This paper reviews the validity of the assumptions and predictions of genet
ic models used for the analysis of quantitative traits. When data on phenot
ypes and pedigrees but not individual genes are available, the traditional,
additive, infinitesimal model makes satisfactory predictions of short-term
response to selection despite the incorrectness of its assumptions. Improv
ements can made in the model such as the inclusion of non-additive variance
components or major genes but gains in the accuracy of prediction have sel
dom been demonstrated. When data are available on individual genes or marke
rs, two models are in common use - a two allele and an infinite allele mode
l. The two models differ in a number of respects such as assumptions about
segregation variance, estimation of fixed or random effects and number of p
arameters to be estimated when multiple traits are used. There is little in
formation on the accuracy with which either model predicts the performance
of future animals. (C) 2001 Elsevier Science B.V. All rights reserved.