Predicting Premarital Relationship Stability: A Methodological Refinement

Author(s): Sally A. Lloyd, Rodney M. Cate and June M. Henton

Source: Journal of Marriage and Family, Vol. 46, No. 1 (Feb., 1984), pp. 71-76
Published by: National Council on Family Relations
Stable URL: https://www.jstor.org/stable/351865
Accessed: 30-11-2018 10:25 UTC

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/351865?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

National Council on Family Relations is collaborating with JSTOR to digitize, preserve and
extend access to Journal of Marriage and Family

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
 Predicting Premarital Relationship Stability:
A Methodological Refinement

SALLY A. LLOYD
University of Utah

RODNEY M. CA TE*
JUNE M. HENTON*
Oregon State University

Predicting premarital relationship stability has been done in a variety of empirical

gations. To date, none of these studies have attempted to controlfor the length of
tionship at initial contact. Thus, such studies may have confounded measures of
ment or commitment with the length of the relationship. The purpose of this stud
predict relationship satisfaction while statistically controlling for the length of the
ship. One hundred thirty-one participants completed assessments of involvement
level, comparison levelfor alternatives, satisfaction, and chance of marriage. Rel
stability was assessed three months and seven months later. Through the use of an
covariance, the variance due to the length of the relationship at the initial con
removed from each of the predictor variables before the main effect (relationship
was assessed. After controlling for length, results indicated that involvement, rewa
and chance of marriage were all significant predictors of stability both at the three
follow-up and at the seven-month follow-up.

Predicting premarital relationship stability have beenhas

drawn concerning the factors that af-
been of interest to researchers for several decades.
fect movement to marriage. Kerckhoff and Davis
The first study of relationship stability was (1962) postulated that there are filtering factors
Burgess and Wallin's (1953) classic study of that operate in premarital relationships. Sup-
engagement and marriage. In the early 1960s this posedly, early in the relationship social status
area of study was entitled "mate selection" and variables are important in predicting "courtship
included such papers as Kerckhoff and Davis's progress," followed later by value consensus and
Filter Theory (1962). By the mid-1970s the empha- need complementarity. Hill et al. (1976) found
sis had shifted from mate selection to attempts to closeness, probability of marriage, amount of
delineate variables that would discriminate be- love, and equality of involvement to be significant
tween couples who remain together and couples predictors of relationship stability.
who break apart (cf. Hill, Rubin and Peplau, Although these studies differ in their content
1976). From such studies a variety of conclusions (i.e., in the variables measured as predictors of
stability), they all share basically similar
methodologies. Each of these studies gathered in-
Data for the study were collected in fall 1979 and spring
1980. The authors would like to thank Dr. Arthur W.
itial information from a group of ongoing couples
Avery for his helpful comments on an earlier draftand of subsequently measured stability or "court-
this paper.
ship progress" by contacting the couples between
Department of Family and Consumer Studies, Universi- 8 and 24 months later to determine which couples
ty of Utah, Salt Lake City, UT 84112. were still intact. However, the use of couples in
ongoing relationships as subjects presents a prob-
*Department of Human Development and Family lem in the interpretation of the results of these
Studies, Oregon State University, Corvallis, OR 97331. studies. For example, at the beginning of the Hill

February 1984 JOURNAL OF MARRIAGE AND THE FAMILY 71

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
et al. (1976) Boston Couples Study, one-third of searchers could place increased confidence in their
the couples had been dating less than five months, results; that is, they could be more confident that
one-third had been dating between five and ten any differences between couples who remain to-
months, and one-third had been dating more than gether versus couples who break apart are not
ten months. It appears logical that the length of merely a function of differences in length of rela-
the relationship at the initial contact might be a tionship at the initial contact of the participants.
confounding factor, especially when one con- Such statistical control would be particularly im-
siders that couples who have been dating only a portant to consider when studying any variable
few months are being compared with couples who that is presumed to be significantly linearly related
have been dating a year or more. Indeed, the issue to the length of the relationship (e.g., involvement
of length of relationship is salient in light of the or chance of marriage).
fact that length of relationship at the initial con-
tact was a significant discriminator between METHOD
couples who stayed together (mean length of rela-
Selection of Variables
tionship at the initial contact was 12.9 months)
and couples who broke up (mean length of rela- Seven predictor variables were examined in
tionship at the initial contact was 9.9 months) present study:' involvement, reward level for
(Hill et al., 1976). Just as studies of marital stabili- self, reward level for the partner, compari
ty point out that length of relationship is negative- level for alternatives (CLalt) for the self, co
ly related to divorce proneness, so Hill, Rubin and parison level for alternatives for the partn
Peplau (1976) have demonstrated length of rela- chance of marriage, and satisfaction. These se
tionship to be negatively related to premarital variables were chosen on the basis of their
break-up. previous use in a study of commitment in pre-
Kerckhoff and Davis (1962) attempted to con- marital relationships (Rusbult, 1980). Rusbult
trol for length of relationship in their study of (1980) found that investment size, alternative
"progress towards permanence" by dividing their value, satisfaction, and reward value were all sig-
sample of couples into "long-term" (dated more nificantly related to commitment to the relation-
than 18 months) and "short-term" (dated less ship. We chose to follow this perspective as well;
than 18 months) subgroups. Although they found to this end, we assessed involvement (which was
differences between the two groups as to what chosen to represent an investment in the relation-
best predicted courtship progress (value consensus ship), reward level, satisfaction, and CLalt. In ad-
versus need complementarity), their division of dition, in keeping with Levinger and Huesmann's
couples into long- and short-term subgroups was (1980) incremental exchange concepts, we in-
somewhat arbitrary (they simply divided the sam- cluded an assessment of the anticipated future po-
ple roughly in half). Furthermore, a major tential of the relationship (i.e., chance of mar-
replication conducted by Levinger, Senn and riage). Rusbult (1980) demonstrated that invest-
Jorgensen (1970) failed to support the findings ofment, rewards, alternatives, and satisfaction are
the Kerckhoff and Davis (1962) study, includingrelated to commitment to the relationship; our
this interaction between length of relationship andquestion of interest was whether these variables
need complementarity/value consensus. and anticipated future potential were related to
In summary, then, a major weakness of longi-the actual continuance of the relationship. Addi-
tudinal studies of premarital relationship stabilitytionally, the seven variables were chosen on the
is their failure to control for the length of the rela- basis of their potential association with the length
tionship at the initial contact of the participants. of the relationship.
It would appear that the ideal means of remedying
Participants
this problem would be to follow each premarital
couple from the very beginning of the relation- Participants were recruited from introductory
ship. Such a study, however, would be difficult Human Development classes at a large North-
because of the time and cost involved. A viable western university. The classes chosen for recruit-
alternative to this approach would be to statisti- ment were elective classes that contained a repre-
cally control for the length of the relationship sentative cross-section of students. All partici-
through the use of analysis of covariance and pants were volunteers and included only those in-
thereby remove the variance in the predictor vari- dividuals who were currently involved in a rela-
able that is due to its association with length. Such tionship.2 A total of 131 (36 males, 95 females)
a statistical control represents a refinement of the participants completed the questionnaires at the
existing method for prediction of premarital rela- initial contact. Mean age of the participants was
tionship stability. By using this method, re- 19.4. Class standing of the participants was as

72 JOURNAL OF MARRIAGE AND THE FAMILY February 1984

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
follows: 34 (26%) freshman, 58 (44%) sopho- angry, and guilty they felt on a scale of I (not at
mores, 21 (16%) juniors, 18 (14%) seniors. all) to 4 (very much). The total satisfaction score
was calculated by summing the content and happy
Measures
scores and subtracting the angry and guilty scores.
Involvement. Involvement was measured with The higher the score, the more content individuals
were with their relationships.
the 10-item Levinger, Rands and Talaber (1977)
involvement scale. This scale measures the dimen-
Procedure
sions of uniqueness, disclosure, outcome corre-
spondence, and emotional caring. Each was ac- During a regularly scheduled class period, par-
companied by a Likert scale that ranged from ticipants
1 were asked to complete a questionnaire
(not true at all of our relationship) to 7 (very consisting
true of the involvement, CLalt, reward
of our relationship). level, chance of marriage, and satisfaction
Reward level. Reward level was measured by measures.
a In addition, the questionnaire included
demographic assessments and questions on cur-
seven-item Likert scale constructed by the present
researchers. The items assessed amount of rent relationship status (casually dating, seriously
rewards in the six resource areas (love, status, dating, ser-
engaged, cohabiting) and length of rela-
vices, goods, money, information) outlined tionship. by
Foa and Foa (1974) and in the area of sexuality. Following this initial data collection, there were
Each question was accompanied by a Likert two subsequent
scale follow-ups. The first follow-up
that ranged from 1 (unrewarding) to 9 (extremely occurred three months after the initial data collec-
rewarding). Cronbach's alpha for the scale tion andwaswas conducted during a regularly
.90. Each participant completed the reward scheduled class period. Participants were asked to
measure twice: first to assess reward level for the indicate in writing their current relationship status
self, and second to assess reward level for the (casually dating, seriously dating, engaged, mar-
partner. ried, cohabiting or separated). In the second
Comparison level for alternatives (CLalt). follow-up, which took place seven months after
Comparison level for alternatives was assessed the initial contact, the subjects were contacted by
through two questions. These questions have been telephone. At this time participants again indi-
outlined by Hatfield, Utne and Traupmann cated their current relationship status.
(1979). Participants were first asked, "If you
found yourself unattached again, for whatever RESULTS
reason, and wanted to find a new partner, how
Three-month Follow-up
easy or difficult would that be?" This question
was accompanied by a Likert scale that ranged Before proceeding with the analysis, partici-
from 1 (very difficult) to 9 (very easy). Par- pants who were contacted at the three-month
ticipants then were asked, "How do you think follow-up were compared with participants wh
you would fare? That is, how would the new part- were not contacted. Results revealed no signif
ner compare with your present partner?" This cant differences between the two groups on any
question was accompanied by a Likert scale that the predictor variables or in length of relationsh
ranged from 1 (worse) to 9 (better). The two at initial contact.
scores were summed to yield a total score. Again, At the three-month follow-up, 111 (85%; 31
each participant completed the comparison level males, 80 females) of the original participants
for alternatives measure twice: first to assess were recontacted to determine their relationship
CLalt for the self, and second to assess CLaltstatus.
for At that time, 82 of the participants were
the partner. still dating their partners, while 29 of the partici-
Chance of marriage. Chance of marriage was pants had broken up with their partners. The
mean length of relationship at the initial contact
assessed by asking the participants, "At this point
in time, what do you feel the chance is of your for these 111 participants was 18 months (ranging
relationship ending in marriage?" Participants in-
from 1 month to 8 years). For the analysis of the
dicated the chance of marriage as a percentage data, participants were grouped into two cate-
from 0 to 100. gories: intact versus separated.
Satisfaction. Satisfaction with the relationshipA series of seven 2 x 2 (relationship status by
was assessed with the Austin Contentment/Dis- gender) analyses of covariance were run. Length
tress measure (Austin, 1974). This measure asksof relationship in months at the initial contact was
the covariate for each of these analyses. Analysis
individuals to think about their relationships and
of covariance was chosen to allow the variance ac-
then assess how they felt about the relationships.
Participants indicated how happy, content, counted for by length of relationship to be re-

February 1984 JOURNAL OF MARRIAGE AND THE FAMILY 73

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
TABLE 1. MEANS AND F VALUES OF VARIABLES AT THREE-MONTH FOLLOW-UP

Means F Values

Variables Together Separated Covariate (Length) Main Effects

Involvement 59.04 50.03 12.55*** 11.82***
Reward level-self 48.65 38.24 5.56* 19.56***
Reward level-partner 48.59 38.31 4.44* 21.87***
CLalt-self 10.11 11.48 1.59 2.94
CLalt-partner 10.72 12.00 5.10* 3.15
Chance of marriage 62.00 38.30 22.98*** 7.92**
Satisfaction 4.17 3.45 9.34** 2.06

*p < .05.
**p < .01.
***p < .001.

moved from each of the seven predictorintact at the three-month follow-up were recon-
variables.
A general linear models approach to ACOVAtacted towas
assess their relationship status. Partici-
pants whoby
used; the covariate was analyzed first, followed were contacted at the seven-month
follow-up
main effects and interaction. No significant main were compared with participants who
effects for gender or interactions were were
found.notFor
contacted. Results revealed no signifi-
simplicity of presentation of the results, only
cant the
differences between the two groups on any of
main effects for relationship status and the the F
predictor variables or on length of relationship
values for the covariate are presented at ininitial
Tablecontact.
1 Of the 48 participants con-
and the text. tacted, 28 relationships were still intact, while 20
The covariate of length was significantly relatedrelationships had broken up.
to involvement, reward level for self, reward level A series of seven one-way analyses of covari-
for partner, comparison level for alternatives for ance were run (due to the small number of males
partner, chance of marriage, and satisfaction (see[n = 6], gender was omitted from the analysis of
Table 1). There were significant main effects for the data). The covariate of length was signifi-
involvement, reward level for the self, reward cantly related to involvement, reward level for the
level for the partner, and chance of marriage. self, reward level for the partner, chance of mar-
These results indicate that, after controlling for riage, and satisfaction (see Table 2). Length was
the length of the relationship at initial contact, not significantly related to comparison level for
participants who were still together versus those alternatives for self or for partner. There were sig-
who had separated had been significantly higher nificant main effects for involvement, reward
in involvement, reward level for self, reward level level for self, reward level for partner, and chance
for the partner, and chance of marriage at the in- of marriage. The main effects for comparison
itial contact. Main effects for comparison level level for alternatives for self and partner, and for
for alternatives for the self, comparison level for satisfaction, were not significant. These results in-
alternatives for the partner, and satisfaction were dicate that, after controlling for the length of the
not significant. relationship at the time of the initial contact, par-
ticipants who were still together with their part-
Seven-month Follow-up
ners versus participants who had broken up with
Seven months after the initial contact, 48 (59%) their partners had been significantly higher in
of the participants whose relationships had been levels of involvement, reward level for the self,

TABLE 2. MEANS AND F VALUES OF VARIABLES AT SEVEN-MONTH FOLLOW-UP

Means F Values

Variables Together Separated Covariate (Length) Main Effects

Involvement 63.32 52.50 6.33* 9.59**
Reward level-self 52.35 39.72 4.48* 10.71**
Reward level-partner 52.27 41.28 5.05" 9.23**
CLalt-self 10.62 12.56 .00 2.76
CLalt-partner 10.15 11.89 .11 2.99
Chance of marriage 72.30 36.30 9.80** 11.70***
Satisfaction 4.62 3.78 4.54* .97

*p < .05.
**p < .01.
***p < .001.

74 JOURNAL OF MARRIAGE AND THE FAMILY February 1984

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
reward level for the partner, and chance of mar- lack of discriminatory power lies in the nature of
riage at the initial contact. the satisfaction variable. Satisfaction may tend to
represent a momentary assessment of feelings
DISCUSSION about the relationship (such a statement applies
also to comparison level for alternatives). In con-
The primary goal of this study was totrast,
incorpo-
reward level and involvement level may take
rate statistical control for length of relationship
into accountat
the history of the relationship, while
initial contact into the prediction of premarital chance of marriage takes into account the future
relationship stability. For both data collection potential of the relationship. Thus, the contrast
points, length of relationship was significantly between there-variables that predicted stability and
lated to involvement, reward level forthose thethat
self,
did not may lie in the difference be-
reward level for the partner, chance oftween marriage,
a momentary assessment of the relationship
and satisfaction. The use of length of relationship and an assessment of the past and anticipated
as a covariate seems to have been warranted, future then,
investment in the relationship.
in light of the fact that length of relationship The at
utilization
in- of analysis of covariance in the
itial contact was able to account for a significant present study has provided us with more informa-
percentage of the variance in each of these tion predic-
on the association between length of relation-
tor variables. ship at initial contact and relationship stability
It is indeed interesting to note that involvement,than would a traditional analysis of variance ap-
reward level for self, reward level for partner, and proach. Analysis of covariance allowed us to ex-
chance of marriage were all significant predictorsamine a continuous concomitant variable (length
of relationship stability at the three-month follow-of relationship at initial contact) with a categori-
up and at the seven-month follow-up, even after cal independent variable (relationship status:
the variation due to length of relationship was sta- together versus separated). Analysis of covariance
tistically removed. Perhaps, then, it is reasonable also can be done with more than one covariate.
to assert that each of these four measures repre- Thus, it is possible to statistically control for the
sents more than a simple temporal marker for the effects of several relevant concomitant variables
depth of involvement in the relationship. In ex-simultaneously. For example, in the study of rela-
change terms, involvement, reward levels for self tionship development, other relevant concomitant
and for partner, and chance of marriage may be variables may be age or number of previous
representative of both an investment in the rela-serious relationships. It is important to keep in
tionship and an assessment of its future potentialmind, however, that the choice of concomitant
which, irrespective of length, predicts who will re-variable should make conceptual and theoretical
main in the relationship and who will break apart.sense; that is, statistical control of a continuous
Such a conclusion is in keeping with Rusbult'svariable through analysis of covariance should
(1980) finding on commitment to a relationship.not be arbitrary. Ultimately, a sound rationale is
Rusbult suggested that the size of the investmentnecessary in choosing the covariate.
in the relationship is a good indicator of the
stability of the relationship; our findings clearly
support such a conclusion. Interestingly, in
Rusbult's (1980) study commitment was neither a
function of satisfaction nor a function of simple
analysis of alternatives. Again, our findings are
parallel; we did not find that satisfaction or com-
parison level for alternatives predicted stability in
premarital relationships.
At first glance the fact that satisfaction with the
FOOTNOTES
relationship did not predict stability in premarital
relationships may appear surprising. Satisfaction, 1. The term "predictor variable" is used in this pap
which is seen as an indicator of the quality of the to denote what is commonly called the "dependen
relationship, is one of the most widely studied variable" in analysis of variance. This was done so as
aspects of the marital relationship. As with to conform to common use of the word "prediction
marital couples, it might be logical to assume that in past studies of premarital relationship stability.

premarital relationships that are high in quality 2. Only one member of a couple was surveyed in th
would also exhibit high stability. In the present study. Consequently, we cannot determine whethe
study, however, satisfaction did not predict prior length of relationship operated differently fo
stability of premarital relationships. Perhaps this each member of the dyad.

February 1984 JOURNAL OF MARRIAGE AND THE FAMILY 75

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms
 REFERENCES in mate selection." American Sociological
Review 27(June):295-303.
Levinger, G. and Huesmann, L. R.
Austin, W. G. 1980 "An 'incremental exchange' perspective on the
1974 "Studies in 'equity with the world': a new ap- pair relationship: interpersonal reward and
plication of equity theory." Unpublished doc- level of involvement." Pp. 165-187 in K. J.
toral dissertation, University of Wisconsin. Gergen, M. S. Greenberg and R. H. Willis
Burgess, E. W. and Wallin, P. (Eds.), Social Exchange: Advances in Theory
1953 Engagement and Marriage. Philadelphia:Lip- and Research. New York:Plenum.
pincott. Levinger, G., Rands, M. and Talaber, R.
Foa, U. G. and Foa, E. G. 1977 "The assessment of rewardingness in close and
1974 Societal structures of the mind. Springfield, casual pair relationships." Technical report
IL:Charles C. Thomas. submitted to the National Science Foundation
Hatfield, E., Utne, M. K. and Traupmann, J. (June).
1979 "Equity theory and intimate relationships." Levinger, G., Senn, D. J. and Jorgensen, B.
Pp. 99-134 in R. L. Burgess and T. L. Huston 1970 "Progress towards permanence in courtship: a
(Eds.), Social Exchange in Developing Rela- test of the Kerckhoff-Davis hypothesis."
tionships. New York:Academic Press. Sociometry 33(4):427-433.
Hill, C. T., Rubin, Z. and Peplau, L. A. Rusbult, C.
1976 "Breakups before marriage: the end of 103 af- 1980 "Commitment and satisfaction in romantic
fairs." Journal of Social Issues 32(1):147-168. association: a test of the investment model."
Kerckhoff, A. C. and Davis, K. E. Journal of Experimental Social Psychology 16
1962 "Value consensus and need complementarity (March): 172-186.

76 JOURNAL OF MARRIAGE AND THE FAMILY February 1984

This content downloaded from 193.140.161.80 on Fri, 30 Nov 2018 10:25:42 UTC
All use subject to https://about.jstor.org/terms