J oumal of Personality and Social Psychology Copyright 1985 by the American Psychological Association, Inc. 1985. Vol. 49. No. 1.85-94 0022-3514/85/$00.75 Physiological and Affective Predictors of Change in Relationship Satisfaction Robert W. Levenson John M. Gottman Indiana University University of Illinois In 1980, 30 married couples had engaged in a low-conflict and a high-conflict conversational interaction while continuous physiological data were obtained. In a separate session each spouse had provided a continuous self-report of affect while viewing the videotape of the interaction. In 1983, 19 of these couples were re-Iocated to determine the change in relationship satisfaction that had occurred over the preceding 3 years. A broadly based pattern of physiological arousal (across spouses, interaction segments, and physiological measures) in 1980 was found to predict decline in marital satisfaction; the more aroused the couple was during the 1980 interactions, the more their marital satisfaction declined over the ensuing 3 years. Several affective variables also predicted decline in marital satisfaction, including a pronounced sex difference in negative affect reciprocity: Marital satisfaction declined most when husbands did not reciprocate their wives' negative affect, and when wives did reciprocate their husbands' negative affect. In an earlier article (Levenson & Gottman, sions, each spouse viewed the videotape of 1983), we used physiological and affective the conversation and provided a continuous data obtained during conversational interac­ rating of his or her affective state during the tions to predict current levels of marital interaction. Both physiological and affective satisfaction in a sample of 30 married couples. measures were found to be highly predictive As we noted there, measures of marital sat­ of the couple's current level of marital satis­ isfaction, which have been developed in the faction. The relation between physiology and sociological literature over the past 40 to 50 marital satisfaction was particularly powerful. years, have reasonable levels of psychometric For example, our measure of physiological reliability, discriminant validity, and predic­ linkage by itself accounted for over 60% of tive validity. Thus, it makes good scientific the variance in marital satisfaction. The sense to attempt to account for variation in strength of these findings, compared with marital satisfaction on some independent ba­ previous work using questionnaires (see Bur­ sis. To accomplish this, while couples engaged gess, Locke, & Thomes, 1971, for a review) in these conversational interactions, we ob­ and behavioral observation (e.g., Gottman, tained continuous measures of the physiolog­ 1979), led us to an unexpected conclusion. ical state of each spouse and derived a mea­ Prediction of how satisfied a couple was with sure of the physiological interrelatedness or their marriage was more accurate if based on "linkage" between spouses. In separate ses- their physiological responses during an inter­ action than if based on a careful observation of the interaction. The authors would like to thank Oliver Oyama and Being able to predict current levels of Cathy Oyama for their dedicated detective work. This marital satisfaction on some independent basis research was supported by National Institute of Mental Health Grant MH39895 to the first author and Grant is undoubtedly useful, but being able to pre­ MH29910 and NIMH Research Scientist Development dict future levels of marital satisfaction may Award K2-00257 to the second author. be even more important. Marital dissatisfac­ Requests for reprints should be sent to Robert W. tion has been shown to be predictive of Levenson, Department of Psychology, Indiana University, & Bloomington, Indiana 47405, or to John M. Gottman, marital instability (Terman Wallin, 1949) Department of Psychology, University of Illinois, Cham­ and of stress in families, which has deleterious paign, Illinois 61820. effects on the physical and _psychological 85 86 ROBERT LEVENSON AND JOHN GOTTMAN health of the spouses (Bloom, Asher, & White, Method 1978) and on the psychological health and Subjects adjustment of children (Emery, 1982). A technique that enables prediction of which The only information we had on the 30 couples from marriages will become more satisfied and Bloomington, Indiana, who had participated in the original study were their 1980 addreSses and places of employment. which will become less satisfied would have Fortunately, the population of southern Indiana is quite great importance both for understanding the stable; a number of the couples still lived in Bloomington nature of relationship satisfaction over time or in the surrounding areas. Others who had moved were and for identifying marriages that are at high tracked down by using such resources as employer records risk for future instability and stress. These and records kept by educational institutions and by contacting people with the same last name in search of marriages would be prime candidates for relatives. In all, 21 couples were located and asked to therapeutic intervention. participate in the follow-up study. One couple declined, Research on marital satisfaction has been and one couple had divorced, with only the wife agreeing overwhelmingly cross sectional. Among the to participate. Thus, we ended up with 19 intact couples for the follow-up (63%). This percentage compares favor­ few longitudinal studies, Terman and Wallin ably with other recent longitudinal studies, such as Bentler J (1949) found that their inventory of marital and Newcomb's (1978) 4-year follow-up (48%) and Mark­ happiness was moderately successful (r = .47) man's (1979) 2lh-year follow-up (54%). Couples electing in predicting marital stability. Bentler and to participate were mailed a package of questionnaires ( to complete in return for a $15 payment. Newcomb (1978) found that a number of demographic and personality variables (e.g., Procedure: 1980 Study older wives, less extraverted husbands) mea­ sured at the time of marriage were predictive The methodological details of the original study were I of marital stability 4 years later, but the presented in Levenson and Gottman (1983) and thus will be reviewed only briefly here. correlations were low (rs = .23 to .33). Mark­ Married couples were recruited by newspaper adver­ man (1979, 1981), in the only longitudinal tisement and were scheduled for three laboratory sessions. study that attempted to predict future marital The first session was scheduled for a time when the satisfaction on the basis of affective self­ couple would not have spoken to each other for at least 8 hours. This session consisted of two IS-min conversa­ report during premarital interactions, found tions, each preceded by as-min preinteractional baseline that the positivity of ratings of the impact of during which they sat in silence. In the first conversation, statements made during these laboratory in­ the couple was asked to discuss the "events of the day" teractions was predictive of marital satisfac­ as if they were home alone at the day's end. In the second 1 conversation, they discussed a conflictive problem area tion in nine couples followed up 2 h (r = .67) in their marriage. In the second and third sessions, each and 51h (r = .59) years later. Skolnick (1981), spouse returned separately to view the videotape of the using longitudinal data from the Berkeley first session's interaction. A continuous rating of affect Growth Study, the Berkeley Guidance Study, was obtained by having the spouse manipulate a rating dial that traversed a 9-point scale (anchored by very and the Oakland Growth Study, found that negative and very positive on the extremes and by neutral coders' ratings of the quality of adolescent at the center). Spouses were instructed to adjust the dial relationships were moderately predictive of as often as necessary so that it always reflected the way subjects' marital satisfaction in midlife. they felt during the interaction. A laboratory computer monitored the position of the dial continuously and Given the paucity of longitudinal research calculated an average every lOs. in this area and our previous finding that Four physiological measures were obtained from each physiological and affective measures were ex­ spouse during the first session's baselines and interactions: cellent predictors of current levels of marital (a) heart rate, measured by the interbeat interval (lBI); (b) pulse transmission time (PTT) to the finger, that is, satisfaction (Levenson & Gottman, 1983), we the time interval between the R-wave on the electrocar­ decided to undertake a follow-up study to diogram and the arrival of the pulse pressure wave at the determine how well the measures we obtained finger tip; (c) skin conductance level (SCL); and (d) general somatic activity (ACT), a global measure of in that study, which was conducted during bodily movement. The laboratory computer monitored the summer of 1980, would do in predicting these physiological variables continuously, averaging them changes in marital satisfaction 3 years later. every !O s. Thus, in summer 1983, we began the process Both affective and physiological variables were enlisted to attempt to predict couples' changing levels of marital of trying to re-Iocate the 30 couples who had satisfaction. The computations used to derive these pre­ participated in the original study. dictor variables can be found in our earlier report PHYSIOLOGICAL AND AFFECTIVE PREDICTORS