Salivary cortisol has been useful for evaluating children's physiological responses to stress and for identifying factors that predict their magnitude and duration. However, results have been somewhat equivocal across studies, and this has motivated researchers to identify sources of variance and error. Here, we examined the prevalence of preschoolers' noncompliance during saliva collection and aimed to learn about noncompliant children in terms of their hypothalamus–pituitary–adrenal function, behavior in other situations, and symptoms of behavioral problems. Results were based on measures of cortisol, children's behavior during saliva collection and a mother–child teaching interaction, and ratings of problem behavior by teachers and parents. Results show that 12% (21/174) of the sample was noncompliant on at least one of the collection trials. Children, who were noncompliant but did not outright refuse saliva collection, had higher cortisol than did compliant children. Children who were noncompliant during saliva collection were likely to be noncompliant during the teaching episode, and they were perceived as having more internalizing symptoms than compliant children. These results suggest that children's noncompliance during saliva collection can be a source of nonrandom missing data or extreme cortisol values, which should be considered in future studies. © 2011 Wiley Periodicals, Inc. Dev Psychobiol 54:113-123, 2012.