Fifteen endoscopic gastric biopsies (GBx) from 12 patients with small lymphocytic infiltrates morphologically raising a differential of indeterminate lymphocytic infiltrate versus mucosa-associated lymphoid tissue (MALT) lymphoma were analyzed genotypically after frozen-section identification of the abnormal lymphocytic infiltrate. Frozen-section immunoperoxidase immunophenotyping was equivocal in each case. All patients had abdominal pain attributable to superficial gastric ulceration, most often antral, without peripheral lymphadenopathy or hepatosplenomegaly. Rearrangement of the immunoglobulin heavy-chain gene (JH-R, seven patients) or kappa light-chain gene (JK-R, eight patients), was found in eight GBx from eight (seven stage IAE; one stage IBE) of 12 patients, establishing, in conjunction with the histologic features, a diagnosis of low-grade B-cell lymphoma. This diagnosis had not been tenable on multiple prior GBx, ranging from one to five per patient, over intervals of 1 month to 6.5 years (median 4.5 months). The T-cell receptor beta-chain gene retained germline configuration in all cases. Insufficient DNA for molecular studies was extracted from the GBx of two patients, one with JK-R (JH-G) on subsequent GBx and one without further GBx. One patient had two GBx, each demonstrating a single additional band in HindIII digests hybridized with the JH probe. No rearrangements were detected in either the BamHI or the EcoRI digests. Uninvolved tissue from this patient was not available for the exclusion of restriction fragment length polymorphism. Three GBx (two patients) showed germline JH genes (JH-G). One had a partial gastrectomy (histology: MALT lymphoma) in 1981 followed by GBx in 1983 (histologically benign) and in 1990 (JH-G), and negative esophagogastroduodenoscopy (EGD) in 1991 without biopsy. The other patient (two GBx with JH-G) had multiple subsequent abnormal EGD, but no biopsies since December 18, 1990. Adequate DNA for gene rearrangement studies can be extracted from GBx samples weighing as little as 20 mg. The two samples with insufficient DNA weighed 1 and 16 mg, respectively. Practically speaking, the remainder of a frozen block from a single GBx is adequate, thus allowing the screening of multiple endoscopic GBx by sequential frozen sections to determine which one contains the most extensive lymphocytic infiltrate for molecular study. Consistent results are obtained on samples weighing 40 to 60 mg. This method is a suitable alternative to kappa/lambda frozen-section immunoperoxidase immunostaining, which can be uninterpretable on endoscopic biopsies or small biopsies from other sites.