In a differential respiratory conditioning paradigm with normal Ss two odors (fresh smelling niaouli and bad smelling ammonia) were used as conditioned stimuli (CS+ or CS-) and 7.4% CO2-enriched air was used as the unconditioned stimulus (US). Three CS+ and three CS- trials were run during acquisition, followed on the next day by the same number of CS+ and CS- only trials. Respiratory frequency, minute ventilation, end-tidal fractional concentration of CO2 and subjective complaints were measured throughout the experiment. While during acquisition all measures were affected, the conditioning effects included only respiratory frequency and subjective complaints. A selective association effect appeared in that the conditioning effects were confined to ammonia as CS+: respiratory frequency increased and more somatic complaints were presented when compared to the CS- condition. The conditioning effect on complaints was not confined to complaints of general arousal, but included respiratory complaints as well. Correlational analyses showed that increases in complaints as caused by the conditioning procedure were predicted by changes in somatic variables, but not by individual differences in Negative Affectivity.