Objective To investigate the effect of familial risk for dementia on verbal learning by comparing cognitively healthy twins who had demented co-twins with cognitively healthy twins who had cognitively healthy co-twins.

Methods 4367 twins aged ≥65 years including 1375 twin pairs (533 monozygotic (MZ), 823 dizygotic (DZ) and 19 unknown zygosity pairs) from a population-based Finnish Twin Cohort participated in a cross-sectional telephone assessment for dementia and in a single free recall trial of a 10-item word list.

Results Cognitively healthy twins with demented co-twins (n=101 pairs) recalled less words than cognitively healthy twins with cognitively healthy co-twins (n=770 pairs) after adjusting for age, sex and education, B=− 0.44, 95%  CI (−0.73 to −0.14), p=0.003. The effect size was similar in MZ (n=31) twins (3.88 vs 4.29 words, B=−0.41, 95%  CI (−0.96 to 0.13)) and DZ (n=66) twins (3.70 vs 4.17 words, B=−0.47, 95%  CI (−0.84 to −0.10)). The heritability estimate of immediate recall (IR) was 0.37, 95%  CI (0.21 to 0.43).

Conclusions The results demonstrate that familial risk for dementia is reflected in the IR performance of cognitively healthy older persons. The finding of poorer IR performance in non-affected siblings compared with the general population, together with substantial heritability of IR, supports IR as a useful endophenotype for molecular genetic studies of dementia.