Participants were individuals from a representative sample of households recruited by the Telepanel Foundation in Amsterdam, The Netherlands. Each of these households is equipped with a microcomputer that is connected to the central computer of the Telepanel Foundation. Regularly, members of the households answer computerized questionnaires presented via the computer. All instructions and questions are presented on the computer screen and questions can be answered by means of a keyboard. The computer administers all answers and, upon completion of the questionnaire, the responses are sent to the computer of the Telepanel Foundation.

The final sample of 722 individuals who participated in the study were currently involved in relationships. They participated in what they thought was a study on "Partner and Relationship." This vague description was chosen in order to avoid specific expectations about the objectives of the study. Their ages ranged from 17 to 56, with an average of 37.1 years. Unfortunately, demographic information regarding marital status and the duration of the relationship was only available from 370 participants (187 women, and 183 men). Of this subsample, 301 participants were married, 51 were cohabitating, and 18 were dating exclusively. Their relationships had an average length of 16.0 years (ranging from 1 month to 35.5 years).

A 2 x 2 factorial design was used in this study. Two levels of Direction of Comparison (enhancing versus deprecating) and two levels of Target under Evaluation (self vs. partner) constituted between-participants factors. The dependent variable involved the extent to which the verbal statements were considered as socially desirable. Participants were randomly assigned to one of the four experimental conditions.

The study was introduced to participants as a study dealing with their opinions and their reactions to other people's public evaluation of their intimate relationship and their intimate partner. Of principal interest here is a set of four scenarios that were developed to serve as the basis for varying the direction of comparison and the target under evaluation. All scenarios described a prior research in which participants discussed their relationship. Participants in each experimental condition were informed that they would be given a short fragment to read from one of the group discussions. They were requested to read very carefully the verbatim fragment that was allegedly derived from a group discussion. The instructions to the participants were worded as follows:

The statements in this fragment conveyed that one of the discussants evaluates the self (as a partner) or the intimate partner as positive and above average, or as negative and below average. The statements in the self-enhancing condition read:

To compare the responses to a self-enhancing evaluation with an evaluation in which the partner was described as positive and in a superior fashion, a partner-enhancing evaluation version was used:

In addition, two discussion fragments were created that involved self- and partner-deprecating evaluations. The following illustrates the self-deprecating condition:

In the partner-deprecating condition the scenario stated:

After reading the scenario, participants were asked to answer a set of questions. Each question was displayed on the computer screen simultaneously with the discussion fragment. They could proceed at their own pace through the computer program.

Social desirability of the verbal statements was measured with two questions. Following Cialdini et al. (1990, 1991), two items were used that required participants to indicate the extent to which they expected most people would do (i.e. the descriptive norm) and what they believed ought to be done in the same situation (i.e. the injunctive norm).

Hence, participants were asked "Do you consider the statements as normal?" and, respectively, "Do you consider the statements as reprehensible?" (reverse scored) on 5-point scales ranging from I disagree strongly to I agree strongly. The two questions obtained a correlation of .66 and were averaged to form one variable, so that higher scores indicated higher social desirability.

Feelings of sympathy for the discussant's statements were assessed by asking participants on a 5-point scale whether they felt sympathy for the discussant. The scale ranged from Not at all to Very much.

Participants then responded to two attribution measures designed to assess self-enhancement attribution and compliance attribution. Self-enhancement attribution was measured by asking whether participants believed that the discussant made his statements (a) to enhance self-esteem, (b) to feel good about the self (c) to make a good impression on others, and (d) to receive full appreciation of others. The four items were closely associated with one another, as indicated by a Cronbach's coefficient alpha of .82.

Secondly, compliance attribution was measured by asking participants whether they believed that the discussant made his statements (a) because the discussant did not wish to put forward a different view, and (b) to avoid others' disapproval. The two items obtained a correlation of .44 and were averaged to form one variable. The items were answered on 5-point scales ranging from Not at all to Very much.