Apatite fission-track length data are used extensively for thermal history inversion. However, several studies have documented instances of poor reproducibility of length data. We address this problem by using c-axis projection to normalize track lengths for crystallographic angle in the extensive laboratory annealing data set acquired by Barbarand et al. (2003a, 2003b). A new simplification reduces the c-axis projection model from six to four fitted parameters. Normalizing for track angle using c-axis projection improves every aspect of length measurement reproducibility examined. It accelerates convergence of mean length in single analyses; increases consistency among replicate measurements by a single analyst; enhances consistency of measurements of the same mounts by different analysts; and improves the match between analyses conducted with and without Cf-irradiation. C-axis projection is also shown to enhance the thermal sensitivity of length data. Based on these results, we assert that c-axis projection is a good means of compensating for observer bias, although it does not overcome differences caused by experimental error.