Genetic evaluation of litter size in pigs is usually performed considering litter size from different parities as repeated measures of the same trait, and fitting an animal model with repeatability, which assumes homogeneous variances and high genetic correlation between parities. To test these hypotheses, a total of 14,774 records from the first seven parities for number of piglets born alive from 3,909 Large White sows were analyzed using a multitrait animal model (MTM) and a random regression model (RRM). The parameters estimated using the MTM evidenced an increasing tendency of heritability (h2) from first (0,07 ±0,01) to sixth (0,28 ±0,04) parities, being its value lower (0,13 ±0,02) in the seventh parity. Values of heritabilities and permanent environmental ratios estimated using the RRM were similar and fluctuate around 0.10 for the different parities. Using both models, the genetic correlations estimated between successive parities were high (?G ? 0,80) and decreased (?G ? 0,30 – 0,40) as the interval between parities increased. The application of MTM and RRM approaches to genetic evaluation of pig litter size would be useful to improve its low persistency in the later parities.