AbstractThe average axial length for eyes grouped according to refractive error must vary as a function of refractive error. In fact, this change in the average axial length can account almost totally for the change in refractive error. This finding is incompatible with the usual interpretation of correlation ametropia. When the categorization of eyes is based on axial length rather than on refractive error, then a more statistically meaningful division of eyes into component and correlation categories is achieved.