive higher grades in physical education classes (Bell, Massey and Dexter, 1997; Cobley, Abraham and Baker, 2008).
The advantage that relatively older youth are more likely to experience may be due to differences in physical size and maturation. Indeed, there is convincing evidence from sport that coaches select athletes who are advanced in terms of physical maturation, and that this provides a probabilistic advantage to relatively older youth (Baker et al., 2010; Brewer, Balsom and Davis, 1995; Sherar et al., 2007). The advantage of advanced physical maturation has been hypothesised to result from the correlation between physical maturation and motor performance proficiency among youth (see Malina, Bouchard and Bar-Or, 2004; Wattie et al., 2008). Once selected to teams, particularly competitive teams, it has been further hypothesised that youth may then be exposed to better coaching, more practice and competition and/or perhaps more positive experiences (see Musch and Grondin, 2001). As such, initial advantages experienced as a result of relative age may propagate future advantages.
To date, few studies have investigated the influence of relative age on more general abilities and performance measures (for exceptions see McPhillips and Jordan-Black, 2009; Roberts et al., 2012). Recently, however, Roberts and colleagues (Roberts et al., 2012) explored the influence of relative age on an indicator of cardiorespiratory fitness in a sample of youth 9 to 12 years of age from England. Utilising a multistage 20 m shuttle run test as a measure of cardiorespiratory fitness, the authors observed that relatively older boys and girls (i.e., those born in the first three months of the selection year) performed better than their relatively younger peers (i.e., those born in the last three months of the selection year). These results are intriguing for several reasons. First, although prior research has emphasised differences in physical maturation as a primary mechanism of the RAE (Helsen et al., 2000; Sherar et al., 2007), relative age trends were still evident even when physical maturation was controlled for. Second, this study was among the first to examine the relationship between relative age and indicators of fitness among a general population of youth. Typically, examinations of relative age in youth populations have focused on sporting samples in general or high performance/elite samples in particular. Therefore, the findings from Roberts et al. (2012) raise interesting questions by challenging one of the purported aetiologies of RAEs (i.e., differences in physical maturation in sport and physical education), as well as challenging relative age’s influence on health-related outcomes.
In this study, our goal was to test the generalisability of the Roberts et al. (2012) findings in a sample of pupils from Germany. More specifically, we examine the influence of relative age on performance on a range of coordination and fitness measures. Such international comparisons are a useful means of exploring the validity and generalisability of theories, methodologies and phenomena (Marsh and Hau, 2003). Indeed, cross-cultural international comparisons have been an important component of establishing the ubiquity of RAEs in sport (Cobley et al., 2009).
Methods
Participants
Participants were recruited from primary school classes in mid-December 2010 and the end of January 2011 in Muenster, Germany. 1 The total sample consisted of N = 1218 participants, 592 of them females and 626 males. The mean age was 9.72 years (SD = 0.45 years). The mean height was 1.43 m (SD = 0.07) and the mean weight was 36.54 kg (SD = 7.50). In Germany, school classes are generally organised using the cut-off date of July 1st; however, there can be exceptions. For this analysis, only pupils who reached the age of nine (n = 346) or 10 years (n = 872) on the day of the test were analysed (i.e., children who were born between 1 July 1999 and 30 June 2000). Pupils were grouped by month of birth into quartiles related to the school cut-off date (Quartile 1: July–September, n = 352; Quartile 2: October–December, n = 328; Quartile 3: January–March, n = 251; Quartile 4: April–June, n = 287).
The standardised tests were conducted by specially trained research assistants who had participated in a one-day certified training programme that included instruction on the theory of the tests, test procedure and quality, and issues of objectivity, reliability and validity (German Motor Ability Test, DMT 6-18: Bös et al., 2009). Parents provided informed consent and participation was anonymous and voluntary throughout the test procedure.
Measures
In addition to having their height and weight measured, participants completed the DMT 6-18 (German Motor Ability Test, DMT 6-18: Bös et al., 2009), a motor ability test recommended by the German Association of Sport Sciences (www.sportwissenschaft.de) to test the general fitness of children between the ages of 6 and 18 (Bös et al., 2009). This test battery included eight tests covering coordination and fitness, and constituted the dependent variables for this study (test-retest reliability 0.52–0.94; factor validity shows a sufficient model fit chi2 (18) = 39.702, p = .01, SRMR = .045, RMSEA = .058 (0.033–0.082), CFI = .96, Bös et al 2009): 20 m sprint: pupils had to run 20 m as fast as possible with time measured in 10ths of a second (s) using a stopwatch (pupils began their run from a standing posture); balancing backwards: pupils had to walk backwards across three balancing beams with decreasing widths (6 cm, 4.5 cm, 3 cm), length 300 cm and height 5 cm. The number of completed footsteps was counted; jumping sideways: within 15 s pupils jumped as quickly as possible from side to side without stepping outside the set field size (50 x 100 cm). The number of jumps was counted; stand and reach: pupils stood on a bench and reached down to their feet with straight legs (“+” further than sole of foot, “0” sole of foot”, “–” above sole of foot in cm); push-ups: pupils completed as many modified push-ups as possible within a 40 s interval. The pupils started lying down with hands folded behind the back. The second position was the top position of the standard push up (i.e., with arms extended). This position was followed by one hand touching the top of the hand of the supporting arm (position 3) before returning to the starting position; sit-ups: pupils performed as many sit-ups as possible in 40 s; standing broad jump: pupils jumped as far as possible using a two legged jump, measured in cm; Six-minute run: on an indoor 54 m circular course, pupils ran as far as possible in six minutes, measured in km.
Test procedure
Pupils completed the test battery in groups of 10 to 12, throughout a normal school day. Eight trained testers carried out the tests. After warming up, pupils started with the 20 m sprint, followed by the balancing backwards, jumping sideways, stand and reach, push-ups, sit-ups, standing broad jump, and finally by the six-minute run. Boys and girls were examined separately, and rest was given when required.
Analyses
For this study, analyses were conducted in accordance with Roberts et al. (2012), and were differentiated between sexes. One factorial (quartiles) between subject analysis of covariance (ANCOVA) tested for differences between quartiles on the varying dependent measures while controlling for pupil body mass index (BMI). SPSS 20.0 and G*Power 3.1 were administered for the descriptive and inferential analyses (Faul et al., 2007). Effect sizes were calculated using Cohen’s f (Cohen, 1992), and the test power was calculated post hoc. Scheffé tests on the basis of ANOVAs were conducted for post-hoc group comparisons. The statistical significance criterion was p < .05.
Results
All data were explored, separately for males and females, for equality of variance (Levene p >.05 for all tests) and for normality. Normality could only be shown for the standing broad jump, side-to-side jump, and stand and reach (only for females). However, the remaining data only appeared non-normally distributed because the large sample size drove the normality tests to statistical significance (inspection of histograms and skewness and kurtosis values suggested the data were adequately normally distributed).
For the boys (Table 1) we found significant differences in height, weight, and 20 m sprint time between the groups. These variables demonstrated a pattern where the relatively oldest boys were significantly taller (post-hoc tests revealed the following significant results: Q1 > Q3 and Q4, and Q2 > Q4) and heavier (significant post-hoc test results: Q1 > Q3 and Q4) and had better sprint performance (significant post-hoc test results: Q1 and Q2 > Q4) than the relatively youngest (Q1 mean = 4.03 s; Q4 mean = 4.15 s), although accompanied by small effect sizes (Cohen’s f = 0.16–0.22). Additionally, the performance in the jumping sideways test reached significance but here the boys from the second quartile performed the best. There were no significant effects in any of the other variables. For the girls (Table 2), the only significant difference between quartiles was for height, with the oldest quartiles being taller than the rest (significant post-hoc test results: Q2 > Q3 and Q4).
Males: Mean (and SD) results of body composition and motor tests differentiated by birth quartiles (controlled for BMI, except for stand and reach p < .01).
BMI = body mass index; SD = standard deviation.
a post-hoc scheffé p < .05: Q1–Q3, Q1–Q4, Q2–Q4;
b post-hoc scheffé p < .05: Q1–Q3, Q1–Q4.
c post-hoc scheffé p < .05: Q1–Q4, Q2–Q4.
d post-hoc scheffé p < .05, No differences.
e when controlling for height as a covariate, p > .05.
Females: Mean (and SD) results of motor tests differentiated by birth quartiles (controlled for BMI, except for stand and reach p < .01).
BMI = body mass index; Q = Quartile; SD = standard deviation;
a post-hoc scheffé p < .05: Q2–Q3, Q2–Q4.
b controlling height as a covariate, height p > .05.
Discussion
Similar to Roberts et al. (2012), we found some influence of relative age on physical variables; however, with only two exceptions (20 m sprint performance and side-to-side jumping in boys), these were limited to height (in both boys and girls) and weight (in boys). Given the relationship between chronological age and height and weight (Malina, Bouchard and Bar-Or, 2004), these relationships are not surprising. The results for the sprint and jumping sideways may also be consistent with previous research. Votteler and Höner (2013) suggest that on tasks high in physiological demand, like sprinting and agility tasks, relatively older children may have advantages because they would have greater explosive power, even after controlling for height differences.
We were surprised, however, that there were no significant relative age effects in the other measures. It seems reasonable that RAEs should have been found in this sample given previous research, particularly given the results of Roberts et al. (2012) and Cobley et al. (2008). However, prior work on the RAE, even those focusing on sporting samples, has typically shown small to moderate effects (see Cobley et al., 2009). Our post-hoc calculation of the effect sizes from Roberts et al. indicates small effects (f = 0.062–0.118), similar to those in the current study. Increasing the heterogeneity of the sample (i.e., when using a general sample versus a sport sample) may reduce or ‘wash out’ the size of the RAE such that our sample may not have been large enough to identify a significant effect. A heterogeneous sample could theoretically include those who have been exposed to more instruction and practice because they have been selected to sports teams, and those that have not been selected for sports teams. This could confound relative age trends and increase the variability within a sample, to the point of decreasing statistical power. Importantly, the statistics concerning test power also do not allow us to rule out a null effect (typically test power should be 0.80 in order to confirm a null effect: Kline, 2005). Given the size of previous effects, we are left with the researcher’s dilemma regarding the value of statistically significant but practically insignificant results. Given a large enough dataset, it may be possible to find significant differences between quartiles but they may be so small as to have no value in real-world contexts.
A second issue relates to the inconsistency between the current results and previous research (Bell et al., 1997; Cobley et al., 2008), which showed significant relative age effects in the distribution of grades in physical education classes. On the face of it, it seems that tests of qualities that should predict performance in physical education courses should reflect the same trends as grades in those courses; however, this may speak to a mechanism underpinning this effect. Specifically, tests of motor performance such as those measured in the current study are objective measures of performance while grades in physical education classes can be subjective and affected by bias (either implicitly or explicitly) on the part of the evaluator. Alternatively, it is also possible that relatively older pupils may be perceived by their peers (and teachers) to be more competent, perhaps because of greater physical size, and as a result become more actively involved in physical education class games. More prominent roles or involvement in games may result in higher grades and/or the development of better sport-specific skills (also resulting in higher achievement). Additional research is needed to test these hypotheses.
One factor, which may need to be considered when exploring the relationship between relative age and performance on tests such as those in the current study, is the selection date discrepancy between education and sport systems. In Germany, selection dates differ for education and sport, and consequently a child’s relative age in education may not be the same as their relative age in sport. This could have confounded the results presented in the current study. Hypothetically, if a child was relatively older in sport, resulting in performance improvements on the battery of tests due to training/experience, but was coded as relatively younger in education (the referent in the current study), RAEs could have been reduced. Indeed, this may account for trends, albeit qualitative ones, where children with their birth date in the second or third quartile have better performance scores than the children born with birth dates in the first quartile. An interesting avenue for future research will be to explore the implications of intersections between sport-based and education-based relative ages.
While the current study has interesting implications for relative age research, there are some limitations that should be acknowledged. Firstly, although the research assistants were trained in data collection methods and theory, inter-rater reliability may have affected the accuracy of data that were originally collected. It would also have been potentially valuable if the current study had data on object-control motor skills (e.g., catching) in addition to the measures of coordination and fitness presented herein, as well as a broader age range of participants. And while this study had strengths, such as its large sample size and analyses of the potential mechanism of RAEs, whether or not these results are generalisable will only be ascertained with additional comparative research in different developmental contexts.
Collectively, these results and those of previous research in this area raise some questions for future work. First, does increasing the homogeneity of the sample (i.e., by performing the same tests on an athlete sample) increase the size of the RAE? Second, to what extent are the inconsistencies between our results and those of Roberts et al. (2012) the result of cultural differences either in the provision of physical education or physical activity practices in general? Finally, are decisions regarding grades in physical education classes more prone to evaluator bias than more “objective” tests such as those used in the current study and by Roberts et al.? Continued examination of RAEs in more representative samples may improve our understanding of this phenomenon.
Footnotes
Acknowledgements
We thank the city of Muenster and their Bureau of Sports Administration, in particular Mr Schirwitz and Mr Imsieke. We also thank the FoSS, Karlsruhe Institute of Technology, Germany, and in particular Klaus Boes, the chairman of FoSS.
Funding
The study is part of a large-scale project within the city of Muenster, Germany under the lead of Bernd Strauss and Maike Tietjens, funded by the Federal Ministry of Family, Children, Youth, Culture and Sport of Northrhine-Westfalia, Germany (see Ghanbari et al., 2012).