Size-selective mortality is common in fish populations and can operate either in a positive size-selective fashion by harvesting larger-than-average fish or be negatively size-selective by harvesting smaller-than-average fish. Through various mechanisms (like genetic correlations among behaviour and life-history traits or direct selection on behaviour co-varying with growth rate or size-at-maturation), size-selection can result in evolutionary changes in behavioural traits. Theory suggests that both positive and negative size-selection without additional selection on behaviour favours boldness, while evolution of shyness is possible if the largest fish are harvested. Here we examined the impact of size-selective mortality on collective boldness across ontogeny using three experimental lines of zebrafish (Danio rerio) generated through positive (large-harvested), negative (small-harvested) and random (control line) size-selective mortality for five generations and then relaxed selection for 10 generations to examine evolutionarily fixed outcomes. We measured collective risk-taking during feeding (boldness) under simulated aerial predation threat, and across four contexts in presence/absence of a cichlid. Boldness decreased across ontogeny under aerial predation threat, and the small-harvested line was consistently bolder than controls. The large and small-harvested lines showed higher behavioural plasticity as larvae and developed personality earlier compared to the controls. The large-harvested line showed increased variability and plasticity in boldness throughout ontogeny. In the presence of a live predator, fish did not differ in boldness in three contexts compared to the controls, but the large-harvested line showed reduced behavioural plasticity across contexts than controls. Our results confirmed theory by demonstrating that size-selective harvesting evolutionarily alters collective boldness and its variability and plasticity.