Seed shelf-life or longevity is difficult to predict or to measure on a practical time scale. Predictive models suggest that water has the same effect on ageing rate for all seed lots within a species and that initial seed quality is the dominating factor explaining within-species variation. These assumptions are used in ‘accelerated ageing’ or ‘controlled deterioration’ (AA/CD) tests, which are commonly used to predict seed longevity in commercial and research applications. In this study, we describe within-species variation of longevity for seeds of Secale cereale and S. strictum (cultivated rye and its wild progenitor) under typical dry storage conditions of a genebank, and show that initial seed quality is an important, but not sole, factor explaining measured longevity. We also test the correlation of seed longevity, measured under humid and dry conditions, using 50 cultivars of rye, wheat (Triticum aestivum) and the intergeneric cross triticale, to assess how well AA/CD tests predict seed shelf-life during dry storage. Known differences in longevity between wheat and rye were confirmed at all water contents, and triticale seeds demonstrated intermediate behaviour. Longevity measured for humid and dry conditions were weakly correlated when comparisons included all grain types and were not correlated in within-grain-type comparisons. Response to moisture varied among cultivars. These findings do not support assumptions made in seed ageing models that use AA/CD tests. Our results suggest that more traits are involved in the expression of seed longevity than those typically measured in studies of initial seed vigour.