The Caenorhabditis elegans genome contains 18 sequences related to mammalian core 2/I N-acetylglucosaminyltransferases. The six most closely related genes (gly-1 and gly-15 to gly-19) likely encode active enzymes, because are all transcribed and do not appear to be pseudogenes. Polypeptide divergence and the gene structures are both concordant with a common ancestor at the time of radiation from mammals that underwent three rounds of duplication and, most recently, a tandem duplication. Polypeptide alignments with mammalian homologues do not indicate whether the enzyme specificities are core 2, 4, or I-like or novel, but do clearly demonstrate the secondary structure characteristics of glycosyltransferases. The six homologues have essentially nonoverlapping expression patterns, unrelated by tissue type or cell lineage. The extent varies widely; gly-15 is expressed only in two gland cells, whereas gly-18 is broadly expressed in diverse cell types. gly-1, -15, -18 and -19 are expressed during adulthood; gly-16 and gly-17 appear to be restricted to embryonic or early larval stages. The parsimonious interpretation of the expression pattern and sequence data is that the catalytic activities are similar but with diverged promoters. Null alleles of three of the genes were generated without causing gross abnormality in homozygous animals. RNA-mediated interference experiments also failed to induce defects in the four genes tested. Nevertheless, the nematode has evolved six diverged core 2 GlcNAc-T-like genes, and we postulate that these arose in response to selection pressures to which C. elegans is not ordinarily subjected in the laboratory.