Sensory processing differences measured by self- or parent-report co-segregate with quantitative autistic traits and have potential endophenotypic properties. It is not known to what extent this reflects generalized sensory dysfunction versus more specific associations involving individual senses or autistic trait domains. We combined Bayesian variable selection with dominance analysis to obtain a more nuanced understanding of modality-specific associations. We recruited two independent samples of adults to complete the Broad Autism Phenotype Questionnaire and the Glasgow Sensory Questionnaire. For each domain of autistic traits (social interaction, communication, cognitive rigidity), we performed stochastic search variable selection using Glasgow Sensory Questionnaire modality subscales as predictors while controlling for uncertainty in other variables. Dominance analysis was applied to the reduced models to evaluate the relative importance of predictors. Only auditory scores reliably predicted all three autistic traits when other modalities were accounted for. The proprioceptive scale, which included motor and interoceptive deficits, predicted communicative autistic traits more than other trait domains. The tactile scale appeared most specific for social autistic traits. Although the findings must be interpreted in light of the limitations of the questionnaires, the study suggests that auditory differences may be more likely than differences in other senses to be a robust sensory endophenotype relevant to autism.