Psychological Intimacy in the Lasting Relationships of Heterosexual and Same-Gender Couples - Psychological Intimacy

Written by Richard A. Mackey, Matthew A. Diemer, Bernard A. O'Brien.

Page 5 of 12

Sample

Couples were recruited through business, professional, and trade union organizations, as well as through churches, synagogues, and a variety of other community organizations. Most couples resided in the northeast part of the country.

The sample was chosen purposively to fit with the goal of developing an understanding of a diverse and older group of heterosexual and same-gender couples in lasting relationships. Couples were recruited who met the following criteria:

1. They were married or in a committed same-gender relationship for at least 15 years.

2. They were diverse in race/ethnicity, education, religious background, and sexual orientation.

Of the 216 partners who were interviewed, 76% were white and 24% were people of color (African-Americans and Mexican-Americans). The religious background of the couples was as follows: 46%

were Protestant; 34% were Catholic; and 20% were Jewish. Fifty-six percent were college graduates and 44% were non-college graduates. The mean age for the sample was 57 years (SD = 10.24): 27% of participants were in their 40s, 33% in their 50s, 26% in their 60s, and 14% in their 70s. Sixty-seven percent of couples were heterosexual and 33% in same-gender relationships. The mean number of years shared together was 30.22 (SD = 10.28): 18% of couples had been together 40 years or longer; 29% between 30 and 39 years; 34% between 20 and 29 years; and 19% less than 20, but more than 15 years. Seventy-seven percent of the couples had children; 23% did not have children. By total gross family income, 7% of couples earned less than $25,000; 25% between $25,000 and $49,999; 29% between $50,000 and $74,999; and 39% had gross incomes of $75,000 or more.

Coding

Each interview was tape-recorded and transcribed to facilitate coding and prepare the data for both quantitative and qualitative analysis. Interview passages were coded for relational themes, which were then developed into categories (Strauss & Corbin, 1990).

Initially, a research team (two women, two men) coded eight transcriptions blindly and individually. Detailed notes were kept and categories were generated. A relationship coding sheet was developed and used in subsequent coding of eight additional interviews. As new categories arose, previous interviews were recoded in keeping with the constant comparative process. Having both genders involved in that process helped control for gender bias and contributed to the development of a shared conceptual analysis. A scoring system was developed to identify themes that evolved from each section of the interviews. There were over 90 categories in 24 topic areas for every participant.

Once the Relationship Coding Sheet was developed, each interview was coded and scored independently by two raters (one male, one female), who noted themes and categories as they emerged from the transcripts. One of the authors coded all 216 interviews to ensure continuity in the operational definitions of variables and consistency of judgments from case to case. The agreement between raters, determined by dividing the number of identical judgments by the total number of codes, was 87%. Cohen's kappa, used as a measure of interrater reliability, ranged from .79 to .93. When discrepancies occurred the raters met to discuss their differences and to re-examine the original transcripts until a consensus was reached on how a particular item was to be scored.

HyperResearch software (Hesse-Biber, Dupuis, & Kinder, 1992) enabled the researchers to perform a thorough content analysis of interview transcripts (totalling over 8,000 double-spaced pages) and identify, catalogue, and organize specific interview passages on which categorical codes were based.

In the second or current phase of the study, we re-examined the codes so as to prepare the data for quantitative analysis. Many variables were re-coded into dichotomous categories. For example, psychological intimacy was originally coded into three categories (positive, mixed, and negative). Because we were interested in understanding factors that contributed to psychological intimacy during recent years, the positive category was retained and compared with a recoded mixed/negative category. Vignettes from the transcripts are used in the following pages to illustrate the meaning of psychological intimacy to participants during recent years.

Data-Analysis

The coded data from the scoring sheets yielded frequencies that were analyzed using SPSS software. Chi-square analysis was used to examine the relationship between the independent variables--which included personal, demographic, and participants' reports of various dimensions of relationships--and the dependent variable of psychological intimacy in recent years. The Alpha criterion was set at .01 for the chi-square analysis.

The chi-square statistic seemed appropriate, since certain conditions were met. First, it has been very difficult to ensure randomness of samples in social and behavioral research, especially in studies that focus on new territory. This nonprobability sample was selected deliberately to include older couples who have been understudied in previous research--namely, heterosexual and same-gender relationships that had lasted an average of 30 years. The goal was to identify factors that contributed to satisfaction from the perspectives of individual partners rather than to test hypotheses. Second, compared to other tests of statistical significance, chi-square has fewer requirements for population characteristics. Third, the expected frequency of five observations in most table cells was met.

To assess the strength of the associations between psychological intimacy and the independent variables, a correlation analysis was conducted. Because of the dichotomous nature of the variables, a phi coefficient was computed for the dependent variable and each independent variable.

Variables that had been related significantly to psychological intimacy in the chi-square analysis and identified in previous studies as having importance to understanding psychological intimacy were selected for building a theoretical model. Based on the phi coefficients, communication was not included in the model (see next section). Two models were tested using logistic regression: one model included the sexual orientation of couples (heterosexual, lesbian, and gay males), the other substituted gender (male and female) for the sexual orientation of couples. Logistic regression was a useful tool in this exploratory research, where the goal was to develop theory rather than test it (Menard, 1995).