Although there are significant intra-individual differences in self-reported diurnal preference, as measured by validated questionnaires, the relative contribution of exogenous and endogenous factors to self-reported diurnal preference largely remains to be investigated. The present study examined which items from the Horne-stberg (H) questionnaire of diurnal preference were better at predicting genotypes in the variable number tandem polymorphism (VNTR) in the coding region of the gene PER3. This polymorphism has previously been reported to associate with diurnal preference, sleep parameters, and cognitive performance markers following sleep deprivation. Participants (n=240, selected from a previously studied population) had completed the H questionnaire and provided a DNA sample, which was genotyped with regard to the PER3 VNTR. A multinomial logistic regression showed that four items significantly increased prediction accuracy between the two homozygotic genotypes, with homozygotes for the longer variant of the gene (PER35/5) associated with answers indicating a stronger morning preference than those chosen by homozygotes for the shorter variant (PER34/4). Only one item, the question of whether the respondent required an alarm clock, discriminated between all three genotypes. Moreover, when the items were divided into those with the strongest versus the weakest genetic association, there was a significant relationship between age and the questions not predicting genotype, but not between age and genotype-predictive questions. This may explain previous findings regarding age-related differences in self-reported diurnal preference. These findings could facilitate the future development of diurnal preference scales especially tailored to the study of specific biological parameters.