Antisense reagents have the potential to modify gene expression by interacting with DNA or mRNA to down-regulate transcription or translation. There have been a number of successful demonstrations of antisense activity in vivo. However, a number of problems must be solved before the method's full potential can be realized. One problem is the need for the antisense agent to form a duplex with the target molecule. We have found that most regions of mRNAs are not open to duplex formation with oligonucleotides because the bases needed for Watson-Crick base pairing are involved in intramolecular pairing. Using arrays of oligonucleotides that are complementary to extensive regions of the mRNA target, we are able to find those antisense oligonucleotides which bind optimally. There is good correspondence between the ability of an oligonucleotide to bind to its target and its activity as an antisense agent in in vivo and in vitro tests. To understand more fully the rules governing the process of duplex formation between a native RNA and complementary oligonucleotides, we have studied the interactions between tRNAphe and a complete set of complementary dodecanucleotides. Only four of the set of 65 oligonucleotides interact strongly. The four corresponding regions in the tRNA share structural features. However, other regions with similar features do not form a duplex. It is clear that ab initio prediction of patterns of interaction require much greater knowledge of the process of duplex formation than is presently available.