Saliva and serum samples were collected from eight healthy volunteers every two hours during a 26‐hour period. Melatonin concentrations were measured by radioimmunoassay after chloroform extraction using radioiodinated melatonin as a tracer. Five of the subjects had high serum melatonin levels at night (peak levels higher than 75 pg/ml); in three subjects the highest serum melatonin concentration was 20‐40 pg/ml. All subjects had low levels (<10 pg/ml) during the day. The correlations between salivary and serum levels were calculated. The regression line y =0.33×+ 3.7 pg/ml, r= 0.95, P <0.001, was obtained for all detectable value pairs (n= 73). The regression and correlation coefficients were almost equal for the peak values of melatonin and during the rising and descending phases of the secretion patterns. However, no significant correlation was found between low daytime salivary and serum concentrations when calculated separately. In the five high‐secretors the melatonin levels in saliva reflected reliably the changes in serum, but in the three low‐secretors the correlation between salivary and serum melatonin was not significant. The proportion of melatonin found in saliva decreased with increasing serum melatonin levels. Circadian rhythm parameters were estimated by single cosinor analysis. The acrophases did not differ significantly within a subject in the concomitant measurements of serum and salivary melatonin. The measurements of salivary melatonin levels seem valid for studies on melatonin rhythms, but the melatonin concentrations measured in saliva do not always consistently reflect the absolute concentrations in blood.