Ninety-four undergraduates at University of Groningen (57 women and 37 men) who were currently involved in a relationship participated in the study (78 dating relationships, 16 cohabiting, none were married). They participated in what they thought was a research on "Test Validation." The average duration of their relationship was 2.2 years (ranging from 1 month to 7.1 years). The mean age of the sample was 22.1 years, with a range from 18 to 28 years. All participants were paid 7.5 Dutch guilders (approximately U.S. $4.5) for their participation.

The design was a 2 x 2 factorial with two levels of target (self vs. partner) and two levels of cognitive set (no-comparison versus comparison). Participants were randomly assigned to list as many as possible self- or partner-evaluations in either noncomparative or comparative terms.

The study was presented via Apple MacIntosh computers, connected to a central server. All instructions and questions were presented on the computer screen and questions could be answered by means of a keyboard. Assurance was given that the data would be treated confidentially. Participants were run in groups ranging from 2 to 8.

On arrival at the laboratory, the experimenter explained that the purpose of the research was the construction and validation of a variety of relationship questionnaires with a computer-administered experiment. The participants were told that they could start the computer-administered procedure by hitting a key and that they could proceed at their own pace through the computer program. Each individual was then led to an individual booth and seated in front of a microcomputer.

Participants answered first a number of questions about their gender, age, duration of their relationship, and other relevant characteristics to familiarize them with the computerized test procedure. Next, the experimental task was introduced.

Participants were told that one purpose of the study concerned the "development of the Groningen Relationship Evaluation Test, for which we need many evaluations of intimate relationships." Hereafter, participants were assigned randomly to one of the following conditions. Participants in the self no-comparison condition (n = 23) were asked to generate as many as possible reasons why they are good partners. Participants in the self comparison condition (n = 23) were asked why they considered themselves as better partners than most others.

Participants in the partner no-comparison condition (n = 23) and in the partner comparison condition (n = 25) were instructed to describe why their boy/girl friends were respectively good partners or better partners than most others. All participants were instructed to write down as many reasons of the specified type as they could think of --but at least five - on a piece of paper. Once they had finished, they were instructed to press 'Return', at which point the computer screen cleared and displayed the instruction to enter the five reasons which they considered to be most important into the computer.

Hereafter, participants completed computer-administered measures of satisfaction and commitment. Next, participants were orally debriefed and asked not to discuss the study with others. After being paid for their participation, participants were thanked and dismissed.

The measures of relationship satisfaction and commitment were modeled on questionnaires used by Rusbult (e.g., Rusbult, 1980, 1983), and were, following Fincham and Bradbury's (1987) recommendation, global, evaluative questions rather than domain-specific questions. Satisfaction was measured by four items: "In general, how satisfying is your relationship?" (1 = not at all and 9 = extremely), "How much happiness and enjoyment do you get out of your relationship?" (1 = very little and 9 = very much), "How does your relationship compare to your ideal?" (1 = far from ideal and 9 = it is ideal), and "All things considered, how negative or positive do you stand on your relationship?," (1 = very negative and 9 = very positive).

Commitment was measured by eight items, such as "For what length of time would you like your relationship to last?" (1 = a month or less and 9 = ten years or more), "To what extent do you feel committed to your relationship, even when you are less satisfied with your partner?" (1 = not at all and 9 = extremely), and "How likely is it that you will end your relationship in the near future?" (1 = not at all likely to end and 9 = extremely likely to end; recoded). Single averaged measures of satisfaction (Cronbach's alpha = .82) and commitment (Cronbach's alpha = .93) were formed.

The questions on satisfaction and commitment were answered by pressing one of the numerical keys (1 through 9) on the keyboard of the computer, after which the next question was presented on the computer screen. The latencies of response (from question onset to response) were recorded to the nearest millisecond by a software clock. Also, to control for individual differences in latencies, baseline response times were assessed by recording the latencies of response to the biographical questions. By doing so, the general speed of responding can serve as a covariate, thus reducing error variance due to individual differences in speed of responding that are unrelated to accessibility, and enhancing statistical power (cf. Fazio, 1990).