Changes in personality are typically modeled linearly or curvilinearly. It is a simplifying – yet untested – assumption that this model form accurately depicts all person-level trajectories in a sample. Given the complexity of personality development, it seems unlikely that imposing a single model form across all individuals is appropriate. Although typical models can estimate individual trajectories that deviate from the average (i.e., random effects), they do not explicitly test whether people differ in the forms (linear, quadratic, etc.) of their trajectories. This heterogeneity is valuable to uncover, though, as it may imply that different processes are driving change. The current study uses data from four longitudinal datasets (N = 26,469) to empirically test the degree to which people vary in their best-fitting model forms for Big Five personality development. Across datasets, there was substantial heterogeneity in best-fitting model forms. Moreover, the type of form someone had was directly associated with their net and total amount of change across time, and these changes were substantially misquantified when the wrong form was used. Variables such as gender, age, trait levels, and number of waves were also associated with individuals’ types of forms. Lastly, comparisons of best-fitting forms from individual- and sample-level models indicated that consequential discrepancies arise from different levels of analysis (i.e., individual versus sample) and alternative modeling choices (e.g., choice of time metric). Our findings highlight the importance of these individual differences for understanding change processes and suggest that a flexible, person-level approach to understanding personality development is necessary.