Data from Norwegian spring-spawning herring Clupea harengus, were used empirically to assess the performance of a regression method aimed at detecting the time of the onset of maturation from growth trajectories. At the level of the whole dataset, the piecewise linear-regression method was accurate and showed only a minor bias (−0·17 years) relative to age at maturation visually read from scales. The method, however, was relatively imprecise and provided an estimate of age at maturation equal to the one read from the scale in less than half of the cases (47·6%). Moreover, bias was strongly dependent on age at maturation: the age at maturation of early maturing fish was often overestimated, whereas the opposite was true for late-maturing fish. Accuracy and precision of the regression method relative to visual readings also depended on the growth type (determined by the nursery area of young C. harengus) and cohort but not on sex. Modifying the original regression method enabled marginally improving the precision of the approach but a strong maturation age-dependent bias persisted. The results with C. harengus suggested that age-at-maturation estimates from the piecewise linear regression method should be treated with caution.