Saliva antimicrobial proteins may interact in a common system for host defence. This study applied multivariate analysis as a means of describing inter-person variation in that system. Samples of stimulated parotid saliva were obtained from 198 subjects. Flow rate was determined, and assays run for total protein, lysozyme, lactoferrin, salivary peroxidase and secretory IgA. Correlation and principal components analysis were used to define the relationships between proteins; cluster analysis was used to identify persons with similar protein concentration profiles. All proteins were significantly correlated at p less than 0.002 (r = 0.20-0.52). Principal components analysis identified a major axis of common variation, defined by lysozyme and salivary peroxidase, and a second axis, defined by secretory IgA. Lactoferrin was associated with both axes. Seven major groupings were obtained by cluster analysis; these were significantly different at p less than 0.001. Such groupings may prove useful in comparing the antimicrobial properties of saliva samples.