The lengthy 5' nontranslated region (5'NTR) of hepatitis A virus (HAV) forms a highly ordered secondary structure, which has been suggested to play an important role in controlling viral translation by allowing for translation initiation by internal ribosome entry. To test this hypothesis, synthetic bicistronic RNAs, with all or part of the HAV 5'NTR in the intercistronic space, were translated in rabbit reticulocyte lysates. In the presence of an upstream cistron designed to block ribosomal scanning, the HAV 5'NTR was capable of directing the internal initiation of translation, confirming the presence of an internal ribosome entry site (IRES). Analysis of various deletion mutants demonstrated that the 5' border of the IRES is located between nucleotides 151 and 257, while the 3' border extends to the 3' end of the 5'NTR, between nucleotide 695 and the first initiation codon at 735. Except for a segment between bases 638 and 694, deletion of stem-loop structures between bases 151 and the 3' end of the 5'NTR inhibited or abolished translation. The addition of a 5' cap structure (m7GpppN) to monocistronic or bicistronic transcripts decreased the translation of a reporter gene downstream of the HAV 5'NTR but enhanced translation of the upstream cistron in bicistronic transcripts. This finding indicates that a 5' cap structure is inhibitory to HAV IRES-directed translation initiation and that the cap structure and the HAV IRES probably compete for the same limiting translation factors. The efficiency with which monocistronic constructs containing the HAV 5'NTR directed translation in reticulocyte lysates was compared with results for monocistronic constructs containing the IRES of the more rapidly growing encephalomyocarditis virus (EMCV). These results indicated that the HAV 5'NTR was more than 25-fold less active than the EMCV IRES in producing translation product. HAV 5'NTR-directed translation was inhibited by the presence of a one-fifth molar quantity of RNA containing the EMCV IRES, while a fivefold molar excess of the HAV 5'NTR did not inhibit EMCV IRES-directed translation. The relatively weak activity of the HAV IRES may thus be due to a reduced affinity for cellular translation factors which are present in limiting quantities in rabbit reticulocyte lysate.