Piglet mortality is a major welfare and economic problem in the pig industry. Despite the use of farrowing crates, piglet crushing remains a major contributor to pre-weaning piglet mortality, which is typically around 12%. Our aims in this study were to quantify variability between sows and consistency across parities in crushing mortality, and to examine the effect of the environment on variability. In our first study, we compared the variability in crushing mortality in 122 primiparous sows (gilts) that farrowed in crates (71) or open pens (51). Certain sows crushed more or fewer piglets than expected by chance. Crushing was more frequent and more variable in pens compared to crates, indicating that crates may mask differences between sows. In our second study, we recorded piglet mortality for 125 sows, which farrowed in crates over several (4–9) parities. After adjusting for litter size, litter weight and parity effects, consistent individual differences between sows were evident. The repeatability of crushing was estimated at 0.14, with estimates of 0.18 and 0.05 for stillborns and total liveborn mortality, respectively. Although these repeatabilities are relatively low, there was a high degree of phenotypic variance (eg sows crushed between 0 and 30.8% of their piglets). Given that sows show some consistency in piglet mortality over parities, this could be used to inform culling decisions. Additionally, if differences in piglet crushing between sows have a genetic component, a breeding programme might reduce mortality from crushing. Because crates restrict maternal behaviour, genetic selection in this system may have relaxed selection for good maternal behaviour. Selection for reduced piglet mortality, and thus improved maternal abilities, could remove a major obstacle to the wider adoption of less restrictive farrowing systems, with positive welfare consequences for the sow and piglets.