A patulin-negative mutant (J1) of Penicillium urticae (N.R.R.L. 2159A) was known to accumulate about 100mg per litre quantities of the 5,6-epoxygentisyl quinone, (-)-phyllostine and another metabolite (UIII). Both were derived from acetate and hence were polyketides. Purified UIII (m.p. 53 degrees C, [alpha](32) (D)+206 degrees , lambda(methanol) (max.) 240nm; epsilon 3806 litre.mol(-1).cm(-1)) was characterized as a partially reduced derivative of (-)-phyllostine and was found to be a diastereoisomer of the known phytotoxin, (+)-epoxydon. Hence its designation as (+)-iso- or epi-epoxydon. From (1)H n.m.r. and c.d. data the stereochemistry of the epoxide ring in (+)-isoepoxydon was determined to be identical with that in (+)-epoxydon (i.e. R,R) but the configuration of the secondary alcohol at C-4 was S rather than R as in (+)-epoxydon. Isoepoxydon (compound UIII) is therefore (4S,5R,6R)-5,6-epoxy-4-hydroxy-2-hydroxymethylcyclohex-2-en-1-one. The boat conformation in which the C-4 hydroxy group is axial is preferred. In the range of 1mm to 5mm, the antibiotic activity of (+)-isoepoxydon against Bacillus subtilis sp. was 56% of that obtained with patulin. Over a period of 1 to 3h, [(14)C]isoepoxydon was efficiently converted into patulin by a shake culture of the parent strain of P. urticae. The precursor relationship of isoepoxydon to patulin was confirmed by feeding unlabelled isoepoxydon (1mm) to a washed-cell suspension of a mutant (J2) in which, over a period of 3 to 5h, a better than 60% conversion into patulin was attained. The enzymic relationship between isoepoxydon and phyllostine and their positions in the late portion of the patulin biosynthetic pathway are discussed.