The Spouse Situation Inventory: A Role-Play Measure of Coping Skills in Women With Alcoholic Partners

Abstract

This article reports on the psychometric evaluation of the Spouse Situation Inventory (SSI), a role-play measure of coping skills in women with alcoholic partners. The study examined the generalizability, alternate form reliability, test–retest reliability, and construct validity of the measure in 472 women from both treatment and nontreatment populations. The SSI had acceptable generalizability and reliability. SSI performance also had predicted relationships with measures of general escape coping, alcohol-related coping behaviors, the woman’s drinking, the partner’s drinking in the treatment group, and the partner’s problem recognition in the nontreatment group. The SSI shows promise as a reliable and valid measure of coping skills in this population and has direct implications for development and evaluation of skill training programs.

Caregiver distress most often is conceptualized from within family stress and family interactional theories (e.g., Hobfoll & Spielberger, 1992). The distress experienced is the result of the family’s frustrated efforts to cope with altered roles, economic responsibilities, direct caretaking, and the disruptive behavior of the impaired person. The caregiver’s ability to resist stress-related problems is viewed, in part, as a function of the social supports available and his or her own coping skills.

The depression and other psychological distress frequently seen among individuals with alcoholic partners also have been conceptualized within family stress (e.g., Moos, Finney, & Cronkite, 1990) and interactional models (e.g., Steinglass, 1987). In fact, the processes of stress and coping in this population share many commonalities with those of individuals caring for partners impaired by a wide range of physical and mental disorders (Moos et al., 1990; Orford, 1986). Women with active alcoholic partners have been found to experience a higher level of depression, trauma, and stress-related disorders (Roberts & Brent, 1982; Svenson, Forster, Woodhead, & Platt, 1995). Their marital distress appears directly related to stress or burden brought on by their partner’s drinking (Moos et al., 1990; Zweben, 1986). Also, the way in which individuals cope with an alcoholic partner appears related to their own functioning (Moos et al., 1990; Moos, Finney, & Gamble, 1982), their partner’s functioning (Orford et al., 1975; Schaffer & Tyler, 1979), and their partner’s recognition of a drinking problem (Sobell, Sobell, Toneatto, & Leo, 1993).

A small but consistent body of research suggests that increased escape–avoidance and withdrawal coping behaviors in this population are associated with either the individuals’ own increased alcohol use and depression or the increased alcohol use of their partner (Moos et al., 1982; Orford et al., 1975; Schaffer & Tyler, 1979). The measures of coping typically used in these studies have been either frequency counts of alcohol-related coping behaviors (e.g., Orford et al., 1975; Schaffer & Tyler, 1979) or measures of general coping style (Moos et al., 1982). Each of these measurement methods can be problematic. Frequency measures are difficult to interpret. A change in the frequency of coping may reflect either a change in actual coping or a change in the partner’s drinking such that the opportunity to engage in different coping behaviors has changed. General coping measures, on the other hand, fail to provide clinically useful prescriptions for effective coping. Neither method fully takes into account the situational specificity of effective behavior.

The Spouse Situation Inventory (SSI) was developed to address these measurement problems. The SSI is a situation-specific role-play inventory of alcohol-related problems encountered by women with alcoholic partners. The inventory was developed via the behavioral–analytic model of scale development (Goldfried & D’Zurilla, 1969; see Rychtarik, Carstensen, Alford, Schlundt, & Scott, 1988, for early work on SSI development). It consists of 24 representative problem situation vignettes confronted by women with alcoholic partners (e.g., dealing with drunken behavior, partner violence, partner’s failure to maintain household responsibilities, and the children’s response to the partner’s drinking). Each situation is accompanied by specific response scoring guidelines developed systematically through the composite judgment of alcoholism professionals.

The SSI takes approximately 40 min to administer. The woman reads and imagines herself in a situation as it is read aloud by an administrator. She pretends that the administrator is her partner and says exactly what she would say to her partner and what she would do in the situation. Her response then is rated on a 6-point scale (1 = not effective at all, 6 = extremely effective) according to developed scoring criteria. This scoring is done either simultaneously with administration or, in the research setting, through later review of recorded responses, which requires approximately 50 min. The woman also rates, on a 5-point scale (1 = not similar at all, 5 = extremely similar), each situation’s similarity to those she has experienced in the past year. The inventory thus provides (a) an overall skillfulness measure reflected in the average situation effectiveness score, (b) a measure of skillfulness in specific situations, and (c) a mean situation similarity score that we interpret as a hardship index. To date, two alternate SSI forms have been developed (see the Appendix for a sample situation, transcribed responses, and associated scoring).¹

In the current study, we assessed the reliability and construct validity of the SSI overall skillfulness measure in both treatment and nontreatment populations. In addition, we explored the influence of administrator gender, SSI similarity, demographics, and treatment status on the measure. With the exception of age (Rychtarik, 1990), there is little prior work to direct the selection of demographic characteristics that may relate to SSI performance. As a result, the joint influence of the common variables of age, education, marital status, employment status, and occupational status was explored.

We assessed SSI reliability from within generalizability theory (Cronbach, Gleser, Nanda, & Rajaratnam, 1972) and supplemented these analyses with alternate form and test–retest analyses. This approach was most appropriate because we conceptualized SSI performance as a sampling of behavior from a wide range of different problem situations.² Generalizability analysis allowed us then to judge how well the behavior sample generalized to behavior on any random set of problem situations that may have been encountered. This approach also is useful for evaluating role-play tests because it simultaneously apportions variance into multiple sources (e.g., person, situation, and rater) and their interactions. Desirable variance in this case was that accounted for by overall skillfulness between individuals and the person–situation interaction.

The construct validity of SSI skillfulness was assessed in a series of analyses designed to test assumptions about how skillfulness is related to general escape coping, frequencies of alcohol-related coping behaviors, the woman’s functioning, and the functioning of the partner. We view escape coping as a general style that influences other, more focused coping efforts. Hence, we first examined the amount of variance in SSI performance accounted for by escape coping. Consistent with prior literature in this area, we predicted that higher levels of skillfulness would be associated with lower levels of escape coping. We also hypothesized that higher skillfulness on the SSI would be associated with fewer behaviors supporting partner drinking and more behaviors supporting partner abstinence. No predictions were made with respect to the SSI’s relationship with the frequency of other behavior classes (i.e., punishes drinking and withdraws from drinking) because the skillfulness of these behaviors on the SSI varied greatly with the situation.

The final construct validity analyses stemmed from family stress theory and the literature previously reviewed. These analyses tested assumptions that high skillfulness would be associated with (a) lower levels of alcohol use and depression in the woman, (b) fewer drinking days in the partner, and (c) less resistance in recognizing a drinking problem among partners not in treatment. Each hypothesis was tested by entering SSI skillfulness into a hierarchical regression model that controlled for demographic characteristics, escape coping, and, in the case of partners’ functioning, women’s functioning variables found in preceding analyses to be associated with the SSI. These tests increased our confidence that a significant contribution of skillfulness to variance in each area would not be explained by its mutual association with other variables. The interaction between SSI performance and treatment status condition also was examined in the first two hypotheses. In the third hypothesis, the interaction between SSI performance and prior treatment history (a common distinguishing characteristic in nontreatment populations) was explored. Given their somewhat exploratory nature, conservative two-tailed tests were used in all analyses.

Method

Participants

Participants were 472 women recruited through newspaper advertisements and flyers for a study on stress and coping in women with alcoholic partners. The sample consisted of 222 women with partners in treatment and 250 women with partners not in treatment. One hundred seventy-nine of the women in the treatment group and 130 of the women in the nontreatment group were accompanied by their partner. Table 1 presents basic demographic and descriptive statistics.

Table 1.

Participant Characteristics by Treatment Status Group

	Treatment status group
	Treatment (n = 222)				Nontreatment (n = 250)
Characteristic	%	M	SD	Skew	%	M	SD	Skew
Demographic
Race
White	47				46
Non-White	53				54
High school education or more	76				75
Marital status
Married	49				51
Cohabiting	51				49
Employed	33				34
Occupational status		3.87	1.96	0.59		3.77	1.84	0.46
Age (years)		34.85	8.38	0.75		35.63	8.99	0.70
Partner prior treatment					54
Escape coping		9.40	5.44	0.27		9.00	5.21	0.26
Significant-other Behavior Questionnairea
Punishes Drinking		2.44	0.77	0.07		2.47	0.79	0.22
Supports Sobriety		2.88	0.68	−0.32		2.69	0.72	0.00
Supports Drinking		1.60	0.54	0.97		1.62	0.48	0.75
Withdraws from Drinking		2.14	0.54	0.23		2.14	0.61	0.43
SSI similarity
Form Ab		2.29	0.82	0.55		2.49	0.72	0.53
Form Bc		2.21	0.79	0.62		2.49	0.81	0.15
SSI performance
Form Ab		2.96	0.32	−0.43		2.98	0.31	−0.06
Form Bc		2.97	0.32	−0.26		2.93	0.36	−0.58
Participant functioning
Quantity-frequency index		1.66	4.30	4.91		1.14	2.97	4.25
Depressiond		25.89	14.08	0.47		25.25	15.16	0.39
Partner functioning
Drinking dayse		74.70	64.04	0.36		119.09	56.52	−.37
Precontemplationf						17.01	4.90	−0.41

Note. SSI = Spouse Situation Inventory.

^aExcludes participants who reported abstinent partners in the previous 6 months; treatment and nontreatment ns = 178 and 241, respectively.

^bTreatment and nontreatment ns = 115 and 125, respectively.

^cTreatment and nontreatment ns = 107 and 125, respectively.

^dTreatment and nontreatment ns = 190 and 224, respectively.

^eTreatment and nontreatment ns = 209 and 242, respectively.

^fNontreatment partners who participated and drank over the previous 6 months; n = 99.

To be eligible for participation, a woman had to report that her partner had a drinking problem and had to be currently residing with a spouse or male partner with whom she had lived for at least 1 year. Women whose partner was either (a) a current inpatient or outpatient in a treatment program or (b) a regular participant in Alcoholics Anonymous (AA) constituted the treatment group. All others were placed in the nontreatment group. Partner participation was not required for study participation’. AH participants’ partners met diagnostic criteria for alcohol abuse-dependence based on either the partner’s own report or the participant’s response to a brief screening instrument (Wieczorek & Miller, 1992).

Overall Design

The study involved a 2 (group: treatment vs. nontreatment) × 2 (SSI form) × 2 (gender of administrator) factorial design. Participants in each of the treatment status groups were randomly assigned to SSI form and to male or female administrator with the provision that approximately equal numbers of treatment and nontreatment participants be assigned to each Form × Gender condition. A random subsample of the participant pool also was selected to return for administration of the alternate (n = 105) or the same (n = 30) SSI form within 2 weeks of the initial administration.

Measures

SSI

Participants were administered either Form A or Form B of the SSI Seven master’s-level research assistants (three men and four women) served as SSI administrators over the course of the study. Administrators received extensive instruction in the conduct of the role-play assessment. They were trained to identify and prompt scorable responses and to respond in a nonjudgmental manner. A scorable response was one that included either an action or a direct statement by the woman to the partner.

SSI responses were videotaped and independently scored by raters trained on individual situation scoring criteria. Neil B. McGillicuddy trained raters to a .80 level of agreement (agreements/[agreements + disagreements]) on an independent sample of responses to each situation. Subsequently, one rater provided ratings on all participant videotapes. A second rater provided a second rating on 72% of the sample. Two additional secondary raters scored the remaining 28%. Total score intraclass correlations between the primary and secondary raters ranged from .75 to .89. Data provided by the three secondary raters were collapsed and treated as those of one secondary rater for subsequent analyses. Individual rater scores were used for all generalizability analyses. For all other analyses, the mean rater score was computed for each SSI situation per participant, and the total mean situation score across the 24 situations was computed and used.

SSI similarity score

This was the mean 5-point similarity rating of the 24 situations. Its internal consistency was .91.

Sociodemographic variables

In addition to age and occupational status, demographic variables were coded as follows: race (0 = White, 1 = non-White), education level (0 = less than high school, 1 = high school or greater), employment status (0 = employed, 1 = unemployed), and marital status (0 = married, 1 = cohabiting). Occupational status was coded 1 (low status) to 9 (high status) via the scale described by Hollingshead and Redlich (1958).

Treatment history variables

The partner’s treatment status group at the time of assessment was coded as 0 (nontreatment) or 1 (treatment). Participants in the nontreatment group also were classified as to whether their partner had ever received help for drinking either through inpatient or outpatient programs or through AA (0 = no previous help, 1 = previous help).

General escape coping

The eight-item Escape-Avoidance subscale of the Ways of Coping Questionnaire (Folkman & Lazarus, 1988) served as the measure of general escape coping.³ On this measure, the participant indicated, on a scale ranging from 0 (does not apply/not used) to 3 (used a great deal), the extent to which she used various escape coping approaches (e.g., “Refused to believe that it had happened”) to respond to a stressful situation that had occurred in the past week. The sum of item responses served as the escape coping variable in the present analyses. Variations of this subscale have been used successfully in prior research in this area (Moos et al., 1982). The internal consistency of the subscale was .79.

Alcohol-related participant behaviors

The Significant-other Behavior Questionnaire (SBQ; Love, Longabaugh, Clifford, Beattie, & Peaslee, 1993) served as an independent frequency-type measure of participant alcohol-related coping skills. The participant indicated, on a 4-point Likert scale (1 = never or only once, 4 = always or almost always), the frequency with which she had engaged in various responses during periods of partner drinking and sobriety in the last 6 months. Items are grouped into four factor-analytically derived subscales representing behaviors that (a) punish drinking (four items; e.g., “Did you have arguments about his drinking?”), (b) support sobriety (eight items; e.g., “Did you spend more time than usual with him when he was not drinking?”), (c) support drinking (six items; e.g., “Did you bring alcoholic beverages home?”), and (d) withdraw from drinking (five items; e.g., “Did you leave the house when he was drinking?”). Subscale internal consistencies were .82, .87, .68, and .65, respectively. The mean item rating on each subscale was used in the present analysis. Because SBQ responses are based on drinking over the past 6 months, we limited relevant analyses to the 419 women for whom 6-month partner drinking data were available and who reported at least some partner drinking over that period

Participant functioning variables

The quantity–frequency index (QFI) and global depression scale of the Health and Daily Living Questionnaire (Moos, Cronkite, Billings, & Finney, 1984) were used as measures of participant functioning. The commonly used QFI was derived from participants’ reports of the frequency and quantity of alcoholic beverages they had consumed over the past year. The QFI represents the average amount of absolute alcohol consumed per day over the period. A QFI log transformation was used in all data analyses.

The global depression score was the sum of scores on 18 depression-related items rated on a 5-point frequency scale (0 = never, 4 = often). This measure has been used in prior research on functioning among spouses of alcoholics (Moos et al., 1982). Its internal consistency was .94.

Alcoholic partner functioning

Participants reported the number of days their partner had consumed alcoholic beverages in a structured interview format adapted from Polich, Armor, and Braiker (1980). The interview was administered to the participant as it applied to her partner. The participant reported the number of days her partner had consumed alcohol over the past 30 days and over each of the previous 6 full months. The sum of the number of days drinking across the 6 months served as the measure of alcohol consumption. Finally, a modified version of the Precontemplation scale of the University of Rhode Island Change Assessment Scale (McConnaughy, Prochaska, & Velicer, 1983) was administered to participating partners and served as a measure of partner recognition of a problem. High scores on this measure suggest less recognition of a drinking problem by the alcoholic partner. For the present study, items were modified to apply directly to the individual’s drinking problem (e.g., “As far as I’m concerned I don’t have any problem with my drinking that needs changing”). The respondent rated the extent of agreement with each statement on a 5-point scale (1 = strongly disagree, 5 = strongly agree). A preliminary psychometric analysis of this modified Precontemplation scale revealed that three of the eight items did not correlate adequately with the full scale and resulted in inadequate internal consistency. Elimination of these three items resulted in a coefficient alpha of .81. This shortened five-item scale subsequently was used.⁴

Results

There were no significant differences between participants randomly assigned to SSI form, gender of administrator, or alternate or test–retest subgroups on demographic, participant, or partner variables. The distribution of SSI total scores on each form was normal and did not differ between forms or treatment status groups. The two forms also did not differ significantly in mean situation similarity rating. Mean SSI skillfulness and similarity ratings are included in Table 1. Preliminary analyses also revealed no significant gender effect or Gender × Form or Gender × Form × Treatment Group interaction. Subsequent generalizability and reliability analyses were collapsed across administrator gender and treatment status.

SSI Generalizability

Separate Person × Rater (primary vs. secondary) × Situation generalizability analyses of variance (Crick & Brennan, 1983) were conducted for each of the two SSI forms. Analyses used the random effects model for each of the three facets. Resulting variance components are presented in Table 2. Differences between individuals accounted for 61% and 63% of the variance for Forms A and B, respectively. The Person × Situation interaction accounted for an additional 24% and 27% of the variance on the respective forms. Sources of error variance were low. The generalizability coefficient was .65 for each form. This intraclass correlation can be viewed as the expected average correlation among pairs of randomly selected sets of any 24 problem situations that could be encountered.

Table 2.

Estimated Variance Components and Proportion of Variance Accounted for in Generalizability Analysis of Spouse Situation Inventory Forms A and B

	Form A (n = 240)		Form B (n = 232)
Source of variation	Estimated variance component	Proportion accounted for	Estimated variance component	Proportion accounted for
Persons	.06500	.61	.07518	.63
Raters	.00003	.00	.00000	.00
Situations	.00739	.07	.00351	.03
Persons × Raters	.00174	.02	.00195	.02
Persons × Situations	.02578	.24	.03211	.27
Raters × Situations	.00001	.00	.00002	.00
Residual (Persons × Raters × Situations)	.00732	.07	.00717	.06

Note. Analyses were based on 2 raters and 24 situations.

Factor analysis typically is not applicable in scale development such as that used for the SSI. However, it can be helpful in further explaining whether a substantial Person × Situation interaction results from a few or several situation-specific factors (Cronbach et al., 1972). A subsequent principal-components factor analysis of the SSI revealed the presence of 10 factors with an eigenvalue of greater than 1.0 on each form. One small but relatively dominant factor accounted for 12.8% and 12.4% of the variance in Forms A and B, respectively. The 9 remaining factors per form each accounted for 4% to 7% of the variance. The substantial Person × Situation interaction thus appeared to result from several small factors, each consisting of only a few situations. The small size of these factors precluded the development of reliable SSI subscales.

Alternate Form and Test–Retest Reliability

In the alternate form sample, there was no significant form order effect, time effect, or Order × Time interaction.⁵ We subsequently collapsed across testing occasions in an overall generalizability analysis with the following variables: persons, form, rater, and situations nested within forms. We used random effects modeling and assumed generalization from one form to all possible forms. The results mirrored those of the previous analyses and revealed no appreciable variance (0%) accounted for by form, the Rater × Form interaction, or the Person × Rater × Form interaction. The Pearson correlation between scores on Forms A and B was .62 for both the A–B and B–A orders. In the test–retest sample, test–retest reliabilities were .73 and .72 for the respective forms.

Construct Validity

Only age and occupational status contributed significantly to SSI performance when the SSI was simultaneously regressed on demographic variables. Both were positively associated with SSI skillfulness. These two variables were included in all subsequent regression models. Table 3 presents the intercorrelation matrix for the remaining variables. The SSI was not associated with SSI form, treatment status group, or prior treatment experience. SSI performance did relate significantly to hardship, as measured by SSI similarity. Individuals reporting higher levels of situation similarity performed more poorly on the SSI. This effect was small but remained after demographic variables had been controlled. As a result, SSI similarity was controlled in all remaining analyses.

Table 3.

Correlation Matrix of Variables Used in Analyses

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
1.	Age	—
2.	Occupational status	.12**	—
3.	Group status	−.04	.03	—
4.	Prior treatment	−.02	−.01		—
5.	SSI form	−.04	−.04	−.02	−.04	—
6.	SSI similarity score	.02	−.08	−.15**	−.02	−.02	_
7.	SSI score	.15**	.18***	.02	.01	−.03	−.17***	—
8.	SSI × Group Status	.11*	.12*	.02	.01	.01	−.10*	.67***	—
9.	SSI × Prior Treatment	−.00	.05		.09	−.02	.02	.74***		—
10.	Escape coping	−.16**	−29***	.04	.04	.00	.23***	−.19***	−.11*	−.13	—
11.	SBQ Punishes Drinking	−.12*	−.05	−.02	.03	.04	.34*	−.11*	−.02	−.09	.21***	—
12.	SBQ Supports Sobriety	−.08	−.01	.13**	.15**	−.06	−.01	−.03	−.01	−.02	.05	.24***	—
13.	SBQ Supports Drinking	−.03	−.15**	−.02	−.02	−.01	.00	−.21***	−.18**	−.19	.14***	−.13**	.13*	—
14.	SBQ Withdraws from Drinking	−.01	−.02	.00	.01	−.02	.28***	−.08	−.03	−.14	.16**	.53**	.00	−.11*	—
15.	Quantity–frequency index	−.06	−.22***	.06	.11	−.03	.07	− 17***	−.13**	−.09	.25***	−.10	.08	.52	−.01	—
16.	Depression	.01	.05	.02	.09	.01	21***	−.07	−.05	.11	.36***	.08	−.17**	−.02	.10	.02	—
17.	Partner drinking days	.15**	−.05	−.34***	−.26***	−.02	.27***	−.18***	−.18***	−.19	.15**	.03	−.16**	.20***	.17**	.23***	.06	—
18.	Partner precontemplation	.17	−.11		.29***	.11	−.04	−.07	−.06	.03	−.18	.08	.00	−.14	−.02	−.12	.09	−.06	—

Note. SSI = Spouse Situation Inventory; SBQ = Significant-other Behavior Questionnaire.

^*p < .05.

^**p < .01.

^***p < .001.

Escape coping

As shown in Table 3, SSI performance had a small but significant negative bivariate correlation with escape coping. In a subsequent hierarchical regression model, escape coping accounted for an additional small but significant amount of SSI score variance, R² change = .01, F(1, 467) = 4.50, p < .05, above and beyond that accounted for by demographic and similarity measures, full model R² = .08, F(4, 467) = 10.30, p<. 001.

Alcohol-related coping behaviors

Only the Punishes Drinking and Supports Drinking subscales of the SBQ had significant bivariate correlations with the SSI, and these correlations were negative (see Table 3). When the set of four SBQ scales was added to a hierarchical regression model with previously entered demographic, similarity, and escape coping measures, the scales accounted for an additional small but significant amount of SSI variance, R² change = .03, F(4, 410) = 3.67, p < .01, and full model R² = .11, F(8, 410) = 6.02, p < .0001. In the full model, however, only the Supports Drinking subscale contributed significantly, β = −.19, t(410) = −3.76, p < .001. Higher SSI scores were significantly associated with lower levels of support for drinking. Although the Supports Drinking subscale was significantly related to SSI performance, we chose not to include it as a control variable in remaining analyses. This decision was made a priori because the SBQ was viewed as measuring the same construct and also would result in the loss of data through elimination of partners who had no reported drinking in the previous 6 months.⁶

Women’s functioning

In these and all subsequent models, z scores were used for the SSI, SSI × Treatment Status, and dependent variables to test for interaction effects. As predicted, the SSI had a significant negative bivariate correlation with the woman’s own alcohol use (see Table 3). As shown in Table 4, high SSI skillfulness also was associated with lower levels of alcohol consumption than those accounted for by other control variables. Contrary to our prediction, SSI performance was not associated with the participant’s level of depression. However, the SSI similarity, score was associated with this measure; women who reported the SSI situations to be more similar also experienced higher depression levels.

Table 4.

Hierarchical Regression Analyses of Participant Quantity–Frequency Index (QFI) and Depression on Demographic, Treatment Status, Escape Coping, and SSI Measures

	QFI		Depression
Variable	R2	β	R2	β
Background	.05****		.01
Age		.01		.05
Occupational status		−.15**		.17***
Treatment group	.05	.06	.01	.03
SSI similarity	.06	.01	.05****	.14**
Escape coping	.09***	.18**	.18****	.38****
SSI score	.10*	−.09	.18	.01
SSI × Group	.10a	−.04	.18b	−.02

Note. Significance levels in R² columns reflect significance of R² change; betas are for the final, full regression model. SSI = Spouse Situation Inventory.

^aOverall F(7, 464) = 7.46, p < .0001.

^bOverall F(7, 406) = 12.39, p < .0001.

^*p < .05.

^**p < .01.

^***p < .001.

^****p < .0001.

Alcoholic Partner Functioning

SSI performance also had a small but significant negative association with partner drinking days (see Table 3). In the regression model, both SSI total score and the interaction between total score and group significantly contributed to variance in the alcoholic partner’s drinking days above and beyond other variables (see Table 5). A closer examination of the interaction effect showed that a significant negative relationship between SSI performance and partner drinking was present for the treatment sample, β = −.22, t(442) = −3.59, p < .001, but not for the nontreatment sample, β = −.04, t(442) = −0.77, ns. A further examination of differences between the treatment and nontreatment groups at different levels of SSI performance indicated that the two groups differed significantly (p = .05) in partner drinking days at untransformed SSI score levels of 2.39 and higher. The two treatment status groups did not differ significantly in partner drinking days at lower SSI performance levels.

Table 5.

Hierarchical Regression Analyses of Partner Drinking Days and Precontemplation on Demographic, Treatment, Participant Functioning, Escape Coping, and SSI Measures

	Days drinking		Precontemplation
Variable	R2	β	R2	β
Background	.03**		.04
Age		.17****		.11
Occupational status		.05		−.15
Treatment group	.14****	−.33****
Previous help			.11**	.31**
Quantity-frequency index	.20****	.21****	.14	−.15
SSI similarity	.24****	.16***	.14	−.06
Escape coping	.25*	.09*	.16	−.17
SSI performance	.26**	−.04	.19	−.32*
SSI × Group	.27*a	−.12*	.21b	.21

Note. Significance levels in R² columns reflect significance of R² change; betas are for the final, full regression model. SSI = Spouse Situation Inventory.

^aOverall F(8, 442) = 20.76, p < .0001.

^bOverall F(8, 90) = 2.95, p < .01.

^*p < .05.

^**p < .01.

^***p < .001.

^****p < .0001.

SSI performance was not related directly to partner precontemplation, and its addition to the hierarchical regression only approached significance (p < .10). It was, however, significant in the full regression model. The partialing out of error variance associated with the interaction term (which was nonsignificant) appeared to enhance statistical precision for the main SSI effect. As predicted, higher levels of performance on the SSI were associated with greater recognition of a problem (lower precontemplation) by the partner.

Discussion

The SSI showed acceptable levels of generalizability and alternate form and test–retest reliability. Consistent with our conceptualization of skillfulness in this area, it appears to be multifactored and dependent, in part, on the situation. The size of the generalizability coefficient, the large amount of variance accounted for by differences in skillfulness between individuals, and the low level of error variance lend solid support for use of the total SSI score. The fact that the score also reflects a significant amount of variance from the Person × Situation interaction only highlights its potential for use in clinical settings with individual women.

SSI performance was affected to a small degree by the woman’s age and occupational status and by the similarity of the SSI situations. One might speculate that older women, through experience, learn to cope somewhat more skillfully than younger ones. Women of higher occupational status also may have learned to be more assertive and may be less dependent on their partner, thus feeling more comfortable with more skillful responses. The level of stress or burden, as reflected in SSI similarity, also may interfere with more skillful performance. Additional research on these factors is needed.

The construct of skillfulness measured here shared a significant but small amount of variance with escape coping. When the SSI contributed significantly to participant and partner functioning variables, however, it did so at a level above and beyond that accounted for by escape coping alone. Also, although the anticipated negative relationship with behavior that supports drinking was found, the predicted association with behavior supporting sobriety was not. The limited number of SSI situations dealing with periods of sobriety may account for our failure to find a significant relationship here. Also, as noted earlier, frequency measures may be insensitive to the situational variations that are reflected in the SSI skillfulness score.

Consistent with our other predictions, skillfulness was related negatively to the woman’s own alcohol consumption. Her depression, however, was predicted only by escape coping and SSI similarity. The latter finding adds support to our interpretation of similarity as a measure of burden. Also, participants with higher levels of skill on the SSI had partners who had been drinking less over the past 6 months. This finding is consistent with earlier research using frequency measures (Orford et al., 1975). The effect was limited, however, to participants whose partners were currently involved in treatment. The results suggest that skills, as measured on the SSI, may have the most impact on the partner’s recognition of the problem before treatment but not necessarily affect his drinking behavior. Once the partner has sought help, however, higher levels of coping in the woman may facilitate reductions in drinking.

It must be cautioned that the size of the contribution of SSI skillfulness to variance in both participant and partner functioning variables was small. This finding, however, must be interpreted in the context of the relatively low and narrow range of skillfulness present in the sample. A broader range of skillfulness may have resulted in larger observed effects. Also, the relationship between the SSI and precontemplation was found only in the full model with the nonsignificant interaction term, suggesting a relationship that is complex and in need of further study and replication. Finally, the causal direction of the relationships identified here is not clear. Predictive validity and skill training intervention studies are needed to elucidate the causal chains implied.

Nevertheless, the SSI has potential for direct application in both research and clinical settings. It shows promise as both a process and outcome measure of the effects of skill training and other interventions. In the clinic, the SSI role-play format and guidelines for effective coping easily can be assimilated into individual or group skill training programs. Therapist familiarity with SSI scoring criteria is important; however, use of video recordings or formal scoring of the SSI is not needed. Formal clinical assessment and monitoring of progress will await the development of alternate SSI administration formats (e.g., multiple choice) that can be more efficiently administered and scored.

Appendix

Sample SSI Situation, Transcribed Responses, and Associated SSI Scoring Situation

Situation

Your partner was spending too much money on drinking, so you had to take control of the family finances. It’s Saturday afternoon, and your partner has been trying to fix the kitchen faucet for about an hour. He asks you if he can have some money to go out and get a part he needs. In the past he has used these occasions to go out and drink. He says, “Give me ten bucks so I can get the part I need. I’ll he back in an hour or so.”

Responses

6-Point Response

“I’d say, ‘Honey, I care about you very much, and I really appreciate the way you’re taking the time to help out and fix the faucet. However, in the past when I’ve given you money to buy things for the house, you’ve spent it on drinking. I really want to be able to trust you, but right now I don’t. I’ll give you the money this time, but if you use it to drink, I will not give you any more money in the future and we’ll have to budget money for a repairman the next time something needs fixing.”

5-Point Response

“I haven’t been able to trust you recently. I feel like I can’t trust you. I don’t want to take any more control here because it’s not good for me and you need to take responsibility for what’s happening here and your drinking is causing money problems. Here’s ten bucks. I hope you go and buy what we need here because if you don’t and the kitchen faucet still doesn’t work, I just don’t know how much more I can take … how much longer I can stand here and watch this go on. I don’t want to take responsibility for these things any longer, you have to do it.”

4-Point Response

“ ‘It doesn’t take an hour to go to the store and get that part, you know. If that’s the case, what part do we need? I’ll go get it and if you want, you can go with me, but you’re not having the money to go out.’ I wouldn’t give it to him.”

3-Point Response

“Honey, tell me what part you need and I’ll go get it. I need the kitchen sink to be fixed and you’re already there doing it and that’s my priority right now to get the kitchen sink fixed, so why don’t you let me go get the part?”

2-Point Response

“I would give him the ten dollars. I wouldn’t say anything. I’d let him go, I really would. I’d say nothing, I know that.”

1-Point Response

“ ‘Oh honey, what part do you need? I’ll go get it and I’ll pick you up a quart of beer on the way home.’ And, if I pick up something to drink he’s usually quiet.”