Predictions of two models were compared. The models relate ovulation rate (OR) and prenatal survival (PS) to litter size (LS): the uterine capacity model (UCM), where maximum LS is limited by uterine capacity (UC), and the threshold model (TM) whereby PS is modelled as a binary threshold trait. Records were simulated according to both models using statistics from Trench Large White gilts. Both models were able to reproduce closely the observed curvilinear relationship between OR and LS, with LS reaching a plateau at high OR. Several genetic correlations (ρg) fulfilling the conditions h20R = 0·34 and h2LS = 0·12 (the residual maximum likelihood estimates in the population) were studied by means of stochastic computer simulation. The genetic correlation between OR and LS was very sensitive to changes in h2uc, whereas ρgLS,PS was always positive, and ρgOR,PS was always negative. The correlation between PS and UC was larger than 0·90, except for very small h2UC This suggests that genes affecting PS have a strong influence on UC and that PS can be a good indirect criterion to select for UC. Both models predicted that the advantage of an index combining OR and LS with respect to direct selection on LS diminishes in successive generations of selection and that the size of the experiment needed to detect significant differences is very large. Records were also simulated by halving the mean and variance of UC, so as to mimic unilateral hysterectomy-ovariectomy (UHO). If the UHO treatment results in halving UC, LS of UHO females should behave very much as half the UC of intact females.