Although the lengthy 5' nontranslated regions (5'NTRs) of other picornaviral RNAs form highly ordered structures with important functions in viral translation, little is known about the 5'NTR of hepatitis A virus (HAV). We determined the nearly complete 5'NTR nucleotide sequences of two genetically divergent HAV strains (PA21 and CF53) and included these data in a comparative phylogenetic analysis of the HAV 5'NTR. We identified covariant nucleotide substitutions predictive of conserved secondary structures and used this information to develop a model of the 5'NTR secondary structure, which was further refined by thermodynamic predictions and nuclease digestion experiments. According to this model, the 5'NTR comprises six major structural domains. Domains I and II (bases 1 to 95) contain a 5'-terminal hairpin and two stem-loops followed by a single-stranded and highly variable pyrimidine-rich tract (bases 96 to 154). The remainder of the 5'NTR (domains III to VI, bases 155 to 734) contains several complex stem-loops, one of which may form a pseudoknot, and terminates in a highly conserved region containing an oligopyrimidine tract preceding the putative start codon by 13 bases. To determine which structural elements might function as an internal ribosome entry site, RNA transcripts representing the HAV 5'NTR with progressive 5' deletions were translated in rabbit reticulocyte lysates. The translation product was truncated, unprocessed P1 polyprotein. Removal of the 5'-terminal 354 bases of the 5'NTR had little effect on translation. However, deletion to base 447 slightly decreased translation, while deletion to base 533 almost completely abolished it. These data indicate that sequences 3' of base 355 play an important role in the translation mechanism utilized by genomic-length HAV RNA. Significantly, this region shares several conserved structural features with the internal ribosome entry site element of murine encephalomyocarditis virus.