Development of a vaccine for acquired immunodeficiency syndrome (AIDS) has proven difficult, and so alternative approaches such as idiotypic manipulation have been suggested. As applied to AIDS, this approach could involve immunizing with an anti-CD4 antibody resembling gp120, to induce anti-idiotypic antibodies which would bind to gp120. The CD4 binding site on gp120 is conserved, and so, such an immune response should protect against all variants. Induction of anti-human immunodeficiency virus (HIV) immunity has been reported using anti-Leu3a, and this result has led to testing in humans. Negative results obtained by others have been attributed to differences in immunization protocols. Because of the importance of this question, we reinvestigated the potential of anti-Leu3a to induce anti-HIV antibodies, compared with control immunizations with OKT4A (another anti-CD4 antibody) and the irrelevant Ig MOPC-21. Responses to anti-Leu3a showed induction of high-titer anti-idiotypic activity, and included combining-site-related activity. Yet sera showed no binding to gp160 above controls and no detectable neutralizing activity in a sensitive HIV plaque assay, so the anti-idiotypes induced were not internal images of CD4. We conclude that the pronounced anti-HIV responses reported with anti-Leu3a cannot be generalized, and thus that anti-Leu3a does not offer promise as an HIV vaccine. However, these results do not negate the promise of the idiotypic approach, and a vaccine for AIDS based on idiotype manipulation remains a possibility.