The objective of this research was to study the amount and pattern of phenotypic diversity among 179 Sinapis alba accessions maintained in germplasm collections at Plant Gene Resources of Canada (PGRC) and the Saskatoon Research Centre of Agriculture and Agri-Food Canada (SRC-AAFC). Accessions were evaluated in five field trials at Saskatoon from 1994 to 1998. Observations were recorded on number of days to flower and to mature, plant height, 1000-seed weight, oil and protein content and selected fatty acids and glucosinolates. Analysis of variance and mean comparisons were used to characterize variation in the germplasm. There was significant variation among accessions for all traits except some minor fatty acids and glucosinolates. Principal component analysis indicated that five or six principal components provided a good summary of the data, accounting for 75–80% of the variation. In all trials, the first principal component axis separated accessions predominantly on the basis of erucic acid (C22:1) and oleic acid (C18:1), with associated C22 and C18 fatty acids. The relative importance of agronomic, morphological and other seed quality traits varied among the trials, but they were always less important than C22:1 and C18:1. Cluster analysis generated 10–13 groups of accessions in each trial except in 1997 (five clusters). Distinct clusters were identified that possessed high or low values for C22:1, C18:1, oil and protein content, maturity, plant height and seed weight. Seed colour was not used as a classification variable; however, brown-seeded accessions were grouped into one distinct cluster due to a significantly higher level of C22:1 in these accessions. This study demonstrates that the S. alba accessions maintained at PGRC and SRC-AAFC are a source of genetic diversity for breeding both condiment (high glucosinolate and C22:1 content) and vegetable (low glucosinolate and C22:1 content) oilseed yellow mustard and for conducting genetic studies. Key words: Sinapis alba, genetic diversity, cluster analysis, principal component analysis