A clinical study was designed to determine if there was a predictable relationship between saliva and plasma codeine concentrations. Drug-free volunteers (n = 17) were administered a 30-mg dose of liquid codeine phosphate. Plasma and saliva specimens were collected at various times for 24 h after administration. Plasma and saliva were analyzed for codeine and morphine by positive-ion chemical ionization gas chromatography-mass spectrometry. The plasma codeine concentrations peaked between 30 min and 2 h after administration and ranged from 19 to 74 ng/mL with a mean of 46 ng/mL. Despite decontamination procedures, elevated saliva codeine concentrations were detected at the early collection times because of contamination of the oral cavity from the liquid codeine. Codeine concentrations in the 15 min specimens ranged from 690 ng/mL to over 15,000 ng/mL. After the initial 2-h period, the mean codeine saliva concentrations declined at a rate similar to that observed in the plasma, but remained 3 to 4 times greater than the plasma concentrations. During the elimination phase, half-life estimates for codeine in plasma and saliva were found to be equivalent, 2.6 and 2.9 h, respectively. However, the area under the curve (AUC) estimate for codeine in saliva was 13 times greater than the plasma AUC. Contamination of the saliva resulted in elevated saliva/plasma (S/P) concentration ratios for the first 1 to 2 h after drug administration. Consequently, S/P ratios in specimens collected in the first 15 to 30 min ranged from 75 to 2580. However, after the absorption phase, a significant correlation between saliva and plasma concentrations was observed (r = 0.809, p < 0.05) and mean S/P ratios remained constant (mean = 3.7). Although small changes in saliva pH were predicted to produce profound changes in the S/P ratios for codeine, this was not observed in the current study. Therefore, saliva codeine concentrations could be used to estimate plasma concentrations through the use of the S/P ratio once the oral contamination has been eliminated. However, these estimates should be made cautiously. One must ensure that oral contamination is not a factor. Also, as with blood-drug concentrations, considerable intersubject variability was observed.