A twin study of infant attachment security at age 24 months was conducted on archival data for a sample of 99 MZ pairs and 108 DZ pairs from the Louisville Twin Study. MZ concordance for attachment was 62.6%, which was significantly greater than the DZ concordance of 44.4%. Concordances were transformed into polychoric correlations, and LISREL was used to conduct a quantitative genetic analysis of the data. Results indicated that 25% of the variability in attachment was attributable to genetic factors, and the remaining 75% was attributable to non-shared environmental effects. No evidence was found for a contribution from shared environmental influences to attachment security. Possible concerns about the validity of twin methodology are addressed and various interpretations of the results are presented.