A Bayesian threshold animal model was applied to evaluate the prevalence over 2 farrowings and genetic background of overall leg conformation score and the presence or absence of 6 specific leg defects (abnormal hoof growth, splay footed, plantigradism, straight pasterns, sickle-hocked legs, and the presence of swelling or injuries) in purebred Landrace and Large White sows. Data sets contained phenotypic records from 2,477 and 1,550 Landrace and Large White females, respectively, at the end of the growing period. Leg conformation data from first and second farrowings were available for 223 and 191 Landrace sows and 213 and 193 Large White sows, respectively. Overall leg conformation deteriorated with age, with statistically relevant differences between females at the end of the growing period, first farrowing (FF), and second farrowing (SF). In a similar way, the prevalence of the 6 specific leg defects increased between the end of the growing period and FF (with the exception of straight pasterns in the Landrace population). Differences between FF and second farrowing were statistically relevant for hoof growth (highest posterior density regions at 95% did not overlap), plantigradism, sickle-hocked legs, and overall leg conformation score in Landrace and for sickle-hocked leg and overall leg conformation score in Large White. The statistical relevance of the genetic background was tested through the Bayes factor (BF) between the model with the additive genetic component and the model with 0 heritability (nonheritable). Heritability (h(2)) was discarded (BF < 1) for sickle-hocked leg in both breeds, whereas decisive evidence (BF > 100) of genetic background was obtained for overall leg conformation score in Landrace and Large White sows (h(2) = 0.27 and 0.38, respectively), hoof growth in both breeds (h(2) = 0.22 and 0.26, respectively), and plantigradism (h(2) = 0.34) and the presence of swelling or injuries in Landrace (h(2) = 0.27). Note that a BF > 100 implies that the model with infinitesimal genetic effects was more than 100 times more suitable than the model without genetic effects, a conclusive estimate within the Bayesian framework. The remaining traits (splay footed and straight pasterns) registered BF values ranging from 11.6 to 35.1 and h(2) values ranging from 0.09 to 0.19. These results indicated a moderate genetic determinism for leg conformation in Landrace and Large White sows.