Since conscriptions, church registers or other traditional sources of historical demography are lacking, research on family structure in late medieval Hungarian towns has to utilize indirect evidence. This article explores the potential of last wills from Eperjes (present-day Presov) in this respect, examining 123 wills from this free royal town from the period between 1446 and 1526. By bequeathing goods to spouses, children, grandchildren, brothers, sisters, and other collateral relatives, or mentioning them in other contexts, the testators could outline the circle of kin that was important for them. However, in Eperjes - contrary to two free royal towns in Western Hungary (Sopron and Pozsony/Pressburg/Bratislava) analysed by the author before - this circle did not comprise all living members of a family. With some simplification, the majority of last wills from Eperjes can be termed 'one-generation wills', in which most goods and chattels are bequeathed to relatives who belonged to the testator's own generation: spouses (eventually with under-age chidren), brothers, sisters and cousins. Parents and grandchildren are hardly ever mentioned, which can be explained partly with a presumably high rate of mortality, and partly with the neolocality and independent households of grown-up children. More significant is the strikingly low number of children occuring in the wills. In 106 families that can be analysed on the basis of the testaments, only 74 children are mentioned altogether. If this had been the real rate of reproduction, the population of the town would have been threatened with extinction within three generations. The relatively high number of brothers and sisters (in most cases two or three), whose inclusion in the wills was made obligatory in local customary law, also contradicts the assumption of too small families. Therefore in Eperjes one cannot estimate the number of surviving offspring on the basis of last wills. The explanation of the rare occurrence of children must be sought not in demographic features, but in the system of inheritance. Through the parallel study of last wills and the only surviving medieval protocol of the town, one can see several examples when property was divided well before both parents' death, so there was little left to be divided on the death-bed. The children's share was handed over when they came of age or married, or when one of the parents remarried. The Eperjes case-study calls for a closer examination of the interdependence of inheritance customs and family structure. One can also notice that even towns within the same country and legal framework show considerable regional variation, which should make one cautious towards too broad generalizations.