Abstract Introduction PALB2 is emerging as a high-penetrance breast cancer predisposition gene in the order of BRCA1 and BRCA2 . However, large studies that have evaluated the full gene rather than just the most common variants in both cases and controls are required before all truncating variants can be included in familial breast cancer variant testing. Methods In this study we analyse almost 2000 breast cancer cases sourced from individuals referred to familial cancer clinics, thus representing typical cases presenting in clinical practice. These cases were compared to a similar number of population-based cancer-free controls. Results We identified a significant excess of truncating variants in cases (1.3 %) versus controls (0.2 %), including six novel variants ( p  = 0.0001; odds ratio (OR) 6.58, 95 % confidence interval (CI) 2.3–18.9). Three of the four control individuals carrying truncating variants had at least one relative with breast cancer. There was no excess of missense variants in cases overall, but the common c.1676A > G variant (rs152451) was significantly enriched in cases and may represent a low-penetrance polymorphism ( p  = 0.002; OR 1.24 (95 % CI 1.09–1.47). Conclusions Our findings support truncating variants in PALB2 as high-penetrance breast cancer susceptibility alleles, and suggest that a common missense variant may also lead to a low level of increased breast cancer risk.