PSYCHOLOGICAL SCIENCE Research Report 384 Copyright © 2002 American Psychological Society VOL. 13, NO. 4, JULY 2002 DO LONELY DAYS INVADE THE NIGHTS? Potential Social Modulation of Sleep Efficiency John T. Cacioppo, 1 Louise C. Hawkley, 1 Gary G. Berntson, 2 John M. Ernst, 3 Amber C. Gibbs, 1 Robert Stickgold, 4 and J. Allan Hobson 4 1 University of Chicago, 2 Ohio State University, 3 Illinois Wesleyan University, and 4 Harvard Medical School Abstract—Loneliness predicts morbidity and mortality from broad- based causes, but the reasons for this effect remain unclear. Few dif- ferences in traditional health behaviors (e.g., smoking, exercise, nutri- tion) have been found to differentiate lonely and nonlonely individuals. We present evidence that a prototypic restorative behavior—sleep— does make such a differentiation, not through differences in time in bed or in sleep duration, but through differences in efﬁcacy: In the study we report here, lonely individuals evinced poorer sleep efﬁciency and more time awake after sleep onset than nonlonely individuals. These results, which were observed in controlled laboratory conditions and were found to generalize to the home, suggest that lonely individuals may be less resilient than nonlonely individuals in part because they sleep more poorly. These results also raise the possibility that social factors such as loneliness not only may inﬂuence the selection of health behaviors but also may modulate the salubrity of restorative behaviors. Positive social relationships contribute to physical and psychologi- cal well-being across the life span (Baumeister & Leary, 1995; Thomp- son & Nelson, 2001). Loneliness and social isolation, however, predict morbidity and mortality from broad-based causes in later life even after biological risk factors are controlled (House, Landis, & Umberson, 1988; Seeman, 2000). Social isolation is typically deﬁned in the epi- demiological literature in terms of marital status, contact with a close friend, religious membership, and membership in voluntary groups (e.g., Berkman & Syme, 1979). The literature on the hypothesized hu- man need to belong, in contrast, has emphasized the psychological im- pact of social interactions and relationships rather than their presence or absence (e.g., Baumeister & Leary, 1995; Uchino, Cacioppo, & Kiecolt- Glaser, 1996). In the present research, we focus on the construct of loneli- ness because it represents the individual’s interpretation of his or her own social circumstances, or more speciﬁcally, the discrepancy be- tween an individual’s desired and actual relationships (Peplau & Perl- man, 1982). The existing research supports a link between loneliness and mortality, as well (Seeman, 2000). For instance, Herlitz et al. (1998) found that among 1,290 patients who underwent coronary artery bypass surgery, ratings of the statement “I feel lonely” predicted survival at 30 days (relative risk ratio = 2.61; 95% conﬁdence interval: 1.15–5.95) and 5 years (relative risk ratio = 1.78; 95% conﬁdence interval: 1.17– 2.71) after surgery independently of preoperative factors also associ- ated with mortality. Understanding the reasons for the effects of loneliness on health has important social consequences. For instance, Luskin and Newell (1997) reported that in 1994 individuals 65 years or older in age accounted for 36% of all hospital stays and 48% of total days of doctor care even though they accounted for only about 11% of the U.S. population. As the number of older adults and the number of individuals living alone increases, the costs associated with broad-based morbidity and mortal- ity are also expected to increase (House et al., 1988). Health behaviors, an early candidate to explain the effects of isola- tion and loneliness on well-being, could not account for the epidemio- logical ﬁndings (e.g., Seeman, 2000). Health and restorative health behaviors (e.g., sleep, exercise), however, are typically measured in terms of amount (e.g., frequency, time), with the quality or salubrity of these be- haviors assumed to be constant across social and cultural contexts—an assumption that has never been tested. We therefore examined whether loneliness might affect health and well-being through its effects on the efﬁciency of a quintessential restorative behavior—sleep. METHOD The design was a 3 (loneliness: lonely vs. middling vs. nonlonely)  2 (gender: female vs. male) between-participants double-blind facto- rial. Participants were 64 undergraduates at Ohio State University who participated in a larger study that required spending a night in the Clini- cal Research Center (CRC). The participants were recruited from the upper, middle, or lower quintile on the UCLA-R Loneliness Scale (Russell, Peplau, & Cutrona, 1980), which was administered during a screening session. We refer to those individuals who scored in the upper quintile (total score  46) as the lonely group, those scoring in the mid- dle quintile (33  total score  39) as the middling group, and those scoring in the lower quintile (total score  28) as the nonlonely group. The UCLA-R Loneliness Scale is a 20-item scale that assesses an indi- vidual’s social circumstances by self-report. Adequate reliability (e.g., Cu- trona, 1982) and validity (e.g., Russell et al., 1980) have been established. Inclusion criteria for the study were that participants (a) scored no higher than 13 on the Beck Depression Inventory (Beck & Beck, 1972), (b) had a body mass index no greater than 27, (c) were enrolled in at least 6 credit hours in the quarter during which they were tested, (d) were not ﬁrst-quarter freshmen during the quarter they were tested, (e) were not last-quarter seniors, (f) were not speech or needle phobic, (g) were not married nor living with a signiﬁcant other, (h) were U.S. citizens, and (i) scored no more than 8 on a 12-point lie scale taken from the Minnesota Multiphasic Personality Inventory that was included to de- termine the credibility of participants’ responses. We used the Nightcap (Ajilore, Stickgold, Rittenhouse, & Hobson, 1995; Mamelak & Hobson, 1989) to measure sleep efﬁciency in the lonely, middling, and nonlonely participants. Using the Nightcap, for in- stance, Pace-Schott, Kaji, Stickgold, and Hobson (1994) found signiﬁ- cant differences in sleep efﬁciency between self-described “good” and “poor” sleepers. Sleep was measured on the ﬁrst night of our study at the CRC and then approximately 2 weeks later for ﬁve consecutive nights at home. A Nightcap (currently available as REMview sleep monitor manufac- tured by Respironics, Inc., Pittsburgh, Pennsylvania) that employs two Address correspondence to John T. Cacioppo, Department of Psychology, University of Chicago, 5848 S. University Ave., Chicago, IL 60637; e-mail: cacioppo@uchicago.edu.