In two studies we applied a new stochastic model for the measurement of latent states and traits to salivary cortisol in healthy subjects. Data were analyzed by Linear Structural Relations (LISREL). In the first study, 48 students collected saliva samples on two occasions at 1300h with an interval of six weeks between occasions. Salivary cortisol levels were found to be mainly influenced by situational and/or interactional effects, which determined approximately 75% of the total variance of the hormone data, while only approximately 21% were determined by the latent trait. Measurement unreliability was found to be approximately 4%. In the second study, 54 young mothers collected saliva at 0800h, 1500h, and 2000h for three consecutive days, starting on the second day after delivery of a healthy baby. Early morning salivary cortisol showed high consistency across the three days. Approximately 60% of the total variance in the early morning cortisol samples was determined by the latent trait. For the 1500h and 2000h samples, however, the corresponding proportions of variances amounted only to 33% and 28%, respectively. These results appear to reflect the absence of a clear-cut circadian rhythm in some of the women on the second day after delivery.