Academic performance can look surprisingly fixed when students are compared with their peers. A new longitudinal study of 16,946 Japanese students suggests that substantial movement up or down the academic rankings during early adolescence is uncommon, and that the behavioural changes associated with those shifts tend to happen alongside them rather than clearly forecasting them years in advance.
The study, published in Scientific Reports on 28 September 2026, followed students in grades 4 to 9 across two contrasting Japanese cities for three to four years. Most students remained broadly stable in their relative academic standing. Only about 9% rose or fell substantially. Those changing trajectories were difficult to distinguish from baseline lifestyle and learning questionnaires, even though changes in study investment, motivation-related responses and digital-media use became visible as academic standing itself changed.
Academic rank is not the same as learning
A crucial detail is that the researchers were not measuring whether children learned more mathematics or language over time in an absolute sense. Achievement scores were standardised each year within each city and grade. The trajectories therefore represent changes in relative standing among peers.
That distinction matters. A student can learn a great deal during a school year while remaining at roughly the same percentile because classmates are learning too. Conversely, moving upward in relative standing means improving faster than the comparison group, not merely gaining knowledge. The study is therefore best read as an investigation of academic mobility within cohorts rather than a measure of whether students were progressing through the curriculum.
Following nearly 17,000 students across two cities
Chihiro Hosoda and colleagues analysed census-style municipal education data collected by local Boards of Education. The final longitudinal sample included 16,946 students spanning grades 4 through 9. Approximate grade-derived baseline ages were 11.9 years in one cohort and 11.4 years in the other, and students were tracked for three or four years depending on the available municipal series.
The researchers combined annually standardised achievement tests with repeated lifestyle and learning questionnaires. The questionnaire information covered aspects of everyday routine, family support, classroom engagement, study behaviour, motivation-related responses and digital-media use. Because the two cities differed in context, analysing both also provided a useful test of whether broad patterns could be observed across contrasting municipal settings rather than within a single school system.
The records had been anonymised before researchers received them. The research team had no access to names, addresses or school identifiers, and the cities were reported only as City A and City B. Tohoku University’s ethics committee approved the secondary analysis and waived individual informed consent because the study used pre-existing anonymised administrative data.
Most academic standing was stable
The dominant result was stability. Most students retained broadly similar relative positions over time, while approximately 9% showed substantial upward or downward shifts. That makes the changing groups educationally important but statistically unusual.
Students with stable high standing also displayed recognisable behavioural patterns. Consistent daily routines, family support and active classroom engagement tended to accompany sustained high achievement. Baseline questionnaire responses could classify stable patterns comparatively well, although the authors stress that much of this apparent classification ability reflects the strong relationship between baseline behaviour and baseline achievement itself.
In other words, it is easier to describe the habits of students who are already performing strongly than it is to identify, years in advance, which individual students will make a major jump or decline in their relative academic position.
The students who changed were difficult to spot in advance
The changing trajectories produced the study’s most important caution. Students who later rose or fell substantially were only weakly classified from their Year 1 behavioural profiles. The behaviours that distinguished them became clearer concurrently, as their academic trajectories changed.
Changes in study investment were among those accompanying academic movement. Motivation-related questionnaire responses also shifted, as did patterns of digital-media use. These associations are meaningful because they show that academic mobility is embedded in a broader behavioural context. They are not, however, evidence that changing one of these behaviours will necessarily cause a student’s academic rank to rise or fall.
The prospective analysis reinforces that point. When the researchers asked whether behavioural information available through Year 3 improved prediction of Year 4 standing beyond prior achievement, the additional explained variance was only ΔR² = 0.002. That is a very small increment.
Prior achievement was therefore overwhelmingly more informative about later relative standing than the accumulated behavioural questionnaire data. The finding pushes against a tempting interpretation of school surveys as early-warning systems capable of reliably identifying future academic movers from lifestyle responses alone.
Why concurrent change is still useful
Weak long-range prediction does not make the behavioural findings irrelevant. It changes what they can reasonably be used for. If study investment, motivation and media habits shift at the same time as academic standing, they may be useful markers of an ongoing transition. Teachers and families may learn more from repeated observation of change than from treating a single baseline questionnaire as a durable forecast.
This also suggests a different way to think about educational support. A student who looks ordinary at baseline may still experience a substantial later change, while a student with a strong baseline profile may remain stable. Monitoring trajectories may therefore be more informative than assigning fixed risk labels early in adolescence.
The study also highlights why digital-media measures require careful interpretation. Media use changed alongside some achievement trajectories, but the observational design cannot establish whether media behaviour drove academic change, responded to it, or reflected other changes in family life, motivation, peer relationships or time allocation.
What the study does not show
The research is large and longitudinal, but it is not an intervention. None of the reported behavioural associations establishes causation. Consistent routines, family support and classroom engagement may contribute to achievement, may partly result from it, or may share underlying causes with it.
The annual standardisation of achievement scores also means that the results concern relative rank. They cannot tell us from these trajectories alone how much absolute knowledge students gained, whether curriculum mastery accelerated, or whether the whole cohort improved or declined.
The two-city design gives the study breadth within Japan but does not make the results automatically generalisable to every educational system. School structures, family expectations, assessment practices and patterns of digital-media use vary across countries. The municipal data were also collected for educational administration rather than designed from the outset as an experiment capable of isolating mechanisms.
Finally, the approximately 9% who changed substantially should not be interpreted as a universal rate of academic mobility. The percentage depends on the study’s trajectory definitions, standardisation procedure, observation window and populations.
The practical lesson is about change, not prediction
The strongest message from the study is not that behaviour is unimportant. It is that behaviour observed at one point in time may be a poor crystal ball for substantial later movement in academic standing.
Stable high achievement was associated with coherent routines and supportive learning environments, but the students whose academic positions changed most were not easily identifiable from their initial behavioural profiles. Their behavioural changes became visible as their academic trajectories unfolded.
For education systems increasingly interested in predictive analytics, that distinction is consequential. Large datasets can describe stable groups accurately without necessarily predicting rare changes well. A model that recognises who is already doing well is not the same as a model that can foresee who will improve or struggle next.
The next research step is therefore causal and prospective. Experimental or carefully designed intervention studies would be needed to determine whether changing specific routines, study investment, motivation or media habits can alter academic trajectories. Until then, these behaviours are best understood as observable companions of academic change rather than proven levers that determine it.
Source Information
Study: Shifting academic fates in early adolescence are marked by concurrent behavioural change rather than baseline profiles
Authors: Chihiro Hosoda, Kenchi Hosokawa, Takuto Matsuhashi, Nanase Takahashi, Yu Hayashizaki, Yutaka Matsuzaki and colleagues, including Ryuta Kawashima
Journal: Scientific Reports
Published: 28 September 2026
DOI: 10.1038/s41598-026-72860-w
Design: Longitudinal secondary analysis of anonymised municipal education records from 16,946 Japanese students followed for three to four years.
