A methodology is developed for the establishment of a core collection based on phenotypic data. A case is worked out for the tulip cultivar collection of the Hortus Bulborum, consisting of approximately 1000 cultivars. The methodology is based on cluster analysis of phenetic characters, selection of cultivars from the resulting clusters according to a combination of the proportional and genetic diversity dependent strategies, optimisation of the selection, and validation by means of principal component analysis and by the diversity indices of Nei and Shannon & Weaver. A core collection with 104 cultivars (approx. 10%: CORE-1) did not give a sufficient representation of the existing diversity. A set of 152 cultivars (approx. 15%: CORE-2) showed a far better representation. The variation in resistance levels for Fusarium is almost completely represented in CORE-2. Although a set of 200 cultivars (approx. 20%: CORE-3) reached a nearly complete representation of the total diversity, a better representation of the resistance levels was not achieved. The newly presented methodology for defining a core collection appears to be a useful approach. The included steps for optimising and validating the choice of accessions makes it a reliable and broadly applicable method.