Thirty six local populations of maize collected in Cantabria, on the Northern part of Spain, were evaluated during two years of trials. With the use of 13 plant, ear and kernet traits showing high heritability and constancy estimates, numerical taxonomy was applied, resulting in ten different taxons. Description of the groups is reported. Discriminant analysis showed that the traits related with kernel characteristics, ear measurements and flowering date were the most important for classification. The first two canonical discriminant functions explained 72.4% of the variation.Peer reviewed