Published guidelines adopted in many countries recommend that women whose family history of breast cancer places them at a risk>or=1.7 times that of the age-matched general population, should be considered for inclusion in special surveillance programmes. However validation of risk assessment models has been called for as a matter of urgency. The databases of the four Scottish Familial Breast Cancer clinics and the Scottish Cancer Registry have been searched to identify breast cancers occurring among 1,125 women aged 40-56, with family histories placing them below the "moderate" level of genetic risk. The observed incidence over 6 years was compared with age-specific data for the Scottish population. Our findings confirm that when there are two affected relatives (one first degree) the relative risk (RR) exceeds 1.7 regardless of their ages at diagnosis. When only one (first degree) relative was affected at any age from 40 to 55, the RR does not reach 1.7 if that relative was a mother but exceeds it if the relative was a sister. The probable explanation is that sisters are more likely than mother/daughter pairs to share homozygosity for a risk allele. Surveillance programmes might therefore accommodate sisters of women affected before age 55. Evidence that "low penetrance" alleles contributing to breast cancer risk may be recessive should be taken into account in strategies for identifying them.