The gal3 mutation of the galactose operon of E. coli, first described by Morse, Lederberg, and Lederberg in 1956, exhibits several unique and unusual properties. This mutation originated spontaneously and also reverts spontaneously, but fails to respond to various chemical mutagens. It is extremely polar, but the polarity is not relieved by nonsense suppressors. The gal+ revertants are of at least two different kinds. One class of revertants is stable and exhibits inducible enzyme synthesis, whereas the other class is unstable and constitutive. The properties of gal3 can be explained by the assumption that this mutation was caused by the insertion of a DNA sequence in the operator-promoter region of the gal operon. This proposal was tested by the construction of a gal3 derivative of a lambdagal+ phage and comparison of the physical properties of the two phages. Density-gradient centrifugation studies indicated that the gal3 mutation was accompanied by a 0.0033 g/ml increase in the buoyant density of the phage. This increase in density corresponds to the addition of 1200 base pairs. This observation was confirmed by electron microscopy of lambdagal+/lambdagal3 heteroduplexes. A single-stranded addition loop, 1160 base pairs in length, was found at a fixed position. It was concluded that the gal3 mutation was caused by the insertion of a sequence similar to IS3 in length. These studies also permit an estimation of the physical size of the gal operon. It is found to be 3600 base pairs long, separated from the attlambda (B.P') site by approximately 14000 base pairs.