The aim of this study was to investigate to what extent other-generated and system-generated cues on social network sites (SNS) influence the popularity and attractiveness of adolescents. In a 2 (friends' physical attractiveness: attractive, unattractive) ×2 (friends' wall postings: positive, negative) ×3 (number of friends: low, average, high) factorial experiment, 497 high school students between 12 and 15 years of age were randomly assigned to one of the twelve experimental conditions. Results revealed that the profile owner of a SNS was perceived as being more attractive when the profile includes attractive friends and positive wall postings. The profile owners' number of friends did not affect the perceived attractiveness of the profile-owner, only the perceived extraversion.
Social network sites (SNS), like Facebook and MySpace, have become immensely popular in the past years. In 2008, 580 million unique people across the world visited a social network site (Comscore.com, 2008). In 2008, about 67% of U.S. Americans between 18 and 30 years of age were a member of a social network site (Kohut, 2008). In the Netherlands, the most popular social network site, Hyves, has over 9 million users, and 75% of all Dutch teenagers between 12 and 17 years old have a profile on this site (Duimel, 2009). SNS are typically used both to keep in touch with existing friends and to develop new friendships (Boyd, 2004; Ellison, Steinfield, & Lampe, 2007). The sites allow members to create a personal profile, observe other profiles and communicate with other members, and develop and/or maintain friendships with these members (Ellison, Steinfield, & Lampe, 2006).
SNS are sources of social information and offer many opportunities for impression formation, such as descriptions, status updates, and photos (Boyd, 2004; Donath & Boyd, 2004; Manago, Graham, Greenfield, & Salimkhan, 2008; Walther, Van Der Heide, Kim, Westerman, & Tong, 2008). SNS offer a mixture of interactive and static features, which give a lot of information about the profile owner (Antheunis, Valkenburg, & Peter, 2010; Tong, Van Der Heide, Langwell, & Walther, 2008). Three sources of information can be distinguished on SNS: self-generated cues, other-generated cues, and system-generated cues (Tong et al., 2008). Self-generated cues are cues that a profile owner has full control over, such as posting visual and descriptive information about the self (e.g., photos and the books and music she or he likes). Other-generated cues are sources of information on a user's profile that come from others, such as messages (i.e., the wall postings on Facebook) left by friends of the profile owner, and pictures of those friends that are visible on a user's profile. Finally, system-generated cues are cues that the computer system chooses to show on a user's profile, such as the number of friends the profile owner has gathered on his SNS (Tong et al., 2008).
People rely on different sources of information about a target person to decide whether they like this person and if they want to become friends with this person (Antheunis et al., 2010; Berger & Calabrese, 1975; Derlega, Winstead, Wong, & Greenspan, 1987; Sunnafrank, 1986). In order to form impressions of others, people tend to place most values in those cues that have a higher warranting value, that is, online information that is reliable information on the basis of which an accurate impression of a person may be formed (Walther & Parks, 2002). On SNS, especially other-generated and system-generated cues may have the highest warranting value. The most salient other-generated cues are the wall postings on one's profile page and the attractiveness of one's friends. These may be manipulated to some extent, but nevertheless provide more warranting value than self-generated information. The most salient system-generated cue is the number of friends that is shown on a user's profile page. Although one has a certain degree of control over how many friends one has, the number is always shown on the user's profile. A user cannot choose not to show the number of friends.
Several studies have investigated the effects of impressions on SNS on interpersonal attraction and popularity (Kleck, Reese, Behnken, & Sundar, 2007; Tong et al., 2008; Walther et al., 2008; Walther, Van Der Heide, Hamel, & Shulman, 2009; Utz, 2010). Kleck et al. (2007) investigated the effects of the number of friends on SNS in different modalities on perceived popularity and attractiveness. They found that the profile owner was perceived as more popular and attractive when their SNS includes a large number of friends compared to a small number of friends. Tong et al. (2008) examined the effects of the number of friends on attractiveness and extraversion and found a curvilinear effect between the number of friends and attractiveness. Walther et al. (2008) have studied the relationship between a) the content of the wall postings and b) the attractiveness of one's friend on the attractiveness and credibility of the profile owner. The physical attractiveness of the profile owners' friends did positively influence both physical and social attractiveness of the profile owner. Walther et al. (2009) investigated the effect of self-generated versus other-generated cues on attractiveness. They found that other-generated cues (i.e., friends' comments) affected the profile owner's attractiveness more than self-generated cues (i.e., profile owner's comments). Finally, Utz (2010) recently examined how far extraversion of the target, extraversion of the target's friends, and number of friends influenced the perceived popularity, communal orientation, and social attractiveness of the target person. She found that all three manipulations of the profile positively affected the perceived popularity of the profile owner. None of the manipulations directly affected the profile owner's social attractiveness.
Although earlier research has investigated the effects of the number of friends (i.e., system-generated cues), the physical attraction of the profile owners' friends and the valence of the wall postings (i.e., other-generated cues) on impression formation on SNS, both these other and system-generated cues have never been investigated in a single study. Investigating these cues together allows for determining the strongest predictors of attractiveness and allows investigating possible interaction effects. Second, other and system-generated cues may affect social evaluation of a profile owner in different ways. For example, perceived extraversion may be mostly determined by number of friends, while physical attractiveness is influenced by the physical attractiveness of one's online friends. Therefore, this study investigates number of friends, physical attractiveness of friends, and valence of wall postings on social attraction, physical attraction, and extraversion in one study.
Moreover, this study focuses on early adolescents, for two reasons. First, adolescents are especially likely to turn to SNS to meet new friends and get to know each other (Hinduja & Patchin, 2008; Subrahmanyama, Reich, Waechter, & Espinoza, 2008; Valkenburg, Schouten, & Peter, 2005). In early adolescence, family relationships become less important and there is a sharp increase in the number and importance of peer relationships (Harter, 1999; Hartup, 1996). This development is reflected in adolescents' internet use. The number of online friends sharply increases in early adolescence (Boyd, 2008; Duimel, 2009). Second, adolescents are most likely to be affected by the impressions others form of them based on their online profiles.The need for online self-presentation is most significant among pre- and early adolescents. Young adolescents often are extremely preoccupied with what they appear to be in the eyes of others (Elkind & Bowen, 1979; Harter, 1999). Peer acceptance and interpersonal feedback on the self, both important features of SNS, are vital predictors of social self-esteem and well-being in adolescence (e.g., Armsden & Greenberg, 1987; Harter, 1999; Hartup, 1996). Indeed, research shows that the reactions adolescents receive on their online profiles affect their self-esteem and well-being (Valkenburg et al., 2005). In sum, adolescents are likely to search for information about others online, and are also most likely to be affected by the impressions other form of them. Therefore, it warrants investigation how adolescents form impressions of others based on information given off by SNS.
Identity claims and behavioral residue
Impressions are commonly formed with a minimal contact or on a minimum of information (Allport, 1937; Tajfel, 1970). Based on limited exposure people make broad generalizations about someone's personality. Allport (1937) described this as thin slices, which are short fragments of social behavior from which people can form impressions. A recent study found that slices of a 60-second video tape of a person were enough to judge someone's personality accurately. Even slices of 5 seconds did significantly improve the accuracy of the judgment (Carney, Randall, & Hall, 2007). This result supports Walther's hyperpersonal model (1996), which states that in situations of limited cue availability, people tend to rely on the limited information that is available. This, in turn, may lead to exaggerated and stereotypic impressions (Hancock & Dunham, 2001).
The Brunswikian lens model (Brunswik, 1956) describes the process by which individuals make inferences about the personality of others. Individuals often leave footprints that reflect their characteristics, on which other people base their impressions. According to this model, elements in the environment can serve as a lens through which observers indirectly perceive underlying constructs (Brunswik, 1956; Gosling, Ko, Mannarelli, & Morris, 2002). Gosling et al. (2002) distinguish between two types of cues that are given off by the environment (cf., Goffman, 1959): identity claims and behavioral residue. Identity claims are controlled cues given off by a person; they are given off intentionally by people in order to present themselves. Behavioral residues, on the other hand, are traces of earlier behavior that are unintentionally left behind that contain cues about past or anticipated behavior (Walther et al., 2008). Both identity claims and behavioral residue are used to form impressions (Gosling et al., 2002).
Profiles on SNS sites contain both identity claims and behavioral residue. On a SNS, self-generated cues such as photos, personal descriptions, and videos posted by the profile owner are examples of identity claims. The profile owner can freely decide which information to disclose in order to craft a desired image (Petronio, 2002). By making a strategic selection of the information they display, profile owners can manage their self-presentation on a SNS to a great extent (Herring & Martinson, 2004; Walther, Slovacek, & Tidwell, 2001). Profiles on SNS, however, also contain other-generated and system-generated cues, such as wall postings, number of friends, and the physical attractiveness of friends. This information is, at least partly, the result of a person's previous actions, but a person has little control over if and how this information is displayed. Although this kind of information is not initiated by the target person himself, it is associated with the target person and may thus also shape people's perception of that target person (Walther et al., 2008). Therefore, other and system-generated cues can be seen as behavioral residue.
Research on SNS and personal web sites has shown that the cues that are available on these sites affect other people's impressions of the website owner (e.g., Vazire & Gosling, 2004). However, the latter source of information, behavioral residue, is seen as more reliable because this information cannot be manipulated by the person to whom it refers (Walther & Parks, 2002). Self-generated cues online are often suspect because people assume that people actively manage their impressions by exaggerating some aspects of their selves (Ellison, Heino, & Gibbs, 2006). Indeed, people tend to rely more on other-generated cues than self-generated cues when forming impression about others online (Walther et al., 2009). Thus, system-generated and other-generated cues are considered to be more reliable sources of information. Studies investigating system and other-generated cues have shown that the number of friends of a profile owner, the physical attractiveness of that person's friends, and the wall postings friends have written all affect the impression other people form of a profile owner (Tong et al., 2008; Walther et al., 2008). However, the effects of other and system-generated cues on impression formation have never been tested in a single study.
The first source of other-generated cues we want to examine in this study is the physical attractiveness of the friends of the profile owner. Although profile owners control the friends they add to their SNS, they have less control about the physical attractiveness of their friends. Moreover, they have no control about their friends' pictures that are visible on their profiles. However, friends' photos on a profile do seem to affect the observer's impression of the profile owner (Walther et al., 2008). When people are surrounded by others who are physically attractive, they are judged to be more physically attractive as well, but only when there is a presumed relationship between these physically attractive others and the target person (Melamed & Moss, 1975). Other studies have since demonstrated that physical attractiveness of friends also affects other dimensions of attractiveness (e.g., Geiselman, Haight, & Kimata, 1984; Walther et al., 2008). Since there clearly is a positive relationship between a profile owner and the photos of his/her friends that are shown on the profile, our first hypothesis is:
H1: There will be a positive effect of online friends' physical attractiveness on the observers' perception of a) social attraction, b) physical attraction, and c) extraversion of the profile owner.
A second source of other-generated cues is the wall postings on the profile owners' webpage. These wall postings are short little messages that friends leave on the profile owner's webpage. Although these postings are not created by the profile owner, they often leave behind cues about the profile owner. Friends can post messages directly related to the profile owner or messages that indirectly contain elements of behavioral residue of the profile owner. This behavioral residue affects the judgment of the target person by the observer (Brunswik, 1956; Gosling et al., 2002; Walther & Parks, 2002). Moreover, these elements cannot be manipulated by the target person (although messages can be removed) and may therefore be judged as more reliable by the observer than self-generated cues (Walther & Parks, 2002).
Empirical research has shown that the tone of these wall postings affect the judgements of the profile owner by the observer (Walther et al., 2008). Results have demonstrated that favorable statements positively influenced the perceived task and social attractiveness of the profile owner, while negative statements decreased social attractiveness. Especially in early adolescence, adolescents are extremely sensitive to opinions of friends and place high value in what their friends say about others (Constanzo & Shaw, 1966; Elkind & Bowen, 1979). Therefore, our second hypothesis is:
H2: There will be a positive effect of the valence of the wall postings on the observers' perception of a) social attraction, b) physical attraction, and c) extraversion of the profile owner.
Number of friends
Finally, we examine the effect of a system-generated source of information: the number of friends of the profile owner. On all SNS, the number of friends is automatically displayed on the profile owner's web page. The number of friends reflects the size of one's network and therefore can act as an indicator for sociometric popularity (Tong et al., 2008). Research has found a positive relationship between the number of friends someone has on a SNS and perceptions of attractiveness of that person (Kleck et al., 2007; Utz, 2010). In the study of Kleck et al. (2007) the number of friends, 15 and 82 versus 261, positively affected impressions of popularity, pleasantness, attractiveness, and self-confidence. Utz (2010) found an interaction effect between the number of friends (82 versus 382) and the extraversion of friends. When the profile owner had introverted friends, the number of friends did significantly alter her social attractiveness. However, SNS became more popular the last few years, and therefore, the mean number of friends is increasing (Walther, et al., 2008). Therefore, Tong et al. (2008) investigated if an inverted U-shaped curvilinear relationship between the number of friends and perceptions of attractiveness existed. Their results showed that perceptions of attractiveness were highest for profile owners with a medium number of friends (i.e., 302) and lower for profile owners with either a small number of friends (i.e., 102), or an excessive amount of friends (i.e., 902) (Tong et al., 2008).
It is unclear if the number of friends will have the same effect on early adolescents' impressions. In early adolescence, traditional family relationships become less important, and a network of friends becomes increasingly important. In these first phases of relationship development, the quantity of adolescents' peer relationships if often judged to be more important than the quality of these friendships (Harter, 1999). The concept of friends on SNS, of course, is different than it is in real life. Friends on SNS relate more to the size of one's entire extended network than that it reflects true friendship (Ellison et al., 2006; Vanden Bogart, 2006). For early adolescents, though, size does matter and the entire size of their network may serve as a cue to attractiveness and popularity. Therefore, our third hypothesis is:
H3: There will be a positive effect of the number of friends the profile owner has on the observers' perception of a) social attraction, b) physical attraction, and c) extraversion of the profile owner.
Respondents and procedure
A total of 559 high school students between 12 and 16 years of age participated in the experiment. Respondents were gathered from four different high schools in an urban area in the Netherlands. In each of the four high schools, respondents were asked to come to the computer classrooms in small groups. They were distributed among the classroom in such a way that it was impossible for them to look at each other's screens. All information was delivered on screen. Respondents were told that they were about to visit an online profile of a girl named Eline and that they were to form an impression of Eline based on her profile. Every participant was then assigned to one of the 12 experimental conditions, each showing a mock-up profile of the most popular SNS in the Netherlands, Hyves1 . Respondents could look at the profile as long as they wanted. Once they had formed an impression of the profile owner, respondents could click through to the posttest questionnaire. After removing respondents who were not Hyves members, 495 respondents remained in the sample, of which 56% were girls. Average age of the sample was 13.58 (SD = 1.00). On average, respondents were active on Hyves about 5 days a week (M = 4.79; SD = 2.28). On an active day, they spent on average 1 hour and 33 minutes on Hyves (M = 93.14 minutes; SD = 116.66).
The experiment had a 3 (number of friends: 6, 269, 665) × 2 (physical attractiveness friends: unattractive, attractive) × 2 (Friends' messages: positive, negative) design. Except for the experimental manipulations, all information on the profiles was consistent amongst the 12 conditions. All profiles contained personal information about the profile owner (e.g., name, age, hobbies, and personal interests), a photo of the profile owner, and a number of messages (see Figure 1). The owner of the profile was Eline, a fictitious 14-year-old girl. We chose to keep the gender of the profile owner constant, because introducing gender as an additional factor would complicate the design and would reduce statistical power because of the doubling of experimental conditions. All profiles showed nine photos of friends: seven girls and two boys. This division reflects the fact that early-adolescent girls usually have more same-sex friends than cross-sex friends (Buhrmester & Furman, 1987).
Number of friends. To determine what a low, average and high number of friends is for adolescents to have on SNS, we examined 280 random profiles of Dutch adolescents between 12 and 15 years of age on Hyves. These adolescents had an average number of friends of 269 (M = 269; SD = 132). We chose this number as the average number of friends in our study. For the large number of friends, we added three times the standard deviation to the average, which is 665 friends. For the small number of friends, we chose nine friends. Hyves always shows a random selection of nine friends on a person's homepage. Having fewer than nine friends results in a different lay-out, adding a possible disturbing factor to the research.
Physical attractiveness friends. To select photos of friends that adolescents deem attractive and unattractive, a selection of 46 photos was pretested by 23 adolescents. The photos were obtained through a French SNS (Amiz), and through rating websites (e.g., hototnot.com). Respondents rated the attractiveness of the adolescents on these photos on a scale from (0) very unattractive to (10) very attractive. The nine photos that were judged to be most attractive (M = 7.31) and the nine photos that were rated most unattractive (M = 3.47) were chosen as the friends that were shown on the profile. One photo of a girl that was rated average (M = 5.82) was chosen to represent the profile owner.
Friends' messages. The 23 adolescents also rated 15 different messages on a scale from (1) this wall posting is a very negative statement to (7) this wall posting is a very positive statement. The two messages that were rated most positive (M = 6.64) and the two messages that were rated most negative (M = 2.00) were chosen as wall postings on Eline's profile. The two positive wall posting were the following: (translated from Dutch; excluding layout & formatting)
“ELINE PARTYGIRL!! Was superfantasticamazing y’day! You’re fun to go out with! Never ever had so much fun! Shall we go again soon?!?!??!”
“Hey Girl. Thanks for your sweet reaction. Im really glad having a friend like U. you’re right, it’ll work out soon, I heard you tried to get everyone there. You’re the best! LY4E”
The negative wall postings were:
“Could you perhaps warn me when you pull a trick like that…?! it was def NOT funny. I thought we had a deal?
“ELINE. How was yesterday? You didn't go out with him again, did you? He is so pathetic.”
Our measures of social and physical attractiveness were based McCroskey and McCain's Measurement of Interpersonal Attraction (1974). The six items measuring social attraction were: “Eline seems pleasant to be with,”“I think Eline could be a friend of mine,”“Eline would fit well in my circle of friends,”“I would like to have a friendly chat with Eline,”“Eline and I could establish a personal friendship with each other,” and “I would like to meet and talk to Eline in person.” The four items measuring physical attraction were “Eline is pretty,”“Eline is very sexy looking,”“I find Eline physically attractive,” and “I like the way Eline looks.” The response categories ranged from (1) completely disagree to (5) completely agree. Cronbach's alpha was .89 for both social (M = 2.53; SD = 0.84) and physical attractiveness (M = 2.32; SD = 0.88).
Extraversion was measured with three bipolar adjectives developed by McCroskey, Hamilton, & Weiner (1974). Respondents were asked to rate Eline on a 5-point scale on the following items: “timid—bold”, “silent—talkative” en “quiet—verbal.” Cronbach's alpha was .80 (M = 3.38; SD = .95). Finally, the respondents were asked if they would consider adding Eline as a friend: (1) yes, (2) don't know or (3) no (M = 1.72; SD = 0.69).
First, to check the manipulation of friends' attractiveness, respondents were asked to rate the attractiveness of Eline's friends on a 5-point scale from (1) very unattractive to (5) very attractive. The manipulation succeeded. Respondents who saw a profile with unattractive friends indeed rated these friends as being less attractive (M = 1.95, SD = 0.80) than respondents who saw a profile with attractive friends (M = 3.33, SD = 0.93), F(1,399) = 256.64, p < .001, η2 = .39. Second, respondents were asked to rate friends' wall postings on a scale from (1) very negative to (5) very positive. Respondents who were shown negative wall postings rated these more negative (M = 1.89, SD = 0.97) than respondents who were shown positive wall postings (M = 4.53, SD = 0.78), F(1,459) = 1042.02, p < .001, η2 = .69.
Third, respondents were asked if they thought Eline had (1) few friends, (2) an average number of friends, or (3) a lot of friends. The perception of the number of friends differed significantly among the three conditions, F(2,409) = 345.38, p < .001, η2 = .63 (9 friends: M = 1.13, SD = 0.41; 269 friends: M = 2.13, SD = 0.63; 665 friends: M = 2.74, SD = 0.52). Finally, the average number of SNS friends for our sample was 233 (M = 233.48, SD = 163.29). Three respondents (0.6%) had 9 friends or less, while 13 respondents had more than 665 friends (2.6%). Therefore, our manipulation of the number of friends (9, 269, or 665) closely matched the distribution of friends in our sample.
The first hypothesis predicted a positive effect of online friends' physical attractiveness on the observers' perception of attractiveness of the profile owner. To test this hypothesis, we conducted a MANOVA with social attractiveness, physical attractiveness, extraversion, and the likelihood of adding Eline as a friend as dependent variables, and physical attractiveness of online friends as the independent variable. The analysis revealed an overall effect of physical attractiveness of online friends, F(4,486) = 4.46, p = .002, η2 = .04. Specific analysis showed significant effects of physical attractiveness of online friends on perceptions of physical and social attractiveness. Respondents who were shown photos of attractive friends found Eline to be more physically attractive (M = 2.48, SD = 0.89) than respondents who were shown photos of unattractive friends (M = 2.16, SD = 0.83), F(1,489) = 17.38, p < .001, η2 = .03. No significant difference were found for social attractiveness, p = .073 , extraversion, p = .269 or the likelihood of adding Eline as a friend, p = .575.
The second hypothesis stated that there were would be a positive effect of the valence of the wall postings on the observers' perception of attractiveness of the profile owner. The analysis revealed an overall effect of physical attractiveness of online friends, F(4,486) = 9.89, p < .001, η2 = .07. There was a significant effect of message valance on perceptions of social attractiveness. Respondents who were shown negative messages (M = 2.34, SD = 0.82) liked Eline less than those who were shown positive messages (M = 2.71, SD = 0.82), F(1,489) = 25.02, p < .001, η2 = .05. An effect was also found for extraversion, albeit in the opposite direction as stated in the hypothesis. Respondents who were shown negative wall posting found Eline to be more extraverted (M = 3.52, SD = 0.99) than those who were shown positive messages (M = 3.24, SD = 0.88), F(1,489) = 11.27, p < .001,η2 = .02. In line with our hypothesis, respondents who were shown negative messages were less apt (M = 1.63, SD = 0.69) to add Eline as a friend than those who were shown positive messages (M = 1.80, SD = 0.68), F(1,489) = 7.93, p = .001, η2 = .02. Finally, the effect of message valance on perceptions of physical attractiveness was not significant, F(1,489) = 3.51, p = .062, η2 = .01.
The third hypothesis predicted a positive effect of the number of online friends on the observers' perception of attractiveness of the profile owner. The main effect of the number of friends was not significant, F(4,486) = 1.60, p = .122, η2 = .01. There was, however, a small effect of number of online friends on perceptions of extraversion. Respondents who thought that Eline had 665 friends, perceived Eline as more extraverted (M = 3.52, SD = 0.98) than those who thought Eline had 269 (M = 3.36, SD = 0.99) or 9 (M = 3.25, SD = 0.85) friends F(2,491) = 2.68, p = .048, η2 = .01. For the other variables, no differences the three conditions were observed.
Results of the MANOVA showed a significant interaction effect between message valence and physical attractiveness of online friends, F(4,486) = 2.97, p = .019, η2 = .01. First, there was a weak effect of message valence and physical attractiveness of online friends on perceived physical attractiveness of the profile owner, F(1,491) = 3.39, p = .067, η2 = .01. Message valence moderates the effect of friends' physical attractiveness on the perceived physical attractiveness of the profile owner (see Table 1). Although the main effect was not significant, the post hoc analyses showed that for respondents who were shown unattractive friends, there was no effect of message valence on the physical attractiveness rating of the profile owner, t(254) = 0.02, p = .981. However, for respondents who were shown attractive friends, positive messages enhanced perceptions of physical attractiveness of the profile owner, t(239) = 2.45, p = .015.
Table 1. Means and standard deviations of physical attractiveness and likelihood of adding Eline as a friend for unattractive vs. unattractive friends and negative vs. positive wall postings
Add as friend
Note. Means with different superscripts differ significantly (within columns) at p < .05.
A similar interaction effect was found for the likelihood of adding Eline as a friend, F(1,491) = 5.17, p = .023, η2 = .01. When respondents were shown unattractive friends, message valence had no influence on the likelihood of adding Eline as a friend, t(254) = 0.41, p = .684. However, respondents who were shown attractive friends, were more likely to add Eline as a friend when they were shown positive wall postings (see Table 1), t(239) = 3.37, p < .001. The result revealed no other significant two-way or three-way interaction effects.
Finally, we controlled for the effect of gender, age, and education level. First, students were divided in two education levels: lower secondary professional education and precollege students. Lower secondary professional education prepares for further vocational training or directly for lower-skilled jobs. Precollege students are being prepared for college. The first are lower-educated students, the latter are higher-educated students. There was a main effect of education level, F(4,468) = 5.92, p < .001, η2 = .05. Specifically, lower-educated students (M = 2.16, SD = 0.84) found Eline less attractive than higher-educated students (M = 2.54, SD = 0.88), F(1,471) = 19.44, p < .001, η2 = .04. Moreover, lower-educated students (M = 3.29, SD = 0.97) found Eline to be less extraverted than higher-educated students (M = 3.51, SD = 0.91), F(1,471) = 7.09, p = .008, η2 = .02.
There was a main effect of respondent's gender, F(4,468) = 2.52, p = .041, η2 = .02. Specifically, boys (M = 2.21, SD = 0.93) found Eline less attractive than girls (M = 2.40, SD = 0.82), F(1,471) = 8.29, p = .004, η2 = .02. Finally, there was a main effect for respondent's age, F(4,474) = 3.70, p = .006, η2 = .02. Raw correlations showed that younger adolescents found Eline to be more socially attractive (r = −.10, p = .032), physically attractive (r = −.10, p = .030), and extraverted (r = −.11, p = .015) than older adolescents, and were more likely to add Eline as a friend (r = −.11, p = .014). No interaction effects between education level, gender, age and the three experimental conditions were observed.
The aim of this study was to compare the effect of other-generated and system-generated cues on SNS on the perceived attractiveness of a profile owner. Specifically, we investigated the effects of friends' physical attractiveness (as can be judged from the photos of those friends that are shown on a profile), tone of wall postings, and the number of friends. In general, our hypotheses were confirmed. Physical attractiveness of friends mainly influenced attractiveness of the profile owner, and positive wall posting led to greater social attractiveness of the profile owner in comparison to negative wall postings. Finally, the number of friends of a profile owner positively influenced perceived extraversion, but no relationship was found for the other measures of attractiveness.
In agreement with earlier research (Walther, et al., 2008; Geiselman, et al., 1984), we found that friends' physical attractiveness positively affected the observers' perceived physical attractiveness of the profile owner. Thus, when a profile owner's friends are more physically attractive, as can be judged from the friends' photos that are shown on that member's profile, they are judged to be more physically attractive as well. In contrast, when friends' photos are less attractive, a profile owner is also seen as less physically attractive. This assimilation effect is well known in interpersonal perception (Melamed & Moss, 1975), but the effects seem to hold even when the relationship between the target person and his friends is expressed only through photos on an online profile.
Wall postings on SNS have an effect on a profile owner's perceived social attractiveness and the likelihood to be added as a friend. The overall increase in attractiveness is in line with the warranting principle (Walther & Parks, 2002) that claims made by others about the target person are used as an indicator for attractiveness of the target person. Furthermore, the direction of the relationship between wall postings and observers' perceived extraversion of the profile owner was contrary to what we expected. When wall postings were negative, Eline was actually rated as more extraverted than when wall postings were positive. An explanation for this finding may lie in the content of the wall postings used in our experiment. One negative statement was about a trick Eline pulled to a friend of hers. The other was about Eline going out with a boy who was disliked by her friend. Although these are not positive acts, they may be seen an extraverted. Taken together, these results about the friends' messages confirm that early adolescents are sensitive to opinions of friends and place high value in what adolescents say about others (Elkind & Bowen, 1979; Constanzo & Shaw, 1966). Hence, these results implicate that people should be more aware of the impact of those short and informal messages posted on one's profile. It might be worthwhile to carefully manage one's self-presentation by strategically removing certain wall postings.
Furthermore, the number of friends of a profile owner affected perceived extraversion, which is in line with earlier research (Tong et al., 2008). Contrary to our expectations, the number of friends had no effect on perceived social and physical attractiveness. Although the size of one's peer network increases significantly in early adolescence (Hartup, 1996), adolescents with a smaller social network are not judged to be less attractive, at least not online. There were also no indicators for an inverted U relationship between number of friends and social attractiveness. For early adolescents, therefore, there appears to be no limit to the number of friends that is considered normal on SNS. This might be explained by the different value between friends on a SNS and friends in real life. Friends on social networks are more related to the size of one's entire network and more seen as superficial, whereas friends in real live reflect more true friendships (Ellison, et al., 2006). These results suggest that the number of friends on a SNS is not an indicator for perceived popularity.
These findings, however, warrant further investigation. First, although our choice of a large number of friends (665) was more than two standard deviations away from the mean in this sample, it might still not be considered an outlier by adolescents. Therefore, in further studies, the number of friends might be increased even further. Second, in line with the explanation of Tong et al. (2008), the presence of other-generated cues about the profile owner might have a stronger relationship with attractiveness judgments than system-generated cues. It is possible that in the absence of other-generated cues such as friends' photos and wall postings, the number of friends a person has may be a potent cue in the determination of attractiveness, because people rely on any information available, however limited, to form impressions of others (Walther & Parks, 2002). Third, it is unclear if a ceiling effect for the number of friends exists at all, even for adults. The results of Tong et al. suggested an inverted U relationship, but the decline in attractiveness and extraversion they observed was not statistically significant.
Finally, we found two small interaction effects between friends' messages and physical attractiveness of friends. First, results demonstrated that message valence strengthens the effect of friends' physical attractiveness on the observers' perceived physical attractiveness of the profile owner. Furthermore, physical attractiveness of friends also interacts with message valence on the likelihood of adding someone as a friend. Although these results are weak, they suggest that wanting to add someone as a friend is highest when someone has both attractive friends and positive wall postings.
In sum, the majority of our expectations in this study were confirmed. Friends' physical attractiveness did positively affect both physical attraction of the profile owner. Friends' messages also affected the physical and social attractiveness of the profile owner. However, the strongest effect was on social attractiveness of the profile owner. Different cues therefore influence different attractiveness dimensions. The number of friends only affected the perceived extraversion of the profile owner, but had no effect on social or physical attractiveness ratings. Overall, other-generated cues (friends' wall postings and friends' physical attractiveness) were a stronger predictor of attractiveness than system-generated cues (number of friends).
Implications of this study and future research
Our study has several implications for future theory and research. First, our study investigated different sources of information on a SNS. Only one study (Utz, 2010) has investigated the effect of both other-generated cues and system-generated cues on attractiveness in a single study. However, that study used the extraversion of target's friends as other-generated cues. In this study other-generated cues are wall postings and friends' physical attractiveness, while system-generated cues refer to the number of friends. Contrary to Utz (2010), our findings demonstrated that other-generated cues are more important indicators of attractiveness of the profile owner than system-generated cues. In agreement with the Brunswikian lens model (Brunswik, 1956), adolescents indeed form impressions based on the surrounding elements and behavioral residue of the profile owners' SNS. However, the system-generated cues (i.e., the number of friends) did not affect the observers' impressions of the profile owner. Future research should further explore this finding by investigating if the number of friends is indeed overshadowed by the physical information also available on the SNS.
This study was the first to investigate the effects of self- and other-generated cues on impression formation among early adolescents. Our results showed that early adolescents too base their impressions for a part on this warranting information. In early adolescence, adolescents are very preoccupied with how they are perceived by others (Elkind & Bowen, 1979). Therefore, the impressions others make of them based on their online profiles may especially affect adolescents' self-esteem and well-being. Previous research showed that especially the reactions adolescents receive on their online profiles positively affect adolescents' self-esteem and well-being (Valkenburg, Peter, & Schouten, 2006). Our findings implicate that profile owners on a SNS cannot fully control their image and therefore should be aware of the information generated by others. This other-generated information might affect the impressions formed by others and therefore could indirectly affect their self-esteem and well-being. Future research should investigate if and how these impressions affect adolescents' social development. For example, research could investigate how the impressions adolescents form of each other based on online profiles affect future interactions.
The important role of other-generated information also has implications for discussions of online privacy. Privacy may be defined as a users' ability to control impressions and manage social contexts (boyd & Ellison, 2008). Earlier research about the effects of information revealing on SNS and privacy concerns was mainly focused on the accessibility of self-generated information, such as the profile image, relationship status, and favorite music (e.g., Debatin, Lovejoy, Horn, & Hughes, 2009; Dwyer, Hiltz, & Passerini, 2007; Gross & Acquisti, 2005; Patchin & Hinduja, 2010). However, this study shows that system-generated cues and other-generated cues online may also be important determinants of social attraction. Contrary to offline other-generated information (i.e., behavioral residue), this online information is easily accessible for a receiver, while for a sender this information is difficult to manage online. Because profile owners only have limited control over these cues, the control over these cues is a privacy issue. Future research concerning privacy on SNS should also take system-generated and other-generated cues into account. Furthermore, adolescents should be aware that people also use these cues to form impressions. A practical suggestion would be to let profile owners themselves decide whether this information is shown on their online profiles. Currently, it is already possible to manage one's wall postings to some extent on most SNS, but must SNS do not allow control over whether profile owners want to show their number of friends.
Our study makes several additions to studies investigating the relationship between impression formation on SNS and interpersonal attraction. First, both other-generated cues and system-generated cues had never been investigated in a single study. By investigating these cues together in one study we were able to determine the strongest predictors of attractiveness and investigate possible interaction effects. Second, other and system-generated cues might affect social evaluation of a profile owner in different ways. By using three different dimensions of social evaluation, we were able to make some refinements to earlier studies about impression formation on SNS. This study shows that warranting information (Walther & Parks, 2002), non-self-generated cues about persons that are used to form impressions of others, are an important factor to take into account in future research online interaction. Especially, future research should investigate how these cues, and the impressions they convey, affect further interaction between individuals.
The authors would like to acknowledge and thank Lisanne de Geus and Bart Monné for creating the stimuli and collecting the data.
Marjolijn L. Antheunis is an Assistant Professor Social Aspects of New Media, Tilburg centre for Cognition and Communication at Tilburg University. Her research interests include social aspects of new media, the effects of online communication on friendship formation, and online impressions management and formation.
Address: Department of Communication and Information Sciences, Tilburg centre for Cognition and Communication, Tilburg University, Warandelaan 2, 5037 AB Tilburg, The Netherlands. Tel: +31 13 466 2971 Email: email@example.com.
Alexander P. Schouten is an Assistant Professor Information Systems at the VU University Amsterdam. His research focuses on the social aspects of information technology, specifically the use of social media, online collaboration, and online impression management and impression formation.
Address: Department of Knowledge, Information, and Networks, Business Administration, VU University Amsterdam. De Boelelaan 1105, 1081 HV, Amsterdam, The Netherlands. Email: firstname.lastname@example.org.