Abstract A questionnaire sent in 1990 to over 2000 pairs of 18 to 26-year-old twins included shortened versions of the Interpersonal Sensitivity Measure (IPSM), Parental Bonding Instrument (PBI) and Intimate Bond Measure (IBM). To evaluate the validity of these short scales, and investigate their stability over time, full versions of the questionnaires were mailed 1 year later, in 1991, to a subsample of 100 twin pairs (25 monozygotic females, 25 monozygotic males, 25 dizygotic females, 25 dizygotic males). After follow-up, replies were received from 51 male and 70 female individuals (60% response overall). From the full length questionnaires used in 1991 we were able to generate test scores for both the shortened and full scale versions, whereas from the 1990 study only short scale scores were available. Correlations between the shortened and full length scales obtained in 1991 were high ( r=0.64–0.99). Correlations between test scores on the shortened versions of the instruments in 1990 and 1991 were moderate ( r=0.47–0.55) for the IPSM, and slightly higher for the shortened PBI scales (0.55–0.66) and the shortened IBM care subscale ( r=0.58). Correlations between the 1990 shortened scales with their full length 1991 counterparts were slightly lower for the IPSM ( r=0.31–0.50), comparable or slightly higher for the PBI ( r=0.61–0.70), and the same for the IBM ( r=0.58). Internal reliabilities of the shortened test versions were slightly reduced but satisfactory in most cases when compared with the full instruments: Cronbach's alpha full length 1991 α=0.81–0.96; shortened 1991 α=0.46–0.96; shortened 1990 α=0.48–0.96. We conclude that some information is lost using the shortened versions of these instruments, and there is also some change over time. However, the relationship between the tests, using either the full length versions or the shortened counterparts, remained constant suggesting that the shortened versions can be used in epidemiological studies.