Background: A number of methods to investigate growth have been reported in the literature, including hierarchical linear modelling (HLM), latent growth modelling (LGM), and multidimensional scaling applied to longitudinal profile analysis (LPAMS). Aims: This study aimed at modelling the mathematics growth of students over a span of 6 years from Grade 3 to Grade 9. Sample: The sample comprised secondary longitudinal data collected in three waves from n = 866 Hong Kong students when they were in Grade 3, Grade 6, and Grade 9. Method: Mathematics achievement was measured thrice on a vertical scale linked with anchor items. Linear and nonlinear latent growth models were used to assess students' growth. Gender differences were also examined. Results: A nonlinear latent growth curve with a decelerated rate had a good fit to the data. Initial achievement and growth rate were negatively correlated. No gender difference was found. Conclusion: Mathematics growth from Grade 6 to Grade 9 was slower than that from Grade 3 to Grade 6. Students with lower initial achievement improved at a faster rate than those who started at a higher level. Gender did not affect growth rate.