Previous studies of the insulin gene–utilized restriction-fragment–length polymorphisms as markers for potential mutations at this locus. This indirect type of analysis could not define the number of variants that might exist within the structural portions and regulatory regions of the gene in non-insulin-dependent diabetes mellitus (NIDDM) patients. New technology has allowed us to examine insulin genes at the single nucleotide level from 100 American black NIDDM patients. Genomic DNA from patients was amplified by the polymerase chain reaction with primers flanking four regions of the gene: 1) the proximal promoter from positions −182 to 42 (including most of exon 1); 2) exon 1 from 14 to 259, which included the rest of exon 1 and all of the 1st intron; 3) exon 2 from 216 to 452; and 4) exon 3 from 1188 to 1433. One of the primers in each reaction was ³²P–end labeled and the resulting products denatured into single strands and electrophoresed on nondenaturing sequencing gels such that mobility was a function of composition and size (single-strand conformational polymorphism or SSCP). Under these conditions, single-base changes in fragments up to 245 nucleotides were detected. Analysis of the proximal promoter region revealed several SSCP patterns in individuals. Direct genomic sequencing of DNA representative of these patterns showed the presence of a common C to G change at position −56 and a C deletion at position −90 in three patients. For exon 1, there were eight SSCP patterns, which on sequencing were found to represent combinations of three polymorphisms within the first intron, a 4-bp insertion (TTGC) at position 46 (close to the 5′-splice donor site of the 1st intron), a G-to-A change at 80, and a previously described T-to-A change at 216. Exon 2 analysis revealed a newly described G-to-A change at 274, resulting in a silent polymorphism within the prepeptide of the gene. For exon 3, two previously described allelic variants in the 3′-untranslated region(C to T at 1367 and A to C at 1380) were found to be polymorphic in this population. In summary, analysis of 200 insulin genes from NIDDM patients revealed the full extent of variability at this locus. Not a single mutation in the coding portions of the gene was found. Four previously described allelic variants were confirmed and four new variants observed, with only one in the coding region. Haplotypes were described for the first time, and seven new alleles identified. Two of four newly described polymorphisms might have biological consequences. The C deletion at −90 in the proximal promoter might alter transcription of the gene. The four-nucleotide insertion at 46 alters the consensus sequence of the 5′-splice donor site of the first intron, which might alter proinsulin mRNA processing. These hypotheses can be tested.