All three of the Web sites had two sections. The first portion was pre-formatted with fill-in-the – blank headers such as education, occupation, hobbies, desired level of commitment and reli – gion/ethnicity. This section was not analyzed for two reasons: (1) The headers were not consistent across the Web sites, and more importantly; (2) Our interest was in the self-presentation section. It is in this unstructured section that participants can draw attention to any aspects about themselves they desire to project—in other words, to engage in impression management. In many cases these narratives included repeating and/or elaborating upon information from the first section. As would be expected, in most cases these narratives also provided additional insights into the person’s activities, interests and lifestyle. It was this section that was coded and content analyzed.
For our purposes, we counted words within the narratives that mapped onto 10 pre-determined categories, for example, love, personality traits, entertainment and physical characteristics (see hypotheses H2a-j above for the complete list of categories). We acknowledge that words interpreted as signaling each of the 10 categories of interest can be debated. For our purposes, words categorized as ‘love’ had to have nurturing or bonding connotations, examples of which would include family or caring; personality trait descriptors would include comments like up for a challenge and likes sports; entertainment services referred to capabilities they had with which they could indulge a partner, such as play the guitar and accomplished dancer; physical characteristics was in reference to words that served to signal one’s beauty or health, such as not that good looking, I have all my own teeth, and charming smile, and so forth.
Four students enrolled in a consumer behavior course who were not informed as to the purpose of the study categorized the narratives. A practice coding test run took place during which students sat separately and then compared their categorizations. The inter-rater reliability was unacceptably low. The source of the low inter-rater reliability was twofold. The first source of discrepancy that was quickly resolved was to limit the analysis to self-descriptions. Thus, coders were informed to not count any portion of a narrative that was directed at someone other than the person writing the narrative. Comments not directed at their self were typically about desired qualities in a mate, but on occasion included comments about family members or friends. The second source of disagreement stemmed from the fact that often narratives had seemingly redundant references, making it difficult to agree on the appropriate number of times to count a comment. Consider, for example, the following verbatim extract from an Australian male:
I have quite a hectic lifestyle… busy most nights of the week … and getting busier. … I love having a good time as well as love going nuts and having fun… I love having a good time…
Arguably this could be counted as just two personality traits, those comments related to a ‘hectic lifestyle’ and those that referred to ‘having a good time’. However, it was decided that the person writing this narrative wanted to emphasize these points and therefore that the redundancies should be counted—in other words, we endeavored to be as liberal in our interpretation of word meaning as was feasible. This narrative was therefore counted as seven personality traits (three of which relate to their hectic lifestyle and four to having fun).
In light of the frequency with which redundancies like the aforementioned appeared, it was decided that the appropriate way to categorize words was by having two research assistants sit together to code each narrative and resolve any differences concerning how to categorize a word at that time. Given that each narrative is an independent sample it is reasonable to assume that should the narratives be categorized by different individuals there would be disparities, but these differences should approximate a normal distribution about the true mean.
For illustrative purposes, two verbatim narratives are presented complete with spelling and grammar errors along with the resultant coding counts for four of the categories of interest: love, personality trait, entertainment and physical characteristics.