Many yeast peroxisome biogenesis mutants have been isolated in which peroxisomes appear to be completely absent. Introduction of a wild-type copy of the defective gene causes the reappearance of peroxisomes, despite the fact that new peroxisomes are thought to form only from pre-existing peroxisomes. This apparent paradox has been explained for similar human mutant cell lines (from patients with Zellweger syndrome) by the discovery of peroxisomal membrane ghosts in the mutant cells (Santos, M. J., T. Imanaka, H. Shio, G. M. Small and P. B. Lazarow. 1988. Science 239, 1536-1538). Introduction of a wild-type gene is thought to restore to the ghosts the ability to import matrix proteins, and thus lead to the refilling of the peroxisomes. It is vitally important to our understanding of peroxisome biogenesis to determine whether the yeast mutants contain ghosts. We have solved this problem by introducing an epitope-tagged version of Pas3p, a peroxisome integral membrane protein (that is essential for peroxisome biogenesis). Nucleotides encoding a nine amino acid HA epitope were added to the PAS3 gene immediately before the stop codon. The tagged gene (PAS3HA) was inserted in the genome, replacing the wild-type gene at its normal locus. It was fully functional (the cells assembled peroxisomes normally and grew on oleic acid) but the expression level was too low to detect the protein with monoclonal antibody 12CA5. PAS3HA was expressed in greater quantity from an episomal plasmid with the CUP1 promoter. The gene product, Pas3pHA, was detected by immunogold labelling on the membranes of individual and clustered peroxisomes; the clusters appeared as large spots in immunofluorescence. PAS3HA was similarly expressed in peroxisome biogenesis mutants peb2 and peb4, which lack morphologically recognizable peroxisomes. Gold-labelled membranes were clearly visible in both mutants: in peb2 the labelled membrane vesicles were generally much smaller than those in peb4, which resembled normal peroxisomes in size.