A selection experiment with a selection and a control line maintained for eight generations was set up to study efficacy of selection for a short interval from weaning to estrus after weaning the first litter and to estimate genetic variation in the Dutch Landrace population. Intervals were recorded without truncation (i.e., intervals up to 234 d were observed). A mating scheme with 10 mating groups was used to avoid inbreeding. Distribution of intervals from weaning to estrus was described by a mixture of a normal and an exponential distribution. Between lines, both underlying distributions were similar, but the contribution to the total distribution was different. From generation 4 onward, intervals were significantly shorter in the selection line than in the control line. Variances within lines, generations, and mating groups varied to a large extent. Response per unit selection differential was calculated for original data and three sets of transformed data, both with and without correction for unequal subclass numbers. Standardized responses were more precise and linear, when data were log-transformed, and when corrected for unequal subclass numbers. The assumed best estimate of the realized heritability was .17. The heritability in the foundation population was estimated at .36 +/- .05, using an animal model including all genetic relationships in the data. It was concluded that genetic selection decreases the average interval from weaning to estrus by reducing the number of sows with a record in the exponential part of the distribution.