Improved winter hardiness is an important breeding objective in the forage grass Lolium perenne. This is a complex trait with several components, including the ability to survive and grow at low temperature, to acclimate to cold, tolerate wind, snow cover and ice encasement. Marker-assisted selection has the potential to increase the efficiency of breeding for improved cold tolerance. Here we describe a genecological approach to identifying molecular markers that are associated with adaptation to low winter temperatures. AFLP was used to assess the genetic diversity in 29 wild populations of ryegrass (Lolium perenne) representing a pan-European temperature cline in terms of their geographical origin. A further 18 populations from a temperature cline in Bulgaria were also analysed. In addition, two varieties and five populations representing parents of mapping families currently in use at IGER were included in the analysis. Principal coordinate (PCoA) and cluster analyses of the molecular marker data showed that the Bulgarian altitude cline populations could be distinguished clearly from the other populations. Two regression analyses were carried out; one to identify AFLP markers that correlated in frequency with low mean January temperature of the geographical origin of the population, and another to identify AFLP markers correlating in frequency with the cold tolerance phenotype of the populations, as determined by LT50 values in freezing tests. In the first analysis six AFLP markers showed significant type II trends with mean January temperature, and in the second analysis 28 bands had a significant univariate relationship with the LT50 value of the accessions. In steps 2 and 3 of the stepwise analysis a further 4 and 5 bands, respectively, improved the fit significantly. The results of the two types of regression analysis are discussed in relation to ecogeography and cold tolerance phenotype of the populations.