We determined whether a classical conditioning paradigm may be used to condition immunologic responses in normal human subjects receiving an optimal immunostimulating dose of recombinant human interferon-gamma (rhIFN-gamma). We conducted a placebo-controlled, double-blind study of 31 normal volunteers in order to determine whether an initially immune-neutral stimulus, oral propylene glycol (PG), could eventually elicit an immune response as a consequence of its being paired with a known immunostimulatory dose and schedule of rhIFN-gamma. Subjects were randomly assigned to one of three groups: (A) rhIFN-gamma injections paired with PG; (B) normal saline injections paired with PG; (C) rhIFN-gamma injections alone. During the 4-week study, subjects received progressively fewer injections so that, by the final week of the study, no injections were given and groups A and B received only PG. The principal outcome measures were serum concentrations of quinolinic acid (QUIN) and neopterin, two nonspecific but sensitive markers of immune activation, and expression of Fc receptors (CD64) on peripheral blood mononuclear cells. RhIFN-gamma injections produced significant and predictable alterations in each of the measured immune parameters. No group B subject made an immune response. Mean serum QUIN levels were significantly higher at the end of week three for subjects in the experimental condition (group A) than for subjects receiving rhIFN-gamma alone (group C) despite receiving identical doses of rhIFN-gamma. Similarly, the predicted decay in mean serum neopterin levels from the end of week 1 to the end of week 2 was seen in group C but not in group A. The exposure of group A to PG blunted the decline of CD64 expression in week four. The data suggest that the pairing of an unconditioned stimulus (rhIFN-gamma) and a conditioned stimulus (PG) permits the conditioned stimulus alone to prolong a cytokine-induced response in normal humans.