The bacteriophage SP6 promoter consists of highly conserved 22 base pairs from -17 to +5. In order to determine the base pairs that are essential to the SP6 promoter function, a library of all possible single base-pair mutations of SP6 promoter was constructed by random mutagenesis. Relative transcription efficiencies of 66 SP6 promoter variants compared with the consensus SP6 promoter(wild type) were determined in vivo and in vitro; in vitro transcription was performed under 6 mM $MgCl_2$ and a linear template condition. Some mutations such as -10T → A, -3A → T, and +5C → G, showed higher transcription efficiency than wild type in vitro. Especially -3A → T mutant was more active than wild type both in vivo and in vitro. Mutations at position -11, from -9 through -5, and +1 seriously reduced promoter activity in vivo and in vitro, while mutations at -17, -16, -1, +4, and +5 had little effect on SP6 promoter function. Some mutations showed substantially different effects in vivo and in vitro; the -13T → C, -2T → A, -1A → C, and +3A → C mutations almost abolished in vitro efficiency but affected little in vivo, while -3A → T mutation almost abolished in vivo efficiency without affecting much in vitro. At several positions, SP6 RNA polymerase displayed a clear hierarchy of base-pair preference. The correlation of in vitro SP6 promoter activity with the presence and/or absence of specific functional groups in the region from -17 to -4 suggest that the followings are necessary for the interaction of SP6 promoter with SP6 RNA polymerase: (1) a hydrogen bond donor at the 4-amino group of the C and a hydrogen bond acceptor at the 6-keto group at the N7 position of the G at -11, -9, -7, and +5; (2) a hydrogen bond acceptor at the 4-keto group of the T or 6-amino group of the A at -9 and -4; (3) the T:A base pairing at -15 and -14.