Journal of Family Psychology 2000, Vol. 14, No. 1,42-58 Copyright 2000 by the American Psychological Association, Inc. O893-32O0/O0/$5.O0 DOI: 10.1037//0893-3200.14.I.42 Predicting Marital Stability and Divorce in Newlywed Couples Sybil Carrere, Kim T. Buehlman, and John M. Gottman University of Washington James A. Coan University of Arizona Lionel Ruckstuhl University of Nevada, Reno A longitudinal study with 95 newlywed couples examined the power of the Oral History Interview to predict stable marital relationships and divorce. A principal- components analysis of the interview with the couples (Time 1) identified a latent variable, perceived marital bond, that was significant in predicting which couples would remain married or divorce within the first 5 years of their marriage. A discriminant function analysis of the newlywed oral history data predicted, with 87.4% accuracy, those couples whose marriages remained intact or broke up at the Time 2 data collection point. The oral history data predicted with 81% accuracy those couples who remained married or divorced at the Time 3 data collection point. This study offers support for causal linkages between perceptual biases and selective attention on the path of marriage. Why do marriages last? Given the grim divorce statistics and the repercussions of divorce, navigating the road of marital stability seems a hazardous task. Nearly one third of all marriages fail within the first 5 years (National Center for Health Statistics, 1991), and between one half and two thirds end in divorce (Cherlin, 1992; Martin & Bumpass, 1989). The consequences of separation and divorce can be severe. Research indicates that individu- als who are separated from their spouses or divorced experience greater rates of psychopa- thology, physical illness, suicide, homicide, Sybil Carrere, Kim T. Buehlman, and John M. Gottman, Department of Psychology, University of Washington; James A. Coan, Department of Psychol- ogy, University of Arizona; Lionel Ruckstuhl, Depart- ment of Psychology, University of Nevada, Reno. We would like to thank the couples who partici- pated in this study. We also thank Catherine Swanson and Alyson Shapiro for their constructive comments on an earlier version of this article. Correspondence concerning this article should be addressed to Sybil Carrere, Department of Psychol- ogy, Box 351525, University of Washington, Seattle, Washington 98195. Electronic mail may be sent to carrere @ u .washington.edu. violence, and mortality from disease (e.g., Berkman & Breslow, 1983; Berkman & Syme, 1979; Bloom, Asher, & White, 1978; Burman & Margolin, 1992). In light of these marital stability statistics and the greater risk for mental and physical health problems among separated and divorced individuals, identifying the factors that help marriages survive has important implications. Furthermore, understanding how these factors influence marital stability will help build a theory of marital quality and stability. This was the objective of the present study. One area of marital research that is receiving increasing attention is marital cognitions. Fin- cham, Bradbury, and Scott (1990) have sug- gested that it is important to understand the role cognitions play in driving emotional expression, behavioral interactions, and satisfaction in marriage. Much of the cognitive research has been in the area of attribution, focusing on the explanations spouses give for behavior within the marriage (e.g., Epstein, Pretzer, & Fleming, 1987; Fincham & Bradbury, 1987; Holtzworth- Munroe & Jacobson, 1985; Kyle & Falbo, 1985; Thompson & Snyder, 1986). Baucom and his associates (Baucom, Epstein, Sayers, & Sher, 1989) provided an overview and structure 42
Transcript
Journal of Family Psychology2000, Vol. 14, No. 1,42-58
Copyright 2000 by the American Psychological Association, Inc.O893-32O0/O0/$5.O0 DOI: 10.1037//0893-3200.14.I.42
Predicting Marital Stability and Divorcein Newly wed Couples
Sybil Carrere, Kim T. Buehlman, andJohn M. Gottman
University of Washington
James A. CoanUniversity of Arizona
Lionel RuckstuhlUniversity of Nevada, Reno
A longitudinal study with 95 newlywed couples examined the power of the OralHistory Interview to predict stable marital relationships and divorce. A principal-components analysis of the interview with the couples (Time 1) identified a latentvariable, perceived marital bond, that was significant in predicting which coupleswould remain married or divorce within the first 5 years of their marriage. Adiscriminant function analysis of the newlywed oral history data predicted, with87.4% accuracy, those couples whose marriages remained intact or broke up at theTime 2 data collection point. The oral history data predicted with 81% accuracythose couples who remained married or divorced at the Time 3 data collection point.This study offers support for causal linkages between perceptual biases and selectiveattention on the path of marriage.
Why do marriages last? Given the grimdivorce statistics and the repercussions ofdivorce, navigating the road of marital stabilityseems a hazardous task. Nearly one third of allmarriages fail within the first 5 years (NationalCenter for Health Statistics, 1991), and betweenone half and two thirds end in divorce (Cherlin,1992; Martin & Bumpass, 1989).
The consequences of separation and divorcecan be severe. Research indicates that individu-als who are separated from their spouses ordivorced experience greater rates of psychopa-thology, physical illness, suicide, homicide,
Sybil Carrere, Kim T. Buehlman, and John M.Gottman, Department of Psychology, University ofWashington; James A. Coan, Department of Psychol-ogy, University of Arizona; Lionel Ruckstuhl, Depart-ment of Psychology, University of Nevada, Reno.
We would like to thank the couples who partici-pated in this study. We also thank Catherine Swansonand Alyson Shapiro for their constructive commentson an earlier version of this article.
Correspondence concerning this article should beaddressed to Sybil Carrere, Department of Psychol-ogy, Box 351525, University of Washington, Seattle,Washington 98195. Electronic mail may be sent tocarrere @ u .washington.edu.
violence, and mortality from disease (e.g.,Berkman & Breslow, 1983; Berkman & Syme,1979; Bloom, Asher, & White, 1978; Burman &Margolin, 1992). In light of these maritalstability statistics and the greater risk for mentaland physical health problems among separatedand divorced individuals, identifying the factorsthat help marriages survive has importantimplications. Furthermore, understanding howthese factors influence marital stability will helpbuild a theory of marital quality and stability.This was the objective of the present study.
One area of marital research that is receivingincreasing attention is marital cognitions. Fin-cham, Bradbury, and Scott (1990) have sug-gested that it is important to understand the rolecognitions play in driving emotional expression,behavioral interactions, and satisfaction inmarriage. Much of the cognitive research hasbeen in the area of attribution, focusing on theexplanations spouses give for behavior withinthe marriage (e.g., Epstein, Pretzer, & Fleming,1987; Fincham & Bradbury, 1987; Holtzworth-Munroe & Jacobson, 1985; Kyle & Falbo, 1985;Thompson & Snyder, 1986). Baucom and hisassociates (Baucom, Epstein, Sayers, & Sher,1989) provided an overview and structure
42
PREDICTING MARITAL STABILITY 43
for the study of cognition in marriage. Theysuggested that, in addition to attributions,cognitive phenomena taking place in intimaterelationships include expectancies (predictingfuture outcomes from current interactions),assumptions (how people think the worldoperates), standards (how the world shouldoperate), and the perceptual process of selectiveattention (the aspects of an event to which anindividual pays attention on the basis of acognitive schema). The present study focusedon perceptual bias and the tendency to selec-tively attend to only certain characteristics ofevents. More specifically, a primary goal was todetermine the influence of the couple's relation-ship perceptions on the stability of the marriage.
According to Gottman (1993, 1994), percep-tion is one of the three domains (perception,physiology, and behavior) that individually andtogether act as interactive thermostats in mar-riage. In this "core triad of balance" theory,Gottman has proposed that each of the threedomains has the potential for balance, ahomeostatic set point associated with harmonyin the marriage. Each of the domains has anegative threshold that, when exceeded, resultsin increased instability within the marriage.Because these three domains are interactive andoverlapping, changes in one domain can causerepercussions in another domain. For example,among newlywed couples who remained hap-pily married over the first 6 years of theirmarriage, a wife's use of humor in a conflictdiscussion is associated with a decrease in thehusband's heart rate (Gottman, Coan, Carrere, &Swanson, 1998).
Perceptions play a special role in this coretriad of balance theory. The perceptual compo-nent of the theory builds on previous cognitivemarital research (e.g., Fincham et al., 1990;Notarius, Benson, Sloane, Vanzetti, & Homyak,1989; Weiss, 1980). Fincham et al. (1990)argued that cognition could be used in marriageto understand past and present events. Thisperceptual framework for past and presentevents shapes the expectations and behaviors ofthe spouses in the future. Fincham et al. used theinformation-processing approach from cogni-tive theory to describe how memory and maritalcognitions are associated. Information is orga-nized and structured in memory on the basis ofwhat is cognitively salient. Likewise, the moresalient and thematically coherent the organiza-
tion of the memory, the more likely theindividual is to retrieve particular types ofevents from memory. Bradbury and Fincham(1987) linked memory and affect, arguing thatindividuals are most likely to retrieve units ofmemory that are congruent with the presentmood they are experiencing. Thus, distressedcouples are more likely to remember negativeevents than positive ones. Fincham et al. (1990)went on to theorize that these negative eventsfrom the past are used by the unhappy spouses tomake sense of present marital interactions and toshape future behavior.
Weiss (1980) described a relationship pro-cess, sentiment override, that includes many ofthe same cognitive linkages described byFincham et al. (1990). Weiss defined sentimentoverride as the tendency to assess one's spouse'sbehavior as either positive or negative on thebasis of more globally held perceptions aboutthe partner rather than the objective nature of thepartner's immediate behavior. Sentiment over-ride is a kind of perceptual filter with which oneviews the behavior of one's spouse. Notariusand his associates (1989) reported that negativesentiment override is associated with distressedwives' tendency to rate their husbands' neutraland negative behavior as more negative than doother wives. This perceptual filter appears toresult in selective attention as well. Whatsomeone pays attention to and remembers is afunction of his or her perceptual filter. Forexample, Robinson and Price (1980) had trainedobservers go to both distressed and nondis-tressed couples' homes and evaluate the behav-ior of the couples. The couples also evaluatedtheir own behavior. Robinson and Price foundthat the distressed couples underreported thenumber of positive events. Thus, these dis-tressed couples selectively attended to negativeinteractions and did not accurately rememberthe number of positive events.
Does the manner in which spouses rememberthe past and their cognitive schema about theirpartner and the marriage help in predicting thefuture stability of the marriage? Our laboratoryhas produced some evidence that it does.Buehlman, Gottman, and Katz (1992) used theOral History Interview (Krokoff, 1984a) tomeasure spouses' global perceptions about theirmarriage and each other (in this study, we referto global marital perceptions as the perceivedmarital bond). In a joint interview, spouses were
44 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
asked to tell the story of their relationship fromthe time they met until the present day. Thecouples were also asked about the good and hardtimes in their marriage, as well as theirphilosophy about marriage. The interview wascoded for how the couple told their story ratherthan based on the content of what they said. Theinterview measured the couple's perceptions byfocusing on the positive or negative qualities ofthe relationship that predominated in the tellingof the story. The coding system is consistentwith Fincham et al.'s (1990) thesis that individu-als are most likely to retrieve units of memorythat are congruent with their present perceptionsabout the marriage.
Using the Oral History Interview, Buehlmanand her associates (1992) were able to predict,with 94% accuracy, those couples who woulddivorce or stay married in a longitudinal studyof 56 married couples. The couple's perceivedmarital bond was associated with maritalstability. Couples in which spouses were morecritical of their partners, disillusioned about themarriage, and believed the challenges of themarriage were outside their personal controlwere more likely to have divorced by the 3-yearfollow-up. Hence, how a couple told the story oftheir relationship could predict their likelihoodof marital stability or divorce.
Buehlman et al.'s (1992) work suggests thatglobal perceptions held by couples about theirmarriage and each other help predict the futurecourse of the marriage, but the generalizabilityof the study outcomes is limited because thestudy was conducted with a sample of marriedcouples at one specific point in the life cycle of amarriage (i.e., couples with preschoolers). Canthe marital perceptions indexed by the OralHistory Interview provide insights about thefuture of a marriage when global perceptions aremeasured at different stages of the marital lifecycle? For example, are perceptions about one'sspouse and the marriage in a greater state of fluxat the beginning of a marriage? The couples inBuehlman et al.'s study had young children andhad been married at least 4 to 5 years beforeparticipating in the research. The global maritalperceptions of the couples in their study mayhave had time to form and solidify. In contrast,early married life appears to be a period ofchange and adjustment. Spouses* perceptions ofeach other and the marriage may be in the
process of forming or may be more fluctuatingin nature as the spouses adjust to each other.
Although some studies show that patterns offuture marital instability can be found inpremarital and early marital relationships (Gott-man, 1979; Markman, 1981; Markman, Floyd,Stanley, & Storaasli, 1988), there is reason tobelieve that newlywed couples may be verydifferent from couples who have been marriedlonger or who have become parents. Researchindicates that newlyweds are in a period of flux,and their patterns of interaction may be moreopen to influence and change (Behrens &Sanders, 1994; Hawley & Olson, 1995). Forexample, Noller and Feeney (1998) found that,during the first 2 years of marriage, coupleschanged their communication patterns. Mark-man, Floyd, Stanley, and Jamieson (1984)theorized that each stage of marriage has its owndevelopmental challenges. They proposed thatthe developmental task for newlyweds is toleam how to communicate successfully and findways to resolve conflict in a constructivefashion.
Gottman (1994) and others (e.g., Fincham etal., 1990) have argued that perceptions areinfluenced by behavior, and thus it is likely thatnewlyweds' communication and conflict-resolu-tion adjustments are influencing and changingthe couple's marital perceptions. By the timecouples become parents, many of their patternsof communication and perceptions about themarriage and their partner may have lost theplasticity of the newlywed phase. This plasticityin newlywed couples may make it more difficultto measure the stable global perceptions thespouses have of each other and the marriage thatwere predictive of future marital quality anddivorce in the Buehlman et al. (1992) research.The reason is that these global marital percep-tions have not yet "jelled."
In addition to the plasticity and openness tochange more apt to be found among newlyweds,these couples are not as likely to have the historyof experience with each other required todevelop a sense of unity and identity as amarried couple. In the original study byBuehlman and her colleagues (1992), a couple'sperception of unity and their belief that theirstruggles against hard times made their relation-ship stronger were both important componentsof what predicted marital stability. Will a shorterrelationship history for newlyweds mean thesecouples have not yet had the opportunity to
PREDICTING MARITAL STABILITY 45
develop a marital bond and an identity as amarried couple? If they have not reached thisstage in the development of their relationships,the Oral History Interview would incorrectlyproduce results indicating that these newlywedsare distressed rather than correctly reflecting theadjustments and beginning stages of developinga sense of unity as a couple. All of this suggeststhat the Oral History Interview may notaccurately assess the global marital perceptionsheld by newlyweds because these maritalperceptions are still fluctuating and becausethese couples have not had enough historytogether to form perceptions designed to bemeasured by certain dimensions of the OralHistory Interview scale (e.g., a sense of unity).
The goal of this study was to examine how theperceptions newlyweds have about their partnerand the marriage in the 1st year of marriageinfluence the stability of the marriage. Wewanted to determine whether the Oral HistoryInterview is a cognitive measure of globalmarital perceptions that has predictive validityfor married couples at different stages of themarital life cycle. It would strengthen theexternal validity of the instrument if similarpatterns in the Oral History Interview predictdivorce and marital stability in both newlywedcouples and couples married for longer periodsof time.
Method
Participants
Between 1989 and 1992, a two-stage samplingprocedure was used to draw a sample of newlywedcouples from the Puget Sound area in Washington.Couples were initially recruited via newspaperadvertisements. Newlyweds interested in participat-ing in the study were asked to phone the laboratoryand leave information so that they could be contacted.The wives in these marriages were then administereda screening telephone interview that included thetelephone version of the Marital Adjustment Test(MAT), a scale measuring marital satisfaction (Kro-koff, 1984b; Locke & Wallace, 1959). For financialand logistical reasons, only the wives were inter-viewed at this screening stage of the study. To beeligible for the study, the couples had to have marriedfor the first time within 6 months of participating inthe study, had to be childless, and had to have noserious illnesses (e.g., cancer, cardiovascular disease,or emphysema). The sample was selected so that therewas an even distribution of marital satisfaction amongthe wives' scores on the telephone version of theMAT. By even distribution of marital satisfaction, we
mean that we had equal numbers of wives at eachpoint of the marital satisfaction distribution. This isin contrast to a bell-shaped distribution that hasgreater numbers of individuals in the middle of thedistribution. This even distribution was chosen so thatwe might oversample both the very happy and thevery distressed couples. The sample was also selectedto match the racial and ethnic demographics of themetropolitan Seattle area (City of Seattle PlanningCommission, 1990).
The newlywed study involved several components,including a laboratory-based marital interactionsession, a 24-hr stay in a studio apartment laboratory,a longitudinal follow-up study (a 1-year follow-upafter the marital interaction session and yearlytelephone interviews and mailed questionnaires be-tween 1993 and 1998), and research on the couples'transition to parenthood. This article focuses on thestudy participants who took part in the maritalinteraction sessions (which included the Oral HistoryInterview and questionnaire data) and the longitudi-nal study.
One hundred thirty couples were initially selectedto participate in the first phase of the study, thelaboratory-based marital interaction session. Thedemographic characteristics of these newly marriedcouples at first contact were as follows: (a) wife'smean age = 25.40 years (SD = 3.50); (b) husband'smean age = 26.54 years (SD - 4.22); (c) wife's meanmarital satisfaction score = 120.45 (SD = 19.69);and (d) husband's mean marital satisfaction score ~115.87 (SD= 18.41). Couples had a combinedmedian income between $25,000 and $39,000. Themean education level for both husbands and wives inthe study was a 4-year college degree. Sixty-fourpercent of the couples had lived together beforegetting married.
Ninety-five couples completed a marital interactionlaboratory session, the Oral History Interview, and thelongitudinal components of the newlywed study. Ofthe original 130 couples, 35 did not complete all threeof these components of the study. Analyses wereconducted to compare the couples who completed allthree components with the couples who did not. Therewas no significant difference in marital satisfactionscores for the husbands (M = 115.21, SD = 17.63,for husbands in the incomplete data group;M= 115.83, SD = 19.00, for husbands in the com-plete data group), *(121) = -0.15, ns. Nor was therea significant difference in the marital satisfactionscores for the wives (M = 115.04, SD = 22.19, forwives in the incomplete data group; M = 121.12,SD = 18.62, for wives in the complete data group),f(121) - -1.44, ns. There was a significant agedifference between the groups of husbands. Thehusbands in the incomplete data group (Af = 24.88years, SD = 4.87) were about 2 years younger, onaverage, than the husbands in the complete data group(M = 26.91 years, SD = 3.87), r(121) = -2.18, p <
46 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
.05. The wives in the incomplete data group(M = 24.10 years, SD = 2.86) were also significantlyyounger than the wives in the completed data group(Af = 25.86 years, SD = 3.70), f(121) = -2.19, p <.05. The couples in the completed data set were morelikely to have lived together before getting married(73% of the complete data couples vs. 46% of theincomplete data couples). The median income forboth groups was between $25,000 and $39,000. Thedegrees of freedom for these /-test analyses werereduced because of incomplete data for some of thecouples on the demographic variables.
Measures and Materials
Oral history coding. The Oral History Interviewis a semistructured interview conducted with both thehusband and wife present (the questions that composethe interview can be found in Buehlman & Gottman,1996). The interview explores the history of thecouple's relationship, the spouses' philosophy aboutmarriage, and how their parents' marriages comparewith their own marriage. Questions about the historyof the relationship focus on the couple's courtship,their wedding, and the good and hard times of theirmarriage. When the spouses discuss their philosophyof marriage, they are asked to think of a goodmarriage and a bad marriage and discuss thedifferences between these kinds of marriage (Buehl-man etal., 1992).
The Oral History Interview coding system mea-sures spouses* global perceptions about the marriageand about each other (see Buehlman & Gottman,1996, for a complete description of the observationalorigins of the coding system and detailed guidelineson the coding rules used for each subscale of thesystem). The thesis of the coding system is consistentwith Fincham et al.'s (1990) proposal that individualsare most likely to retrieve units of memory that arecongruent with their present perceptions about themarriage. Rather than coding the content of theinterview (e.g., how long the couple dated beforebecoming engaged, whether the couple has children,and whether the couple has a good relationship within-laws), the coding system indexes how the coupletells the story of the relationship. More specifically, itfocuses on the positive or negative nature of what thespouses choose to recall from the history of theirrelationship. For example, some couples minimizenegative aspects and emphasize the romance ornaturalness of the relationship. Other couples canonly remember how hard it was to get together andwhat a struggle the marriage has been.
The coding system also measures how each spousedescribes and talks about his or her partner in thetelling of the story. Again, the focus is on the tenor ofthe description over the course of the interview. Forexample, when they are asked to describe what firstattracted them to their partner, do spouses seem
unsure, or do flattering descriptions of their partner'spersonality or appearance readily come to mind forthem? In a similar fashion, the coding system takesinto account how the spouses interact as they tell thestory of their relationship. For example, do they teaseeach other? Do they finish each other's sentences andvalidate what the other person has said? Alternatively,do they snipe at each other, argue about the history ofevents, or describe their spouse or the history of themarriage in cynical or disillusioned tones?
The coding system taps into global maritalperceptions via eight dimensions-subscales (Fondness/Affection, We-ness, Expansiveness, Negativity, Disap-pointment and Disillusionment, Chaos, Volatility, andGlorifying the Struggle). These subscales originallycame from Oral History Interview observations (seeBuehlman & Gottman, 1996; Buehlman et al., 1992).Buehlman and Gottman (1996) created the codingsystem to determine whether the way in whichspouses talked about their marriage and each othercould provide insights about marital stability ordivorce trajectories. Buehlman et al. (1992) tested thepsychometric properties of the instrument in theirstudy of married couples with young children andfound the Oral History Interview coding system tohave good internal construct validity as well as strongpredictive validity. The current study extends theconstruct and external validity of Buehlman et al.'s(1992) research.
Each of the eight subscales of the Oral HistoryInterview is made up of six to nine items. The coderscores the spouses on each item using a 5-pointLikert-type scale ranging from strongly agree tostrongly disagree. Three of the subscales are positivein nature: (a) Fondness/Affection, (b) We-ness, and(c) Expansiveness. Each of these three positivesubscales is rated separately for the husband and forthe wife, but the items are identical for both thehusband and the wife. The Fondness/Affection scalerates each spouse's expressions of pride, fondness,and affection for his or her partner. This subscaleincludes items such as "Husband compliments wifeduring the interview" (husband's subscale) and"Wife is proud of her husband or specific qualitiesabout her husband" (wife's subscale). The We-nessscale reflects the degree to which each spouse usesterms during the interview that indicate unification inthe marriage. Examples of items from this subscaleinclude "Wife emphasizes 'we' as opposed to 'he' orT " (wife's subscale) and "Husband emphasizes thesame beliefs, values, and goals as his wife"(husband's subscale). The Expansiveness scale mea-sures how expressive and expansive the spouses arein the interview. This dimension indexes not only howexpressive each spouse is but also how the spousesrespond to and expand on what their partner is saying.This is in contrast to spouses who respond toquestions with a few short sentences, seem with-drawn, and do not add to what their partner says.
PREDICTING MARITAL STABILITY 47
Example items include "Wife recalls easily their firstdate, proposal, wedding, etc." (wife's subscale) and"Husband expands on what his wife is describing orsaying" (husband's subscale).
Two of the coding dimensions are negative: (a)Negativity and (b) Disappointment and Disillusion-ment. Both of these negative coding dimensionsinvolve separate subscales for the husband and thewife, but the items are identical. The Negativity scaleindexes the extent to which spouses are critical oftheir partner, are vague about what attracted them totheir partner, and display negative affect toward theirpartner. Examples of the items used to code thissubscale include "Husband does not know whatattracted him to his wife" (husband's subscale) and"Wife disagrees with husband during the interview"(wife's subscale). The Disappointment and Disillusion-ment scale assesses the degree to which each memberof the couple has given up on the marriage byexpressing depression about the relationship or notbeing able to articulate what makes the marriagework. Example items include "Husband is depressedwhen talking about his marriage" (husband's subs-cale) and "Wife mentions how difficult their marriageis or marriage in general is" (wife's subscale).
Three subscales evaluate information about howthe couple reports handling marital conflict: (a)Chaos, (b) Volatility, and (c) Glorifying the Struggle.Each of these subscales involves one score percouple. The Chaos scale rates the degree to which thecouple feels out of control of their lives and buffetedabout by elements outside of their control. Exampleitems include "The couple has a lot of unexpected orout of control marital conflict" and "The couplebelieve unexpected problems have weakened theirrelationship." The Volatility scale measures theintensity, both positive and negative, of the spouses'feelings for each other. Highly volatile couplesexpress feelings of great passion and yet fightfrequently. Examples of items that are used tomeasure this subscale include "The couple mentionthey like a good fight (having it out)" and "They feelemotions and other things intensely." The Glorifyingthe Struggle scale assesses the extent to which acouple has gone through difficult times but perceivesthe marriage to be stronger because of theseexperiences. Their marriage is the center of theirlives, and they are proud of the struggles they havegone through. Example items include "The coupleexpress that marriage is a struggle, but it is worth it"and "The couple is proud of the hardships they havebeen through."
The coding system also evaluates the degree towhich couples* gender roles within the marriage arestereotypical and the degree to which their beliefs andvalues are traditional (Gender Stereotypy Scale; onescore per couple). Items used to index this dimensioninclude "The wife is the emotional gatekeeper in the
marriage" and "The husband is invested in being theprovider, main wage earner."
Intercoder reliability was calculated via intraclasscoefficients. A perceived marital bond measure wasderived from the Oral History Interview (described indetail in the Results section). Its intercoder reliabilityis presented here as a matter of consistency with theother data on reliability. The intraclass correlation forthe Oral History Interview perceived marital bondmeasure was .75. The intraclass correlations for theother Oral History Interview subscales were asfollows: Husband's Fondness/Affection, .76; Wife'sFondness/Affection, .76; Husband's We-ness, .71; Wife'sWe-ness, .76; Husband's Expansiveness, .47; Wife'sExpansiveness, .66; Husband's Negativity, .81; Wife'sNegativity, .71; Husband's Disappointment and Disil-lusionment, .79; Wife's Disappointment and Disillu-sionment, .72; Chaos, .68; Volatility, .56; Glorifyingthe Struggle, .37; and Gender Stereotypy, .35.
Marital satisfaction. The MAT (Krokoff, 1984b;Locke & Wallace, 1959) was administered to thewives during the initial telephone interview, and thepaper-and pencil version of the scale was adminis-tered to both the husbands and wives a week or twobefore the laboratory session. The scores on thepaper-and-pencil version were used in the analysesdescribed in this article. The MAT assesses maritalsatisfaction and is frequently used in marital researchbecause of its strength in reliably and validlydistinguishing between happily and unhappily mar-ried couples. The telephone version was used tointerview wives about their marital satisfactionduring the sample selection phase of the study. Asdescribed earlier, the sample was selected so thatthere was an even distribution of marital satisfactionamong the wives' scores on the telephone version ofthe MAT. This version includes the same items usedin the paper-and-pencil version (Krokoff, 1984b).Higher scores on the MAT represent greater maritalsatisfaction.
Social desirability. The Marlowe-Crowne SocialDesirability Scale (Crowne & Marlowe, 1960) wasused to test the discriminant validity of the OralHistory Interview coding system. The Marlowe-Crowne Social Desirability Scale is a well-established instrument used to measure the need ofindividuals to engage in culturally acceptable andapproved behaviors. It was used as a construct thatcould offer a competing explanation for couples'scores on the different subscales of the Oral HistoryInterview coding system.
Procedure
Newly wed couples (N = 130) were recruited forthe study over the course of 3 years. During each ofthe 3 years, approximately 35 to 45 couplesparticipated in marital interaction laboratory proce-dures. These marital interaction procedures are not a
48 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
part of the analyses described in the present article;however, a report of the procedures can be found inGottman et al. (1998). As part of a larger packet ofquestionnaires, couples were given a paper-and-pencil version of the MAT and the Marlowe-CrowneSocial Desirability Scale to fill out and bring withthem to their first marital interaction session in thelaboratory. The MAT questionnaires the couples filledout and brought to the first laboratory session wereused in the analyses described in this study.
Oral History Interview. As a result of the designof the original study, The Oral History Interview wasconducted in one of two types of settings. Of the 95couples included in the analyses described in thisarticle, 40 were selected to be studied in more detail.These 40 couples were asked to spend 24 hr in ourapartment laboratory while we videotaped theirwaking behaviors, indexed their heart rate, and tookendocrine and immunological measures. For logisticand financial reasons, we did not include all couplesin these procedures. The couples who participated inthe apartment laboratory procedures were given theOral History Interview in the early evening of theirstay in the lab. These interviews in the apartmentlaboratory were videotaped. The remaining 55couples included in the analyses described here wereinterviewed in their home, and the interviews wererecorded via audiotape.
A multivariate analysis of variance (MANOVA)was run on the Oral History Interview variables todetermine whether there were significant differencesbetween the couples interviewed at home and thoseinterviewed in the apartment laboratory. We wantedto determine whether the setting of the interviewinfluenced the coding. Eleven of the subscales werecombined into one variable (perceived marital bond)because of concerns about multicollinearity andredundancy of the dependent variables. The perceivedmarital bond variable was based on a principal-components analysis described later. The other threeOral History Interview variables included in theMANOVA were glorifying the struggle, volatility,and gender stereotypy. Using Wilks's lambda as thecriterion, we found no significant differences betweenthe couples interviewed in the apartment laboratoryand those interviewed in their homes on perceivedmarital bond scores, F(4,90) = 2.01, ns.
Longitudinal follow-up. Couples were recon-tacted by telephone and mail each year between 1993and 1998. Each spouse was interviewed separately bytelephone, and spouses described their marital status.Divorce rates were based on husbands' and wives'self-reports of actual divorce.
Results
Data analyses addressed four objectives.First, principal-components analysis was used to
establish the internal construct validity of theOral History Interview. Second, correlationaldata analyses were used to examine thediscriminant construct validity of the scale.Third, discriminant function analyses were usedto examine the predictive validity of the OralHistory Interview at two time points and tocompare its predictive strength as an instrumentwith that of the MAT. Finally, an analysis ofvariance was used to explore the differences inOral History Interview scores among threegroups of couples: those divorcing by the 1995rime point, those divorcing after the 1995 timepoint but before the 1998 time point, and thoseremaining married.
Internal Construct Validity Criteria
A principal-components analysis was con-ducted to determine the latent variables presentin the newlywed couples' oral history and toreplicate the scaling steps taken by Buehlman etal. (1992). The couple was used as the unit ofanalysis.1 The first component in the analysisrepresented 53% of the total variance (see Table1). Only those subscales that had a loading of.71 (absolute value) or higher were used.Comrey and Lee (1992) suggested that variableloadings of .71 or greater are excellent indica-tors of the underlying component. The subscalesmeeting this criterion were Husband's Fondness/Affection, Wife's Fondness/Affection, Hus-band's Negativity, Wife's Negativity, Husband'sWe-ness, Wife's We-ness, Husband's Expansive-ness, Wife's Expansiveness, Husband's Disap-pointment and Disillusionment, Wife's Disap-pointment and Disillusionment, and Chaos. Inthe previous study of married couples withchildren, Glorifying the Struggle was among thesubscales that had a loading above .7; however,it had a lower loading in the analysis of thenewlywed data set (.53). Unlike our previousstudy, three subscales measuring the wife's rolein the marriage (Fondness/Affection, Negativity,
1 The couple was used as the unit of analysisbecause the Oral History Interview coding systemmeasures spouses' global perceptions of the relation-ship and of each other. Because the couple was theunit of analysis, there was no theoretical problemwith using variables from the interview indexing eachspouse and the couple as a whole in the principal-components analysis (R. Abbott, personal communi-cation, February 9, 1999).
PREDICTING MARITAL STABILITY 49
Table 1Principal-Components Analysis of Oral HistoryInterview Variables: Perceived Marital Bond
and Expansiveness) had loadings greater than .7.The variables for this principal componentrepresent the perceived marital bond present inthe relationship. The correlation matrix for thiscomponent's subscales is presented in Table 2.The other two principal components resultingfrom this analysis were volatility, with only theVolatility subscale loading above ,7 (volatilityloaded at .84), and gender stereotypy, on whichonly the Gender Stereotypy subscale loadedabove .7 (gender stereotypy loaded at .90).
Discriminant Construct Validation
Marital satisfaction. Marital satisfaction wastheorized to be a construct related to but notidentical to marital bonding. To examine thishypothesis, we calculated correlations betweenthe Oral History Interview variables and maritalsatisfaction scores from Time 1 (see Table 3).2
There were moderate but significant relation-ships between marital satisfaction and several ofthe perceived marital bond variables of the OralHistory Interview as well as the factor score forperceived marital bond (the regression approachwas used to calculate the perceived marital bondfactor score, as recommended by Tabachnick& Fidell, 1996). These moderate correlationsbetween marital satisfaction and perceivedmarital bond suggest that the two constructs arerelated but distinguishable from one another.
Social desirability. Social desirability wasconsidered a construct that would have no
relationship to marital bonding theoretically butcould offer a competing explanation for thecouples' behavior during the interview. As ameans of testing this assumption, the individualmarital bond variables loading greater than .7and the marital bond factor score were corre-lated with the spouses' scores on the Marlowe-Crowne Social Desirability Scale. As can beseen in Table 4, the relationships between thesevariables were nonsignificant, and the majorityof them had near-zero correlations. Chaos(—.11), Husband's Expansiveness (-.16), andHusband's Disappointment and Disillusionment(-.16) scores had the strongest correlations, buteven they were small and nonsignificant.
Predictive Validity Criterion
Two discriminant function analyses wereconducted to determine whether scores on theOral History Interview could correctly classifycouples as stable or divorced as of 1995 (Time2) and 1998 (Time 3). Seven of the 95 coupleshad divorced by 1995. The discriminant func-tion analysis of 1995 marital status included 6 ofthe 11 codes of the marital bonding factorloading higher than .70. Only 6 of the codeswere used because discriminant analysis re-quires that there be fewer predictor variablesthan the sample size of the smallest group(Tabachnick & Fidell, 1996). The predictorvariables were chosen on the basis of whichwere most effective in predicting membership inthe divorced group (Tabachnick & Fidell, 1996).The discriminant analysis of the 1995 maritalstatus data correctly categorized 87% of thecouples as married or divorced (see Table 5). Ofthe 7 couples who had divorced, 6 (86%) werecorrectly categorized. Only 11 of the 88 couplesstill married in 1995 were incorrectly classified.The canonical correlation for this analysis was.42, x2(6, N = 95) = 17.35, p < .01.
A second discriminant analysis was con-ducted with the 1998 marital status data todetermine whether the Oral History Interviewvariables had similar strength in predicting
2 This analysis involved a subsample of the original90 couples (n — 76) because not all of the coupleswho completed the Oral History Interview completedthe MAT and the Marlowe-Crowne Social Desirabil-ity Scale.
50 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
marital stability over a longer period of time. Atthis time point, 16 couples had divorced, so wewere able to use all of the variables loadingabove .7 on the perceived marital bond principalcomponent as predictor variables. When the 11codes of the marital bonding factor that loadedhigher than .70 were used, the discriminantanalysis was able to correctly categorize 81% ofthe couples as married or divorced (see Table 6).Of the 16 couples who had divorced, 13 (81%)were correctly classified. Only 15 of the 79couples still married in 1998 were incorrectlyclassified. The canonical correlation for thisanalysis was .53, x2(H, N = 95) = 28.23, p <.005. The standardized canonical discriminantfunctions for these analyses are also provided inTables 5 and 6.
Two discriminant function analyses usingboth the husbands' and wives' marital satisfac-tion scores (MAT paper-and-pencil versionscores; Locke & Wallace, 1959) at Time 1 topredict marital status at Time 2 and Time 3 wereconducted to compare the MAT's predictivestrength with that of the Oral History Interview.The results of the discriminant function analysisusing the Time 1 marital satisfaction scores topredict marital status at Time 2 are summarizedin Table 7. Marital satisfaction variables wereable to correctly classify 75% of the couples asmarried or divorced at Time 2. The results werestatistically significant, but the MAT was not assuccessful as the Oral History Interview incorrectly categorizing the divorced couples (4 ofthe 7 divorced couples were incorrectly catego-rized), nor was it as accurate in predicting whichcouples would remain married (77% correct forthe marital satisfaction measures, as comparedwith 88% for the Oral History Interview). Thecanonical correlation was .28, x2(2, N = 92) =7.08,/?>.05.
The discriminant analysis using Time 1marital satisfaction (MAT) scores to predict the1998 (Time 3) marital status of the newly wedcouples showed that the MAT did a poor job ofpredicting group membership (60% accuracy).Only 7 of the 16 divorced couples were correctlycategorized (44%). The MAT was also a poorindex for predicting those couples who wouldstill be married in 1998 (63% accuracy). Thecanonical correlation for this analysis was .11,X2(2, N=92) = 1.10, ns. These results aresummarized in Table 8.
:§
>-4
c
a
4*5as
1 1 1
I I I
9 (o B 9 ***
.> « .tS
a -a
"H W I
I
PREDICTING MARITAL STABILITY 51
Table 3Correlations of Oral History InterviewVariables With Locke-Wallace MaritalAdjustment Test Scores (n — 76)
A one-way analysis of variance was used tocompare the perceived marital bond factorscores for three groups of couples: couples stillmarried in 1998, couples who had divorced by1995, and couples who divorced after 1995 butbefore 1998. The perceived marital bond factorscore was calculated via the principal-compo-nents analysis previously described. Coupleswere divided into these three groups because
Table 4Correlations of Oral History InterviewVariables With Marlowe-Crvwne SocialDesirability Scale Scores (n = 76)
previous research (Gottman & Levenson, inpress) showed that couples divorcing early havevery negative emotional communication pat-terns, whereas couples who divorce later displayan absence of positive emotion in their commu-nication (but not significantly high levels ofnegative emotion). The present analysis allowedus to determine whether there was a similardifference in global marital perceptions betweencouples who remained married and couplesdivorcing after different lengths of marriage.The perceived marital bond factor score wasused because of the high correlation between theindividual variables making up this factor(Tabachnick&Fidell, 1996).
There was a significant linear trend inperceived marital bond among the three types ofcouples, F(2, 92) = 7.40, p < .001, with thosecouples still married in 1998 having the highestperceived marital bond scores and those coupleswho had divorced by 1995 having the lowestscores. A Scheffe" test was used to compare thedifferences among the three sets of couples(Keppel, 1982). The couples divorcing by 1995had a significantly lower marital bond score(M = -1.2, SD = 0.98) than the couples stillmarried in 1998 {M = 0.16, SD = 0.94;p < .05);however there were no significant differencesbetween the couples divorcing after 1995
52 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
Table 6Discriminant Function Analysis PredictingMarital Stability and Divorce in 1998 From theOral History Interview
{M = -0.33, SD = 0.94) and the other twogroups of couples.
Discussion
The perceptions newlywed spouses haveabout their partner and their marriage predict thestability of the marriage with 87% accuracy atthe 4-6-year point and do so with 81% accuracyat the 7-9-year point. These results lend supportto the theory that perceptual biases, especiallythe ways in which spouses selectively attend topositive or negative aspects of the marriage andtheir partner, shape the future marital path.
One of the questions raised by the presentstudy is the following: What does the perceivedmarital bond construct measure? We proposethat perceived marital bond is a measure ofspouses1 perceptual biases about each other andthe marriage. It is a construct that taps intomarital cognitions described by other investiga-tors, such as sentiment override (Notarius et al.,1989; Weiss, 1980) and selective attention tospousal behavior (Robinson & Price, 1980).Although it is related to the construct of maritalsatisfaction, as one might expect marital quality
to be associated with perceptions about therelationship, it is not identical to maritalsatisfaction (see Table 3). How, then, does it tapinto the perceptual biases that the spouses hold?
The Oral History Interview coding systemindexes a variety of behaviors of couples (e.g.,how spouses talk about each other and how theyinteract with each other in the interview) andwhat they selectively attend to in the past historyof the marriage, in the present state of themarriage, and in their partner's qualities andbehaviors. All of these elements making up theperceived marital bond seem to be intercon-nected by the positive-negative valence ofspouses' perceptual bias about the marriage. Theways in which they interact, talk about eachother, and describe the history of the marriageare strongly interrelated. This supports Finchamet al.'s (1990) thesis that how spouses rememberthe past corresponds with how they behavetoward one another in the present. As suggestedby Baucom and his colleagues (1989), individu-als pay selective attention to aspects of asituation that fit with a cognitive schema theyhold (i.e., the history of the marriage). Thesesubjective and biased perceptions of events areproblematic because people do not realize theyare operating with only a subset of informationabout those events. Baucom et al. suggested thatit is this process that results in present and future
Table 7Discriminant Function Analysis PredictingMarital Stability and Divorce in 1995From the Marital Adjustment Test (MAT)
Predicted groupmembership"
™°- Married Divorcedof
Actual group casesb n % n %
MarriedDivorced
Coefficient*Husband marital
satisfaction (MAT)Wife marital
satisfaction (MAT)
85 66 78 19 227 4 57 3 43
0.43*
0.69*
Percentage of cases correctly classified: 75%.bNumber of couples who had both husband and wifescores on the MAT at Time 1 (n - 92). Standardizedcanonical discriminant function.*p < .05.
PREDICTING MARITAL STABILITY 53
Table 8Discriminant Function Analysis PredictingMarital Stability and Divorce in 1998From the Marital Adjustment Test (MAT)
Predicted groupmembership4
No.of
Married Divorced
Actual group
MarriedDivorced
CoefficientHusband marital
satisfaction (MAT)Wife marital
satisfaction (MAT)
76 48 63 28 3716 9 56 7 44
1.26
-0.75
'Percentage of cases correctly classified: 59.8%.bNumber of couples who had both husband and wifescores on the MAT at Time 1 (n = 92). Standardizedcanonical discriminant function.
behavior being influenced by selective attentionand perceptual biases. The Oral History Inter-view Perceived Marital Bond scale appears tocapture this perceptual bias of couples. Itsstrength in predicting marital stability comesfrom the links marital cognition researchershave proposed between perceptual biases andfuture behavior within the marriage.
The predictive strength of the Oral HistoryInterview may also come from indexing bothwhat spouses report about the marriage and howthey interact with each other in the interview.This may be the advantage of using a "narra-tive" interview in which the spouses tell theirstory rather than using either a questionnaire orinterview with explicit questions about themarriage. VerofT, Sutherland, Chadiha, andOrtega (1993) suggested that direct, specificquestioning may reflect a person's social self-presentation in contrast to narratives, which areless inhibiting and more consistent with howpeople organize their experiences. Veroff and hisassociates also suggested that the "meaning"that spouses give to their relationship in thetelling of their story may be "diagnostic" ofhow they will function as a couple. Our researchwith the Oral History Interview supports thisperspective. The interview allows observationof how the couple operates as a unit andprovides insights about how their perceptions
and behaviors are indicative of what will takeplace in the marriage over time.
There is other evidence that global maritalperceptions shape spousal behavior. Notariusand his associates (1989) found that sentimentoverride, a perceptual filter spouses bring toevaluating the behavior of their partners, wasassociated with how wives judged the neutraland negative behaviors of their husbands.Notarius et al. (1989) found that distressedwives more negatively evaluated their husbands'neutral and negative behavior than nondis-tressed wives (as compared with objectiveratings of the husbands' behaviors by behavioralcoders). In our laboratory, Hawkins, Carrere,and Gottman (1999) found that the newlywedwives from the present study who scored low onthe Oral History Interview perceived maritalbond factor rated their husbands* anger, humor,and affection more negatively than did wivesscoring high on the factor. Shapiro, Gottman,and Carrere (2000) found that newlywed wives(from the same sample described in this article),who made the transition to parenthood witheither stable or increasing rates of maritalsatisfaction had higher scores on Oral HistoryInterview variables associated with perceivedmarital bond. Shapiro et al. theorized that theperceptions wives had about their marriages andspouses helped buffer them against the declinein marital satisfaction more typically observedduring the transition to parenthood. The resultsof these analyses suggest that spouses, particu-larly wives, are more likely to evaluate theirpartner's behavior on the basis of global maritalperceptions. These perceptual biases may helpin times of demanding marital transitions (e.g.,becoming parents) and influence day-to-dayinteractions. Our findings lend support totheories of causal relationships between cogni-tions and behavioral outcomes.
This study also replicates the findings ofBuehlman et al.'s (1992) analysis of the OralHistory Interview, this time with a sample ofnewlywed couples; as such, the results extendthe external validity of the original study'soutcomes. Furthermore, the present study clari-fied and strengthened the construct validity ofthe instrument by examining its internal struc-ture and its ability to be discriminated fromcompeting explanations of its construct (socialdesirability) and overlapping constructs (marital
54 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
satisfaction). The analyses in the present studyindicate that the predictive validity of theinstrument appears to taper off over time.However, the perceived marital bond scale ofthe Oral History Interview still provides moreaccuracy than self-reports of marital satisfactionin predicting the future course of marriage.
Buehlman et al. (1992) were able to predictmarital stability and divorce in a sample ofmarried couples with young children. Our goalwas to determine whether their results could begeneralized to couples at a different stage of themarital life cycle. The Oral History Interviewwas successful in identifying those newlywedcouples who would still be married or would bedivorced in 1995 (4 to 6 years after theirmarriage began). This population of couples hada comparatively short history of married experi-ence. Our concern was that the length of acouple's relationship might influence the OralHistory Interview measures of perceived maritalbond (e.g., "we-ness," expansiveness, andchaos). The ability of the interview to capturethose elements of the perceived marital bond innewlywed marriages that predict stability anddivorce is provocative. It is also worth notingthat the discriminant function analysis correctlyclassified six of the seven marriages that endedin divorce by Time 2 in this sample.
Overall, the results of this study strengthenour confidence in the construct validity of theOral History Interview. As before, we foundevidence in the Oral History Interview for theprincipal component of perceived marital bondin couples who were either low or high inexpressed fondness, "we-ness," expansiveness,negativity, and disappointment in the marriageas well as in the degree to which couplesdescribed their marriages as chaotic. In ournewlywed sample, the Oral History Interviewdimension of whether couples glorify thestruggle had a loading value too low to justifyinclusion into this principal component. Giventhe relatively short experience these couples hadin their marriage, it is understandable that avariable measuring the extent to which they hadstruggled through hard times would not be asuseful as other variables in predicting the healthof their marriage. In the present study, bothwives' and husbands' scores on the subscalesscored individually for each spouse (Fondness/Affection, We-ness, Expansiveness, Negativity,and Disappointment and Disillusionment) had
sufficiently high loading values to include in theperceived marital bond principal component.This is in contrast with the findings of Buehlmanet al. (1992) that only couple scores and husbandscores on the subscales had loading values highenough to include in the principal component.
The ability of the Oral History Interviewcoding system to predict the stability of therelationships did drop over time (from 87% inYears 4-6 of marriage to 81% by Years 7-9 ofmarriage). The analysis of variance comparingthose couples still married in 1998, thosecouples divorced by 1995, and those couplesmarried in 1995 but divorced in 1998 was a posthoc examination of the differences among thethree groups on the perceived marital bondfactor score. There was a significant linear trendin marital bond scores, with those couples stillmarried in 1998 having the highest perceivedscores and those couples who had divorced by1995 having the lowest scores. Those coupleswho had divorced by 1995 also had a signifi-cantly lower perceived marital bond score thanthe couples who remained married in 1998. Thelinear trend in the perceived marital bond scoresfor these three groups of couples would imply,as suggested by Gottman (1993, 1994), thatmarital instability is created when a couple'sperceptions about the marriage and about eachother exceed a particular negative threshold (thecouples who had divorced by 1995). However, itis less clear what factors led to maritaldissolution in the case of couples who divorcedafter 1995 but before 1998. Their perceivedmarital bond scores were lower than those of thecouples still married in 1998, but they were notsignificantly lower. One possibility is thetendency of global marital perceptions to changeor remain constant over time. Future researchcould examine the extent to which maritalperceptions are stable or fluid over time.
The present study has limitations. The coupleswho did not complete all of the components ofthe study (Oral History Interview, Time 1 MAT,and longitudinal telephone interviews on maritalstability) differed from the couples completingthese components, and these differences mayhave made the marriages of couples withincomplete data more unstable. There were nosignificant differences at Time 1 between thetwo groups on marital satisfaction, nor werethere differences in income. The couples withcomplete data were more likely to have lived
PREDICTING MARITAL STABILITY 55
together before marrying, a factor usuallyassociated with greater risk for divorce. Thecouples with complete data sets were signifi-cantly older (2 years on average) than the othercouples. Younger marriages are frequentlyassociated with greater rates of marital distress(Belsky & Rovine, 1990; Kurdek, 1991; Raschke,1987). Perceived marital bond may not have thesame predictive validity with the kinds ofcouples who did not complete the data.
Another limitation of the data is that we didnot look at change in marital satisfaction overtime as an outcome. Karney and Bradbury(1995) highlighted the importance of consider-ing both marital quality and stability as outcomemeasures in longitudinal studies of marriage.Although we were able to complete telephoneassessments of marital stability with 95 of thestudy couples, we had a lower success rate withmarital satisfaction measures. We decided tofocus solely on marital stability rather thanfurther reduce the sample size in the analyses.
One thought-provoking outcome of the studyis related to those couples who were incorrectlycategorized as divorced or married by thediscriminant analysis. The discriminant analysisinvolving the perceived marital bond variablesincorrectly classified 10 of the 83 couples stillmarried at Time 2 as divorced. This analysissuggests that thfe low perceived marital bond inthese 10 marriages could lead to maritaldissolution and that these marriages are at risk.However, 7 of these 10 couples were stillmarried in 1998. It would be interesting to knowhow the marital perceptions of these 7 couplesin the high-risk group changed or remainedstable in the years after the Oral HistoryInterview was conducted. As pointed out earlier,an important research direction would be todetermine how fluid or stable marital perceptualbiases are over time.
Our findings should be interpreted with somecaution because of the low number of divorcesthat occurred over the course of the study. Thisis one of the major challenges faced bylongitudinal studies of marital stability (Gott-man, 1994). This low base rate may be due to thelength of time it takes an unhappy marriage toprogress to divorce. The problem may also bedue to sampling issues; that is, those distressedcouples whose marriages are more likely to endin divorce may be less likely to participate in astudy of marriage.3
Conclusions and FutureResearch Directions
The present study offers support for causallinkages between marital cognitions and maritaloutcomes, particularly the impact of perceptualbiases and selective attention on the stability ofmarriage. The research presented here, incombination with research conducted by otherinvestigators on sentiment override (Notarius eta l , 1989; Weiss, 1980) and spouses' selectiveattention to each other's behavior (Robinson &Price, 1980), provides further clues as to howperceptual biases may influence the course ofmarriage. Specifically, they may influence mari-tal interactions in the present and subsequentlyresult in trajectories toward marital stability ordissolution. Future research could explore thecausal pathways between selective attentionresulting from perceptual biases about past andpresent events and future marital processes. Forexample, research in our laboratory suggeststhat Oral History Interview perceived maritalbond scores may be linked to spouses' rating ofeach other's behaviors (e.g., during conflictinteractions or during times of high duressresulting from demands from outside themarriage) and to subsequent changes in maritalquality and marital stability. In addition, ourresearch indicates that there is a drop in thepredictive validity of the Oral History Interviewperceived marital bond scale over time. It wouldbe useful to know how stable or fluid perceptualbiases are in marriage. It would also be helpful
3 When trying to predict a rare event such asdivorce, the goal is to correctly identify thosemarriages that will result in divorce because of thecost of being wrong. We would suggest that the mostproblematic event would be to predict that a couplewould remain married when in fact they divorced.This kind of event would prevent the possibility ofmarital intervention. An analogy would be a false-negative test for cancer, in which the actual presenceof cancer was missed by the diagnostic tool. In thisstudy, it was most important to correctly predict whowould divorce. It is far better to have a false-positivetest for cancer, in which cancer was diagnosed but didnot exist; likewise, it is better to have an instrumentsuch the Oral History Interview that might identifysome couples that appear to be on the road to divorce,when in fact their marriages will remain stable. Thereis less chance for harm in this case and everyopportunity to make a marital intervention that mightenhance the relationship.
56 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
to know how these perceptual biases are formedand what processes result in changes in thesebiases. Gottman (1994) suggested that percep-tions, behavior, and physiology are all inter-linked in marital processes and that majornegative changes in one of these three core triadindicators of marital health can lead to negativechanges in the other indicators, resulting in adecline in the strength of the marriage. Researchthat manipulates these different indicators ofmarital health could more specifically delineatethe relationships between perception, physiol-ogy, and behavior and subsequent maritaloutcomes.
Implications for Applicationand Public Policy
We have developed a model of research in ourlaboratory that moves from discerning the rulesthat underlie interpersonal relationships (throughobservation and measurement) to developingintervention research that tests those theoreticalrules of behavior. It is a research process that webelieve helps investigators apply and test theirbehavioral research findings. Our research withthe Oral History Interview is a case in point. Theability of the instrument to tap processespredictive of marital stability or dissolutionmakes it a powerful tool for identifying thosecouples on a trajectory toward divorce. Both thepresent study and the research by Buehlman andher associates pinpoint elements of a maritalrelationship that build and maintain the maritalbond (e.g., fondness and admiration for one'spartner and a sense of unity). Gottman recentlydeveloped a workshop for couples to help themstrengthen their marriage, based in part on whatthe Oral History Interview reveals about themarital bond (Gottman, 1999). During theworkshop, spouses build the marital bond bystrengthening their friendship, developing asense of unity, and creating a reservoir ofpositive regard for each other. We are currentlyconducting research in our laboratory to assesschanges in behavior, physiology, and percep-tions associated with couples' participation inthis workshop. We think it is important for thereto be a "marriage" between research on theprinciples of behavior and intervention research.Insights gained from empirical studies ofbehavior usually initiate ideas for interventionsthat are based on the results of such research.Results of intervention studies can generate both
broader applications of the intervention (e.g.,clinical training programs, community interven-tion projects, and legislative policy) and ques-tions about behavior that require further clarifi-cation through research. We think such amarriage between basic behavioral research andintervention research helps move research out ofthe university and into the community.
References
Baucom, D. H., Epstein, N., Savers, S., & Sher, T. G.(1989). The role of cognitions in marital relation-ships: Definitional, methodological, and conceptualissues. Journal of Consulting and Clinical Psychol-ogy, 57,31-38.
Behrens, B. C , & Sanders, M. R. (1994). Preventionof marital distress: Current issues in programmingand research. Behavior Change, 11, 82-93.
Belsky, J., & Rovine, M. (1990). Patterns of maritalchange across the transition to parenthood: Preg-nancy to three years postpartum. Journal ofMarriage and the Family, 52, 5-20.
Berkman, L. E, & Breslow, L. (1983). Health and theways of living: The Alameda County Study. NewYork: Oxford University Press.
Berkman, L. R, & Syme, S. L. (1979). Socialnetworks, host resistance, and mortality: A nine-year follow-up study of Alameda County residents.American Journal of Epidemiology, 109, 186-204.
Bloom, B., Asher, S., & White, S. (1978). Maritaldisruption as a stressor: A review and analysis.Psychological Bulletin, 85, 867-894.
Bradbury, T. M, & Fincham, R D. (1987). Affect andcognition in close relationships: Toward an integra-tive model. Cognition and Emotion, 1, 59-87.
Buehlman, K. X, & Gottman, J. M. (1996). The OralHistory Coding System. In J. Gottman (Ed.), Whatpredicts divorce? The measures (pp. OHI1-OHI118). HMsdale, NJ: Erlbaum.
Buehlman, K. T, Gottman, J. M , & Katz, L. F.(1992). How a couple views their past predicts theirfuture: Predicting divorce from an oral historyinterview. Journal of Family Psychology, 5, 295-318.
Burman, B., & Margolin, G. (1992). Analysis of theassociation between marital relationships and healthproblems. Psychological Bulletin, 112, 39-63.
Cherlin, A. (1992). Marriage, divorce, and remar-riage. Cambridge, MA: Harvard University Press.
City of Seattle Planning Commission. (1990). Popula-tion changes in Seattle 1980-1990, Seattle, WA:Author.
Comrey, A. L., & Lee, H. B. (1992). A first course infactor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.
Crowne, D. P., & Marlowe, D. (1960). Anew scale ofsocial desirability independent of psychopathology.Journal of Consulting Psychology, 24, 349-354.
PREDICTING MARITAL STABILITY 57
Epstein, N., Pretzer, J. L., & Fleming, B. (1987). Therole of cognitive appraisal in self-reports of maritalcommunication. Behavior Therapy, 18, 51-69.
Fincham, F. D., & Bradbury, T. N. (1987). The impactof attribution in marriage: A longitudinal analysis.Journal of Personality and Social Psychology, 53,510-517.
Fincham, F. D., Bradbury, T. N., & Scott, C K.(1990). Cognition in marriage. In F. D. Fincham &T. N. Bradbury (Eds.), The psychology of marriage(pp. 118-149). New York: Guilford Press.
Gottman, J. M. (1979). Marital interaction: Experi-mental investigations. New York: Academic Press.
Gottman, J. M. (1993). A theory of marital dissolutionand stability. Journal of Family Psychology, 7,57-75.
Gottman, J. M. (1994). What predicts divorce?: Therelationship between marital processes and maritaloutcomes. Hillsdale, NJ: Erlbaum.
Gottman, J. M. (1999). Couple's handbook: MarriageSurvival Kit Couple's Workshop. Seattle, WA:Gottman Institute.
Gottman, J. M., Coan, J., Carrere, S., & Swanson, C.(1998). Predicting marital happiness and stabilityfrom newlywed interactions. Journal of Marriageand the Family, 60,5-22.
Gottman, J. M., & Levenson, R. W. (in press). Therole of positive affect in long-term marital stability.Journal of Marriage and the Family.
Hawkins, M., Carrere, S., & Gottman, J. (1999).Marital interaction as observed objectively andperceived subjectively: A study of sentiment over-ride. Manuscript submitted for publication.
Hawley, D. R., & Olson, D. H. (1995). Enrichingnewlyweds: An evaluation of three enrichmentprograms. American Journal of Family Therapy, 23,129-147.
Holtzworth-Munroe, A., & Jacobson, N. S. (1985).Causal attributions of married couples: When dothey search for causes? What do they concludewhen they do? Journal of Personality and SocialPsychology, 48, 1398-1412.
Karney, B. R., & Bradbury, T. N. (1995). Thelongitudinal course of marital quality and stability:A review of theory, method, and research. Psycho-logical Bulletin, 118, 3-34.
Keppel, G. (1982). Design and analysis: A research-er's handbook (2nd ed.). Englewood Cliffs, NJ:Prentice Hall.
Krokoff, L. J. (1984a). The anatomy of blue-collarmarriages. Unpublished doctoral dissertation, Uni-versity of Illinois at Urbana-Champaign.
Krokoff, L. J. (1984b). A telephone version of theLocke-Wallace test of marital adjustment. Unpub-lished manuscript, University of Illinois at Urbana-Champaign.
Kurdek, L. A. (1991). Predictors of increases inmarital distress in newlywed couples: A 3-year
Kyle, S. O., & Falbo, T. (1985). Relationship betweenmarital stress and attributional preferences for ownand spouse behavior. Journal of Social and ClinicalPsychology, 3, 339-351.
Locke, H. J., & Wallace, K. M. (1959). Short maritaladjustment and prediction tests: Their reliability andvalidity. Marriage and Family Living, 21, 251-255.
Markman, H. J. (1981). Prediction of marital distress:A five-year follow-up. Journal of Consulting andClinical Psychology, 49, 760-762.
Markman, H. J., Floyd, F. J., Stanley, S. M., &Jamieson, K. (1984). A cognitive-behavioral pro-gram for the prevention of marital and familydistress: Issues in program development and deliv-ery. In K. Hahlweg & N. Jacobson (Eds.), Maritalinteraction (pp. 396-428). New York: GuilfordPress.
Markman, H. J., Floyd, F. J., Stanley, S. M., &Storaasli, R. D. (1988). Prevention of maritaldistress: A longitudinal investigation. Journal ofConsulting and Clinical Psychology, 56, 210-217.
Martin, T. C , & Bumpass, L. L. (1989). Recent trendsin marital disruption. Demography, 26, 37-51.
National Center for Health Statistics. (1991). Ad-vance report of final marriage statistics, 1988(Monthly Vital Statistics Report 39). Hyattsville,MD: Public Health Service.
Noller, P., & Feeney, J. A. (1998). Communication inearly marriage: Response to conflict, nonverbalaccuracy, and conversational patterns. In T. N.Bradbury (Ed.), The developmental course ofmarital dysfunction (pp. 11-43). New York: Cam-bridge University Press.
Notarius, C. I., Benson, P. R., Sloane, X>., Vanzetti,N. A., & Hornyak, L. M. (1989). Exploring theinterface between perception and behavior: Ananalysis of marital interaction in distressed andnondistressed couples. Behavioral Assessment, 11,39-64.
Raschke, H. J. (1987). Divorce. In M. B. Sussman &S. K. Steinmetz (Eds.), Handbook of marriage andthe family (pp. 597-624). New York: Plenum.
Robinson, E. A., & Price, M. G. (1980). Pleasurablebehavior in marital interaction: An observationalstudy. Journal of Consulting and Clinical Psychol-ogy,48, 117-118.
Shapiro, A. F., Gottman, J. M., & Carrere, S. (1999).The baby and the marriage: Identifying factors thatbuffer against decline in marital satisfaction afterthe baby arrives. Journal of Family Psychology, 14,59-70.
Tabachnick, B. G., & Fidell, L. S. (1996). Usingmultivariate statistics (3rd ed.). New York: Harper-Collins.
Thompson, J. S., & Snyder, D. K. (1986). Attributiontheory in intimate relationships: A methodological
58 CARRERE, BUEHLMAN, GOTTMAN, COAN, AND RUCKSTUHL
review. American Journal of Family Therapy, 14, intervention. In J. P. Vincent (Ed.), Advances in123-138. family intervention, assessment and theory (Vol. 1,
Veroff, J., Sutherland, L., Chadiha, L., & Ortega, R. pp. 229-271). Greenwich, CT: JAI Press.(1993). Newlyweds tell their stories: A narrativemethod for assessing marital experiences. Journalof Social and Personal Relationships, 10, 437-457. Received May 27, 1997
Weiss, R. L. (1980). Strategic behavioral marital Revision received November 5, 1999therapy: Toward a model for assessment and Accepted November 5, 1999 •
