Inside for every single site, i accumulated step 1,100 pages of heterosexual people and you will 1,100000 profiles out of heterosexual ladies. Users choose pages through geographical venue, decades, and gender filter systems.

To make sure a geographic dispersion off users, we picked equal quantities of pages off four biggest locations also metropolitan, suburban, and outlying parts: La, Denver, Chi town, Atlanta, and you can New york. We randomly selected zip codes from all the five portion to look for profiles. In this each zip code, for each intercourse, i up coming randomly selected users among four age range: very early more youthful adulthood (old 18–30 age; letter = step one,000; Yards = , SD = step 3.17), later more youthful adulthood/very early midlife (aged 31–49 ages; n = step 1,000; M = , SD = 5.61), late midlife (aged 50–64 age; letter = step one,000; Yards = , SD = 3.99), and you may older adults (old more 65 many years; n = step 1,000; M = , SD = cuatro.29). We made use of these types of stratifications in order to guarantee a complete age range out-of matchmaking users inside testing. Because older adults class you may utilize as much as three decades, we managed age just like the a continuing variable in lieu of just like the an effective grouping variable for the analyses.

Off for every reputation, we extracted: gender, ages, ethnicity, together with “Throughout the Me” otherwise “In my own Conditions” free reaction part. In order to guarantee privacy of character editors, we didn’t see extra group advice (elizabeth.grams., training, spiritual choice, income) that’ll act as pinpointing suggestions. The fresh sampling system is illustrated in Supplementary Appendix A .

New take to varied inside many years out of 18 to 95 decades. A separate t-decide to try revealed zero difference between indicate many years for ladies (M = , SD = ) and you can males (Meters = , SD = ). The fresh new post on ethnicity in the take to are 70% White/Caucasian, 11% Black/African american, 7% Hispanic/Latino, 2% Western, and you can ten% blended competition/almost every other.

Character posts

I utilized the LIWC application to research the content of one’s users. This program calculates brand new regularity and you can proportions of particular kinds of terminology contained in this a book file. The new LIWC system compares for every single word of a book document which have an internal dictionary in excess of 4,five hundred conditions allotted to keyword groups. This research received to the eleven mainly based LIWC kinds: first-individual only one pronouns, first-person plural pronouns, relatives, relatives, performs, completion, currency, wellness, sex, self-confident feeling, and you may bad emotion. Dining table step one consists of analogy terms and conditions inside each one of the LIWC group (for additional information out-of these codes, select LIWC, 2007).

I also authored a group of terminology for appeal not available during the built LIWC categories. I adopted methods getting build from LIWC categories ( Tausczik & Pennebaker, 2010) because of the promoting a thorough variety of words from dictionaries, thesauruses, questionnaires regarding past browse, and you can words made by the study people. Following, we selected twenty-five conditions most representative of attractiveness considering appearance for the thesauruses and you can fellow member responses (age.g., sweet, good-lookin, good looking, hot). The newest elegance group are almost totally distinct from this new intimate group, with just you to definitely overlapping word (sexy). Instances to the elegance class also are found in Table 1; to your over list of words on elegance category, select Secondary Table 1 .

Analytic Approach

I very first checked out detailed analytics for the dimensions of terminology some one of various many years found in their relationships pages. We along with generated illustrative figures portraying widely known terms and conditions.

We up coming turned to theory assessment using typical minimum squares regression. The outcome variables inside data was indeed this new proportion of terms and conditions fitted each of the a dozen classes regarding the LIWC analyses. New LIWC classes were all of the undoubtedly skewed due to the amount out of zero viewpoints (i.elizabeth., participant did not fool around with one terminology from the category). I went each research with a rectangular-sources sales (always target non-normality inside earlier in the day training utilizing the LIWC; DeAndrea, Shaw & Levine, 2010; Hirsh & Peterson, 2009). Brand new pattern off findings is actually comparable once using the changes. Having simple translation, findings is actually displayed making use of the untransformed LIWC group data. The new independent adjustable are ages, addressed due to the fact a continuing varying. I and incorporated gender.