An intercomparison study on the determination of vitamin B-1, B-2 and B-6 was performed as part of the EU MAT project involving 16 laboratories. Each laboratory was requested to analyse three different food samples (lyophilized pig's liver, mixed vegetables and wholemeal flour, respectively) using their 'in-house' method as well as an 'optimal extraction protocol' and using a common batch of takadiastase enzyme. High-performance liquid chromatography (both normal-phase and reversed-phase HPLC), microbiological assay and chemical methods were used. For vitamins B-1 and B-2 the agreement between laboratories (using their 'in-house' methods) was generally good (vitamin B-1: CV(w) < 5%, CV(b) = 16-28%; vitamin B-2: CV(w) < 8%; CV(b) = 12-40%, respectively, where CV(w) is the within-laboratory coefficient of variation and CV(b) is the between-laboratory coefficient of variation) with no apparent differences between the 'in-house' and 'optimal' extraction protocols. For vitamin B-1 microbiological results tended to be higher (e.g. ca 20-50%) than HPLC results. Some further improvement could result from optimalization (i.e. lowering) of the sample-enzyme ratio. Enzyme efficiency needs to be checked with each new batch of enzyme. For vitamin B-6 results were much more at variance (CV(w) = 6-13%, CV(b) = 36-63%), especially with the pig's liver. Some discrepancy was found between laboratories using dilute mineralic acid for extraction without enzyme treatment and those using trichloroacetic acid (TCA) and/or included takadiastase treatment. It is concluded that for vitamin B-6 further standardization and method improvement is needed in order to obtain acceptable variation between laboratories. In particular, B-6 vitamer interconversion during extraction and peak interference needs further study.