The RNA world hypothesis proposes that nucleic acids were once responsible for both information storage and chemical catalysis, before the advent of coded protein synthesis. However, it is difficult to imagine how nucleic acid polymers first appeared, as the abiotic chemical formation of long nucleic acid polymers from mononucleotides or short oligonucleotides remains elusive, and barriers to achieving this goal are substantial. One specific obstacle to abiotic nucleic acid polymerization is strand cyclization. Chemically activated short oligonucleotides cyclize efficiently, which severely impairs polymer growth. We show that intercalation, which stabilizes and rigidifies nucleic acid duplexes, almost totally eliminates strand cyclization, allowing for chemical ligation of tetranucleotides into duplex polymers of up to 100 base pairs in length. In contrast, when these reactions are performed in the absence of intercalators, almost exclusively cyclic tetra- and octanucleotides are produced. Intercalator-free polymerization is not observed, even at tetranucleotide concentrations > 10,000-fold greater than those at which intercalators enable polymerization. We also demonstrate that intercalation-mediated polymerization is most favored if the size of the intercalator matches that of the base pair; intercalators that bind to Watson-Crick base pairs promote the polymerization of oligonucleotides that form these base pairs. Additionally, we demonstrate that intercalation-mediated polymerization is possible with an alternative, non-Watson-Crick-paired duplex that selectively binds a complementary intercalator. These results support the hypothesis that intercalators (acting as 'molecular midwives') could have facilitated the polymerization of the first nucleic acids and possibly helped select the first base pairs, even if only trace amounts of suitable oligomers were available.