Human immunodeficiency virus type 1 (HIV-1) recombination occurs during reverse transcription when parts of the two co-packaged RNAs are used as templates for DNA synthesis. It was previously hypothesized that HIV-1 Gag polyproteins preferentially encapsidate the RNA from which they were translated (cis-packaging hypothesis). This hypothesis implies that mutants encoding Gag that cannot efficiently package viral RNA are selected against at two levels: these mutants do not generate infectious virus, and these mutants are not efficiently rescued by the wild-type virus because the mutant RNAs are packaged at much lower levels than are those of the wild-type genome. Therefore, genetic information encoded by gag mutants can be rapidly lost in the viral population. To test this prediction of the cis-packaging hypothesis, we examined several gag mutants by measuring the efficiencies of the mutant RNAs in being packaged in trans in the presence of wild-type virus and determining the rates of recombination between gag mutants and wild-type viruses. We observed that the viral RNAs from the nucleocapsid zinc finger or the capsid truncation mutant were packaged efficiently in trans, and these mutant viruses also frequently recombined with the wild-type viruses. In contrast, viral RNAs from mutants containing a 6-nucleotide substitution encompassing the gag AUG were not efficiently encapsidated, resulting in a low rate of recombination between the mutants and wild-type viruses. Further analyses revealed that other, more subtle mutations changing the gag AUG and abolishing Gag translation did not interfere with efficient encapsidation of the mutant RNA. Our results indicated that neither the gag AUG sequence nor Gag translation is essential for viral RNA encapsidation, and Gag can package both wild-type and gag mutant RNAs with similar efficiencies. Therefore, we propose that HIV-1 RNA encapsidation occurs mainly in trans, and most gag mutants can be rescued by wild-type virus; therefore, they are unlikely to face the aforementioned double-negative selection.