An experiment was conducted to test the hypothesis that phenotypes differing in germination rate and the presence or absence of secondary dormancy at low temperature were not genetically different. Seed of oilseed rape was germinated at 4, 10 and 19 degrees C, where selections were made in the percentile ranges 1-10 (early), 45-55 (intermediate) and 91-100 (late). Secondary dormancy occurred only in the late selections at the two lower temperatures. Thermal weighting of curves of cumulative germination on time gave circumstantial evidence that early percentiles were similar at all three temperatures and that seeds with secondary dormancy came largely from later percentiles above the 50th. To test for genetic differentiation between phenotypes, 5'-anchored simple sequence repeat primers were used to generate DNA marker profiles of seedlings raised from seed from each category. Principal coordinate analysis, and more detailed comparisons using the most discriminating markers, confirmed that the early germinators at the three temperatures were not associated with different banding profiles, but seeds entering secondary dormancy, particularly at 10 degrees C, were genetically distinct from germinators at the same temperature. Secondary dormant seeds at low temperature appear to originate mainly from the late germinating seed at higher temperature. Effects of temperature history and the requirement for alternating temperatures to break secondary dormancy were quantified. The results confirm the existence of genetically discrete sub-populations differing in ecologically significant traits.