Psychoneuroendocrinology (2004) 29, 1153–1162 www.elsevier.com/locate/psyneuen Social variables predict between-subject but not day-to-day variation in the testosterone of US men Peter B. Gray a, * , Benjamin C. Campbell b , Frank W. Marlowe a , Susan F. Lipson a , Peter T. Ellison a a Department of Anthropology, Peabody Museum, Harvard University, Cambridge, MA 02138, USA b Department of Anthropology, Boston University, 232 Bay State Road, Boston, MA, 02215 USA Received 22 July 2003; received in revised form 27 January 2004; accepted 27 January 2004 KEYWORDS Testosterone; Marriage; Fatherhood; Mating effort; Evolution; Circadian rhythm Summary Previous research has shown lower testosterone (T) levels associated with involvement in committed, romantic relationships (‘‘paired’’ men) and paternal care in eight studies of North American men. An unanswered question is whether differences in male T levels associated with relationship status better reflect state (e.g., a man has lower T levels because he is involved in a relationship) or trait (e.g., low T men are more inclined toward such relationships) effects. Toward addressing this question, this paper presents data on male salivary T levels among a sample of 65 men varying in marital and parental status. Subjects collected saliva samples (at approximately waking, 17:00 and 21:00 h) and filled out ques- tionnaires concerning their activities on four days. Each subject collected samples in two settings that varied in social interactions: for unmarried men, two working and two non-working days; for married non-fathers, two days spent mostly with their wives and two days spent mostly away from their wives; and for married fathers, two days spent mostly with their young children and two days spent mostly away from their children. Analyses revealed no significant within-subject T differ- ences between these different conditions. However, between-group analyses revealed that married men had lower evening T levels than unmarried men, corro- borating existing North American studies of male T and relationship status. These results suggest that day-to-day differences in social interactions may not be associa- ted with differences in T levels, and lend further support to the growing body of evidence that hormone–behavior effect sizes may be greater in the afternoon and evening than in the morning. # 2004 Elsevier Ltd. All rights reserved. 1. Introduction Previous research has shown an association between male testosterone (T) levels and relationship status. North American findings indi- * Corresponding author. Division of Endocrinology, Metab- olism and Molecular Medicine, Charles Drew University, 1731 East 120th Street, Los Angeles, CA 90059, United States. Tel.: +1-323-563-4811; fax: +1-323-563-9352. E-mail address: pegray@cdrewu.edu (P.B. Gray). 0306-4530/$ - see front matter # 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.psyneuen.2004.01.008