Some preparations of beta-galactosidase from strains of Escherichia coli carrying point mutations in their lacZ genes did not precipitate with antibody as effectively as wild-type enzyme, but did not appear to be chain-terminating mutations as judged by polarity measurements and suppression. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis of crude extracts of induced Lac+ strains revealed that the monomer of beta-galactosidase ran as a band uncontaminated by other cellular proteins. This method was used to identify missense mutations in the alpha and beta portions of the lacZ gene. Six of 13 mutations investigated were judged to be missense by this criterion. Measurement of the degree of polarity, the ability to complement a nonsense mutation at the operator-distal extremity of the gene (omega-complementation), and suppressibility by 12 nonsense suppressors allowed the assignment of six other mutations as either number or ochre. The protein figments produced by these six nonsense mutations appeared to be degraded in vivo. One mutation that could not be classified was either a missense mutation whose protein product was degraded or a very leak nonsense mutation. Two lacZ alleles were suppressed by the ochre suppressors supM and supN, although they were missense by other criteria. The ability of supM to suppress both nonsense and missense mutations can be explained if it is derived from a tyrosine transfer ribonucleic acid with a modified base in the first position of the anticodon. The mutations assigned to the missense class were not suppressed by the missense suppressors supH, supQ, glyV, glyU, or glyT. Our results suggest that the criteria used in the past to distinguish between nonsense and missense mutations may not be conclusive even when used together.