Subcortical morphometric findings in autism are heterogeneous, and diagnosis-level comparisons may obscure associations involving symptom dimensions and age. We examined whether Autism Diagnostic Interview-Revised reciprocal social interaction, verbal communication, and restricted or repetitive behavior scores were associated with seven bilateral subcortical volumes and whether age moderated these associations. This cross-sectional secondary analysis included 355 autistic participants aged 7-49 years from 16 Autism Brain Imaging Data Exchange I sites. Public precomputed FreeSurfer v6 volumes were analyzed using log-transformed measures and total brain volume. Models adjusted for age, sex, total brain volume, and site, with HC3 standard errors. Benjamini-Hochberg correction was applied separately to 21 age-independent domain-region tests and 21 domain-by-age interaction tests. None of the age-independent associations survived false-discovery-rate correction. One age interaction survived correction: the reciprocal-social-interaction association with bilateral caudate volume became more negative with increasing age (beta = -0.002933 per 1-SD score per year, 95% CI [-0.004804, -0.001061], p = 0.0021285, q = 0.044699). The coefficient remained negative and nominally significant in all 16 leave-one-site-out analyses. However, the interaction was not supported in participants aged 18 years or younger (N = 287; beta = 0.001548, p = 0.600141), and the estimate weakened when the upper age range was restricted. These findings provide qualified evidence of cross-sectional age heterogeneity in the reciprocal-social-interaction-caudate association, rather than an age-independent subcortical symptom signature. The pediatric null result, site-age imbalance, threshold-proximal corrected result, and absence of external replication require cautious interpretation and longitudinal confirmation.