Outline of Predicting Divorce at Marital Therapy Intake: A Preliminary Model

By: D.R. Crane, J.N. Soderquist, & R.L. Frank

I.                    Purpose: to predict divorce in a martial-distressed and therapy-seeking population, using data from two forms of marital assessment.

A.                 MSI – Marital Status Inventory (see below III.B.1.)

B.                 MAT – Marital Adjustment Test (see below III.B.2.)

II.                 Introduction.

A.                 What predisposes individuals and couples to divorce?

1.       Prior studies have yielded complicated models that are not generalizable.

a.       Data from huge national surveys has been, for the most part, cross-sectional (rather than longitudinal) and nearly all respondents were women.

b.       Longitudinal studies are expensive and not pragmatic (distressed couples are not always willing and able to complete assessments).

B.                 This study focuses on clinically distressed couples in marital therapy.

1.                   What distinguishes couples that stayed married from those that got divorced?

III.               Method.

A.                 Subjects – 235 couples receiving therapy from BYU’s MFT Clinic.  Data was collected from their closed files.

B.                 Assessment Instruments.

1.                   MSI (see above I.A.) is a true/false test that reliably and validly measures divorce potential, partially based on each partner’s distress.

2.                   MAT (see above I.B.) measures overall marital quality and accurately and consistently differentiates between distressed and not distressed couples.

C.                 Other variables helped to form the model – age at marriage, children, religion, employment, education, and prior therapy.

IV.              Results.

A.                 Total variance in divorce outcome was between 21% and 28% for the various models.

1.                   This number is low, but better than the findings of previous research at 10% variance.

B.                 MSI scores were significantly different for married and divorced couples.

V.                 Discussion.

A.                 Variables found significant in this study are similar to previous studies’ findings.

1.                   Note contrary findings:

a.       This study found that women who married at an older age were more likely to get divorced.

b.       Marital quality (as tested by the MAT) was not found to have a strong or direct relationship to marital stability.

i.                     MAT (measuring marital quality) may be unable to detect small differences in marital quality between two groups who both had low scores (i.e. the basement effect).

B.                 The best model includes the individual contributions of both partners, but studied as mutually influential variables.

C.                 Wives’ distress level (measured by the MSI) is the most important predictor.

1.                   This is consistent with previous research that showed that a woman’s willingness to consider divorce is made in the face of graver consequences than those that a man usually faces after divorce.  Thus, when the wife is considering divorce, her feelings and her decisions should not be taken lightly.

D.                 In every case (of every couple), the best model (see above V.B.) was significantly more accurate than chance at assigning marital status.

E.                  Utility of this study: to design and apply therapy for couples, especially those at high-risk for divorce as measured by the MSI.

Liz Price

Gina Bright

Alana Williams

Jolie Randal


Russell Crane, D., Soderquist, J.N., Frank, R.L. (1995).  Predicting divorce at therapy

intake: a preliminary model, The American Journal of Family Therapy, 23:3,



Over the years, a significant amount of literature has been collected regarding the variables that seem to predict which individuals and couples are most likely to divorce.  The majority of these studies used large national surveys. (i.e. the US census and the General Social Surveys conducted by the National Opinion Research Center (NORC)).  Other methods for collecting data on divorce related topics have included live observation, interviews, telephone surveys, or return mail questionnaires.  However, there were only a few people who chose to participate or respond creating a relatively small sample and significant limitations. 

The first limitation mentioned was that most of these studies are cross sectional rather than longitudinal.  This causes an overrepresentation of women and divorced individuals.  “In addition, such surveys (because they yield what amounts to a “snapshot” of current marital status) focus primarily on factors, such as educational level, that are measured at the time of the survey, even though marital dissolution may have occurred much earlier.  Thus, the temporal order of the variables is the opposite of what it should be for inferring the effect of such factors upon marital outcomes” (Russell Crane, Soderquist & Frank, 1995, pp.228). The second limitation is that the data was gathered from only female subjects.  Very few research projects have included men or couples within their sample. This gives a very incomplete picture of marriage.  The last limitation noted was that none of the studies focused on or used clinically distressed or treatment seeking individuals or couples.  The information gathered from the general population may or may not be useful to clinically distressed individuals in couples therapy. 

“The purpose of this study was to begin to identify the characteristics that distinguished couples who remained married from couples who divorced and to predict divorce in a clinical sample.  Given an accurate predictive model, clinicians would be able to identify which couples are likely to divorce and which are not.  They would also have a greater ability to (a) apply appropriate treatment to couples in both categories, and (b) design treatments for those at higher risk for divorce ”(Russell Crane, Soderquist & Frank, 1995, pp.229).   

The subjects within this study included 235 couples seeking therapy from the Brigham Young University Marriage and Family Clinic.  The assessment instruments used included the Marital Adjustment Test (MAT) (Locke &Wallace, 1959) and the Marital Status Inventory (MSI) (Weiss & Cerreto, 1980).  The MAT is used to asses marital quality and is internally consistent.  It also accurately differentiates between distressed and non-distressed marital relationships. The MSI is used to measure divorce potential.  It is a 14 item, true / false instrument.  It is a reliable and valid test.  The Spearman-Brown split-half reliability for the MSI was found to be .86 and .87 respectively.   The MSI was also found to have discriminant validity.  The variables included where obtain from clinical records and included age at marriage, psychological distress, number of children, number of preschool children, remarriage, number of stepchildren, religion (all were members of the Church of Jesus Christ of Latter Day Saints), wife’s employment, education, and prior therapy experience.  Using closed files from the clinic records, 235 couples were randomly chosen to participate within this study. Out of this sample 171 remained married, while 64 were divorced.  Random sampling was used to create equal groups of married and divorced men and women.  Another group of 64 was formed and set aside for a cross-validation analysis.  The units of analysis were husbands, wives, and couples.

The study used four procedures in order to create the prediction models.  The first procedure used was the Maximum R-square Regression (MAXR).  The MAXR was used to assure that the maxim amount of variance was accounted for, while reducing the amount of variables within the study.  The second procedure used was the Stepwise Discriminant Analysis (SDA).  This was used for further elimination of variables.  The Descriptive Discriminant Analysis (DDA) was then used to enter the data that was selected and ordered by the SDA to indicate the differences between married and divorced husbands and wives. The last procedure was the Predictive Discriminant Analysis (PDA).  The variables found within the DDA were then entered into the PDA to see whether or not they could classify subjects as to future marital status.  The Cross-Validation Sample was used to compare and check the accuracy of the models in predicting future marital status.  

This study found that overall the couples model might be more useful in understanding divorce.  “It may be that the best picture of the divorce process is in the individual contributions of both partners, but studied as mutually influential variables”  (Russell Crane, Soderquist & Frank, 1995, pp. 234).  It was also found that the distress level of the wife is a more important predictor of divorce than the husband’s level of distress.  This finding is consistent with other research within this field.  Models within the PDA were also able to predict marital status and divorce with a significant level of accuracy.  However, the researchers suggest that the models be tested against a new sample of married and divorced subjects for further accuracy.  Generally, the most important variable within this study was the distress of the wives.  These scores were found using the Marital Satisfaction Inventory (MAI). 

Limitations to this study include that there could be more information about the subjects prior to testing (i.e. premarital pregnancy and education) and that the subjects were all members of the same religious group, the LDS church.  However, “since religious homogeneity was previously found to be correlated to divorce (Chan & Heaton, 1989; Corley & Woods, 1991), using a sample from a single religious denomination while achieving results similar to other studies may identify variables that are more important than religion alone” (Russell Crane, Soderquist & Frank, 1995, pp. 235).