Results of this study showed that self-reported PA tracked moderately between the ages of 12 and 13. Previous reviews have indicated low to moderate tracking of PA in childhood and adolescence [8, 9], although comparison between studies is complicated among other things by the different lengths of follow-up. Adjusted coefficients were found to be clearly higher than unadjusted ones , as was found in this study. However, adjustment was done in very few previous studies. Tracking of PA in children and adolescents has been well discussed in the literature, and will therefore not be repeated here. The documented decline in PA among girls is also in accordance with findings in children in this age group . The decline noted in this study was nevertheless very small. A Norwegian study with a large population-based sample has shown that PA levels were higher in the Spring compared to the Fall season among 9 year-olds . What appears to be an increase in PA between BL and T1 is thus most probably related to these seasonal variations.
Despite the long known tracking of PA behaviour, investigation of the stability of factors associated with it has been lacking. PA is influenced by different factors including genetic predisposition as well as psychological and social-environmental factors [4, 8, 10–12, 37]. From the standpoint of public health promotion, particular attention to factors amenable to change should be given. This study shows that potential psychological and social-environmental correlates of PA appear to be well established at a young age, with moderate to moderately-high tracking. Higher tracking coefficients in both genders were found for self-efficacy and parental support, both robust correlates of PA. Low to moderate tracking of psychological and social-environmental correlates of PA was reported in a study of young adults and over a 7 year period . However, no study of tracking of correlates of PA in this age group could be identified.
Small but significant changes in enjoyment of PA and teachers’ support in both genders and in friends’ support and perceived environmental opportunities in females in a direction unfavourable to PA were detected. The changes documented in this study are rather small and probably not of much practical significance in the studied period. Nevertheless, they might be indicative of a trend which is getting established. Unless addressed, these changes might increase further and adversely impact PA and might get exacerbated when the children go through school transition at the end of 7th grade, as indicated in a previous study .
In light of the moderate tracking of potential correlates of PA together with the unfavorable changes noted, early health promotion efforts to make cognitions, beliefs and perceptions related to PA more positive would seem important. Such efforts can include the enhancement of social support for PA, as well as getting children to begin exercising and providing them with feedbacks, which is believed to be critical in the enhancement of self-efficacy for PA . Tracking is a positive phenomenon for children with high levels of PA and with positive cognitions, beliefs and perceptions related to PA. Health promotion efforts should thus also be aimed at such children in order to help them maintain these levels.
Finally, no effect of pubertal timing on PA at BL and 20 months later was noted. Sherar et al. 2010 summarized the results of the few existing studies looking at the impact of biological maturity on PA . They found inconsistent results for both genders. Associations, when noted, were generally low, as was also found in this study. They concluded that differences may stem from different factors, including the methods used to assess both maturity and PA. Evidence of validity and reliability of the method used to assess pubertal stage in this study has been documented . However, the measure gives a rough assessment of pubertal status and has weaknesses compared to more direct measures, such as the possibility of social desirability bias related to the self-report. PA was also self-reported. However, the differences between studies might also reflect contextual differences since pubertal maturation has important societal and cultural dimensions , which might in turn determine whether or not it will influence PA. The absence of an association between the timing of pubertal maturation and PA beliefs and perceptions at age 11 among females, and the few weak associations among males should also be interpreted in light of this. Whereas early maturation in girls is hypothesized to have many potential negative psychosocial impacts, including less enjoyment of PA as well as less social support from parents and peers , findings of this study do not support those hypotheses, at least with regards to how these are perceived by children. Corresponding literature for boys is limited; but early maturation in boys is believed to have advantages, because their body changes conform to the socio-cultural ideal . In addition, early maturation and advanced physical development might lead to a better athletic ability  and thus early maturers can more easily be positively reinforced for their sport and physical activity achievements. These facts might at least in part explain the slightly higher enjoyment of PA and perceived support for PA from parents and teachers in early maturing boys in this study. Worth emphasizing is that results might have been different if comparison between extreme maturity groups was conducted, which was impossible in this study because of low numbers of children in these groups, due to the age at which pubertal status was assessed. A progress to higher maturity stages when pubertal changes become more visible is likely to be related to more personal and social influences on the children.
Strengths and weaknesses
The study has both weaknesses and strengths. Self-reported measures in children present problems of validity and reliability. Nevertheless, good test-retest reliability was obtained for the measures used. Good and stable estimates of internal consistency reliability of the instruments used to measure the correlates were also found across all the time points. In their review of studies of tracking of PA, Telama et al. 2009 indicated that many studies did not account for low reliability of measures used and suggested the use of reliability information in order to correct correlations for measurement errors , which was done in this study.
One methodological consideration is that the children might give more accurate answers after being previously exposed to the same instrument, and also with increasing age, although it is difficult to estimate the magnitude of such effects. This can influence the tracking coefficients, and the changes noted.
The question assessing PA was general and intended to provide a measure of total PA. It was mainly used to look at trends over time in this study, and the use of self-reported measures of PA for such purposes appears justified . In addition, studies have shown that simple, overall self-report PA questions correlated significantly with other measures, and have relatively satisfactory validity . However, such a measure is unlikely to capture sporadic activity, as an accelerometer would have. As there are known seasonal variations in PA among children in Norway , stability and change in PA were assessed only between T1 and T2 which represented similar seasons. Tracking is influenced by the duration of follow-up. The follow-up period in this study is relatively short. Nevertheless, the transition into adolescence is a period during which rapid changes are expected to occur. It is also a critical period from the perspective of physical education and lifestyle development . The study is also limited to a single geographical area making generalizability limited.
The study provides new insights into the tracking and change of potential correlates of PA in young children, and is, to our knowledge, the only study addressing this issue in this age group. The sample size for the study was large and the rate of retention was also high.