How Do Adolescents Evaluate Online Dietary Information in
a Healthy Living Blog?
The Role of Homophily, Physical Attractiveness, and Credibility Cues on
Source Likeability, Perceived Source Credibility and Message Evaluation
Pascale Kwakman 10530649
Supervised by: Dr. Corine Meppelink
Master’s Thesis
Graduate School of Communication Research Master Communication Science
Abstract
Online health seeking behaviour is becoming increasingly prevalent. Adolescents,
specifically, frequently search for dietary information using social networking sites, such as healthy-living blogs. Despite concerns about the accuracy of online health information, little is known of the heuristic mechanisms that are used by adolescents. Understanding these mechanisms can aid the development of successful online health campaigns. This study (N = 217) examined the effects of demographic homophily, attitudinal homophily, physical attractiveness and credibility cues on (1) source likeability, (2) perceived source credibility, and (3) the evaluation of a health message. In addition, the moderating roles of health literacy, personal relevance, and sex on the relationships of interest are examined. Analyses with structural equation modelling revealed that contrary to our predictions, neither demographic homophily, nor the presence of credibility cues had a significant effect on adolescents’ perceived source credibility. Similarly, attitudinal homophily had no significant effect on either source likeability nor on perceived source credibility. Demographic homophily was found to negatively influence source likeability; an effect that can most likely be explained by the blogger’s age. Older bloggers were more likely perceived as authorities, and therefore better liked and accredited with greater credibility. Source likeability and perceived source credibility were identified as the prime drivers of message evaluation, with both factors showing significant positive effects on adolescents’ message evaluation. Moreover, a halo-effect was found, indicating that well liked bloggers are also perceived as being more credible sources. Lastly, these effects may differ substantially for male and female adolescents, and for those with different levels of health literacy and personal relevance.
Keywords: online health information, healthy-living blogs, dietary information, homophily, physical attractiveness, credibility, source likeability, message evaluation, adolescent health behaviour
How Do Adolescents Evaluate Online Dietary Information in a Healthy Living Blog? The Role of Homophily, Physical Attractiveness, and Credibility Cues on Source Likeability,
Perceived Source Credibility and Message Evaluation
With the emergence of Web 2.0, the dynamics of health information sharing drastically changed (Eysenbach, 2008; Macario et al., 2011). Instead of purely receiving health information from health authorities, the majority of people now use social networking sites (SNSs), blogs (Miller & Pole, 2010; Carrotte, Vella & Lim, 2015) and vlogs (Huh et al., 2014) to share, locate and evaluate health information (Tennant et al., 2015). Through these SNSs, health seekers are frequently exposed to social influencers: (lay) opinion leaders, who share health information through online narratives (Dodd, 2018; Liu et al., 2017; Macario et al., 2010).
The lifestyle-ization of health information has shifted the focus from evidence-based information to unsubstantiated emotional appeals (Lewis, 2006), which has led to increased medical and scientific concern about online health misinformation (“Pew Research Center”, 2000). Health seekers are often exposed to inaccurate and contradictory messages, which may lead to the establishment of misguided health beliefs (Lee & Cho, 2017). Such message’s high accessibility, makes it crucial to understand message-receivers’ evaluation of influencer-based health information (Schönfeldt & Hall, 2012).
Adolescents, specifically, are known to frequently use the Internet as their main source of health information (Borzekowski & Rickert, 2001; Gray et al., 2005a, 2005b). In fact, 75% of adolescents indicate to have used the internet as a health information source at least once (Rideout, 2001). Since peer-generated health information, in the form of ‘healthy-living’ blogs (HLBs), is ubiquitous on SNSs (Boepple & Thompson, 2014; Lu, 2013), it is not strange that 34% of adolescents report to have accessed the internet to search for diet, nutrition or exercise information (Borzekowski & Rickert, 2001). Nutrition and exercise
information on SNSs has, however, been criticized for its inaccuracy and the potentially harmful health outcomes on its audience (Carrotte et al., 2015; Myrick & Erlichman, 2019). For instance, dietary restriction and excessive exercise are often encouraged on HLBs and other SNSs (Boepple & Thompson, 2014; Holland & Tiggemann, 2017).i More importantly, adolescents’ vulnerability to peer pressure regarding health-risk behaviour, makes them a critical target group to study the evaluation of dietary information on HLBs (La Greca, Prinstein & Fetter, 2001; Paxton et al., 1999).
Building on the reasoning above, there is a need to understand the heuristic cues used by adolescents to assess peer-generated online dietary information in HLBs. While few studies have identified heuristics that explain adults’ (Myrick & Erlichman, 2019; Wang et al., 2008) and university students’ (Young, 2015) acceptance of such messages; to my
knowledge, no studies have been performed that explore adolescents’ dependency on heuristic cues during the evaluation of peer-generated HLBs, in combination with the receiver
characteristics that may affect this evaluation process.
Understanding the heuristic cues and receiver characteristics at play in adolescents’ evaluation of influencer-based health information, could aid in designing effective expert-based health campaigns targeted at those who are at risk to adopt the beliefs of healthy-living bloggers. Such campaigns may in turn lead to an increase in well-informed health behaviours. It is imperative for health professionals to include evidence-based mechanisms in their
campaigns, for if they do not efficiently penetrate the noise of misinformation spread by lay influencers, no real change is expected to occur (Macario et al., 2010). Given these points, the research questions of this study are: (1) Which heuristic mechanisms explain the effect of peer-generated dietary information in a HLB on adolescents’ evaluations of these messages? (2) Which adolescents are more highly susceptible to positively evaluate peer-generated dietary information in a HLB?
Theoretical Background Source Perceptions
Perceived source credibility.
Perceived source credibility is considered one of the basic elements of effective interpersonal communication (McCroskey, Richmond & Daly, 1975). In short, credibility relates to the believability of a source. The source-credibility model (Hovland, Janis & Kelly, 1953) theorizes that credibility has two main dimensions: expertise and trustworthiness. Whereas expertise relates to a source’s qualifications and/or ability to know the truth about a topic, trustworthiness relates to the reader’s perception of the source’s motivation to tell the truth (O’Keefe, 2002; Petty & Brinol, 2010). In the case of peer-generated health information, expertise is often lacking, while trustworthiness could easily manifest itself among the readers of these messages due to the perception that influencers are peers who produce authentic and trustworthy content (Colliander & Dahlén, 2011; Domingues Aguiar & Van Reijmersdal, 2018). This perception, however, also heightens the risk of misinterpreting incorrect peer-sourced health information as valid information.
Numerous credibility cues exist, such as the provision of credentials, references to an institution, or official seals of approval (e.g. the Health on the Net Foundation’s HONcode)ii (Eastin, 2001; Neubaum & Krämer, 2015). Such cues can be used while evaluating messages: a process called heuristic processing. Following both Chaiken and Eagly’s (1993) heuristic systematic model and Petty and Cacioppo’s (1986) elaboration likelihood model, people either process messages systematically or heuristically. Whereas systematic processing involves a critical analysis of the arguments given in a message, heuristic processing relies entirely on using message cues as a shortcut to judge the quality of a message (Chaiken & Eagly, 1993; Petty & Cacioppo, 1986). Given their vulnerability to social norms, adolescents are known to frequently use heuristics during message processing (Reyna & Farley, 2006). It
is thus proposed that adolescents will have a stronger perception of source credibility when being exposed to a HLB that contains credibility cues, compared to a HLB without credibility cues (H1).
Source likeability and physical attractiveness.
Source likeability originates from McGuire’s (1985) source-attractiveness model, which postulates that message effectiveness is based on three main messenger-based mechanisms that determine a source’s attractiveness: familiarity, likeability and similarity. Erdogan (1999) defines likeability as “affection for the source as a result of the source’s physical appearance and behaviour” (p. 299). In addition, physical appearance is not only related to physical attractiveness, rather it includes any number of desirable characteristics (e.g. athletic prowess or intellectual skills; Erdogan, 1999).
Erdogan’s (1999) description clarifies that source likeability encompasses positive evaluations of an individual, as well as social attraction. For the purpose of this study, a distinction is made between physical attractiveness of the source, and source likeability, since the former may positively influence the latter (Chaiken, 1986). This presumption seems logical, since studies have frequently shown a relationship between a person’s physical attractiveness and their likeability (Joseph, 1982; Timmerman & Hewitt, 1980). A similar effect is predicted to occur in this study: perceived attractiveness of the source is positively associated with source likeability (H2a).
Another effect that could possibly exhibit itself when strong likeability is experienced, is the halo-effect. The halo-effect implies that people tend to assume that those who rank high on one dimension, such as likeability, also excel on other dimensions (Erdogan, 1999;
Ohanian, 1990). McCroskey and Richmond (1996) explain that “when we like a source, we are more likely to perceive him or her to be competent and trustworthy” (p. 117). Influencers often do not have any qualifications in nutrition, yet they are frequently used as spokespersons
for health campaigns and relied upon for advice by the public (“Soa Aids Nederland”, 2016; “Voedingscentrum”, 2018). Liking influencers could thus lead to the false belief that they are a credible source of health information. Based on the above reasoning, it is expected that perceived source likeability is positively associated with perceived source credibility (H2b).
Homophily.
Briefly explained, homophily constitutes source–receiver similarity. Research shows that the influence of peers on adolescent health behaviour is often stronger when adolescents perceive a source to share similar physical attributes, values, beliefs, experiences or
personality traits (McCroskey, Richmond & Daly, 1975; Shoham et al., 2012; Simpkins, Schaefer, Price & Vest, 2013; Zhang et al., 2015). Overall, homophily can either be based on similarities in values and beliefs (i.e. attitudinal homophily), or on demographic similarities (i.e. demographic homophily; Simons, Berkowitz & Moyer, 1970). O’Keefe (2002) explains that only source–receiver similarities or dissimilarities that are relevant to the receiver, can influence judgements of the communicator’s likeability and credibility. Empirical research, however, remains inconclusive about the level of relevance ascribed to attitudinal and demographic homophily; for different studies render different results.
To be specific, both Lu (2013) and Wang and colleagues (2008) found a significant effect of attitudinal homophily on perceptions of information credibility. Hu and Sundar (2010), in contrast, found no such effect. Moreover, positive effects of demographic
homophily on adolescents’ physical activity and dietary behaviours were found by Shoham and colleagues (2012), Simpkins and colleagues (2013), and Zhang and colleagues (2015). These effects could, however, not be replicated by Sakib, Zolfagharian and Yazdanparast (2019). These inconclusive results regarding demographic and attitudinal homophily need to be studied further, specifically in a relevant adolescent sample.
Demographic homophily.
Many adolescents follow social influencers online (Silverman, 2016), a phenomenon that may be due to adolescents’ perception that these influencers are demographically similar to them (Domingues Aguiar & Van Reijmersdal, 2018). This leads us to hypothesize that demographic homophily is positively associated with increased source likeability (H3a). Moreover, demographic homophily is predicted to positively influence perceived source credibility (H3b), since due to their relatability, influencers are often perceived as trustworthy sources (Domingues Aguiar & Van Reijmersdal, 2018).
Attitudinal homophily.
Attitudinal homophily is described in the literature as a stronger driver of source liking and source credibility, compared to demographic homophily, because of its greater relevance to the receiver (O’Keefe, 2002; Simons et al., 1970). Zchorlich and colleagues (2015) argue that social media users are more likely to trust information that corroborates their own
experiences and beliefs. Moreover, attitudinal homophily may indirectly influence credibility perceptions through liking of the source (Ma & Atkin, 2017; O’Keefe, 2002). Since the literature provides no clear direction of attitudinal homophily’s effect on message evaluation, the following research questions are posed: Does attitudinal homophily indirectly influence adolescents’ evaluation of the nutritional information in the HLB through increased liking of the source (RQ3a)? Does attitudinal homophily indirectly influence adolescents’ evaluation of the nutritional information in the HLB through increased perceived credibility of the source (RQ3b)?
Receiver Characteristics Health literacy.
Health literacy, or eHealth literacy, is one of the key elements of health behaviour decision-making (Fleary, Joseph & Pappagianopoulos, 2018). Following Norman and Skinner
(2006), eHealth literacy can be defined as “the ability to seek, find, understand, and appraise health information from electronic sources and apply the knowledge gained to addressing or solving a health problem” (e9). Little research on adolescent health information evaluation has considered eHealth literacy, even though adolescents often display autonomy regarding their own health behaviour (Fleary et al., 2018; Ghaddar et al., 2012).
Many adolescents indicate having difficulties in accessing, understanding and
evaluating the trustworthiness of online health information (Gray et al., 2005a, 2005b). These adolescents may not know that key information is missing, or that biased information is presented (Cline & Haynes, 2001). Considering these difficulties, low health literacy impedes the critical evaluation of health information; hence healthy-living bloggers could be
considered credible sources of health information, even when no features of credibility are present. Because of their low health literacy, many adolescents might depend on heuristic cues, rather than on a critical analysis of the message (Chaiken & Eagly, 1993; Petty & Cacioppo, 1986). It is therefore proposed that adolescents with lower health literacy will have a greater perception of source credibility when no credibility cues are present, compared to adolescents with higher health literacy (H4a). It is also hypothesized that the effects of heuristic cues (i.e. demographic homophily, attitudinal homophily and physical
attractiveness) on message evaluation, mediated by likeability and perceived source credibility, will be stronger for adolescents with low health literacy than those with high health literacy (H4b).
Personal relevance.
Similar to health literacy, the amount of personal relevance ascribed to a certain health topic can influence the type of information processing (i.e. heuristic or systematic; Petty, Cacioppo & Goldman, 1981). Research shows that when personal relevance is low, message-receivers rely less on the critical scrutiny of message elements, and rely more on heuristic cues
such as homophily, source attractiveness and source likeability (Petty & Cacioppo, 1979a, 1979b). It is therefore hypothesized that the effects of heuristic cues (i.e. demographic
homophily, attitudinal homophily and physical attractiveness) on message evaluation, mediated by likeability and perceived source credibility, will be stronger for adolescents with low
personal relevance toward a healthy diet, compared to those with high personal relevance toward a healthy diet (H4c).
Sex.
Sex is also perceived to have a moderating effect on adolescents’ dependency on heuristic cues. For one thing, women tend to be more relationship-oriented compared to men (Cross & Madson, 1997), and they mainly engage in friendships with demographically similar others (McPherson, Smith-Lovin & Cook, 2001; Ibarra, 1992). Moreover, females are more likely to comply with the request of others with whom they share seemingly insignificant similarities (i.e. shared names, birthdays or fingerprints; Burger et al., 2004). Therefore, it seems logical to assume that the effect of demographic and attitudinal homophily on source likeability and perceived source credibility will be stronger for female adolescents compared to male adolescents (H4d). See Figure 1 for the hypothesized model as discussed above.
Figure 1. The influence of heuristic cues on the evaluation of health information through the mediators source likeability and perceived source credibility.
Method Study Design, Procedure and Participants
A 2 (source credibility: low vs. high) x 2 (demographic homophily: low vs. high) between-subjects factorial experiment was conducted. Two secondary schools participated in the current study. Active informed consent was obtained from the heads of the schools, and passive informed consent was obtained from the parents.
A total of six junior classes (n = 181) and two senior classes (n = 36) were allocated to one in four conditions, while trying to account for an equal distribution of participants across the conditions (N = 217). One lecturer and one researcher administered the questionnaire to the student during scheduled classroom periods. Participants were informed that they were free to end their participation in the experiment at any time, that their participation would be completely anonymous, and that the information would be handled confidentially. After participants provided active informed consent, they were asked several demographic
questions. Subsequently, the adolescents were told to open a link from their e-mail, read the blog post ‘three suggestions for a healthy diet’, and to fill out the rest of the questionnaire (see Appendix B).
Table 1 below represents details of the student characteristics in each condition. Significant differences between each group were detected for participants’ age and education. The effect size of these differences is considered small to medium (Cohen, 1992). A residual analysis (Sharpe, 2015) showed that students with a middle educational level are
underrepresented in condition one (i.e. high demographic homophily, low credibility), and overrepresented in condition 4 (i.e. low demographic homophily, high credibility). In addition, students with a low educational level are overrepresented in condition one and underrepresented in condition four. These variations in educational level must be considered during further analyses.
A pairwise comparison using Bonferroni correction, revealed a number of significant differences in age between the conditions (p < .001). The main difference lies between condition one (i.e. high homophily, low credibility), in comparison to the other conditions: participants in condition one are on average 0.87 to 1.23 years older compared to participants in the other conditions. This difference must be taken into account during further analyses. Gender, ethnicity and health influencer following did not differ significantly between groups. Stimulus Materials
All blogs and blog posts were created by using www.wix.com. The fictitious HLB closely resembled an actual HLB, and participants could actually navigate the blog (see Figure 2). The blog post used in this study provides three suggestions for a healthy diet: having a breakfast that is rich in protein and fibre, drinking more water, and substituting energy-dense foods for nutrient-dense foods. An example of the blog post stimulus can be seen on
https://degezondetoer.wixsite.com/blog/post/3-tips-voor-een-gezond-voedingspatroon.
Demographic homophily was manipulated by either showing participants a fictitious source of a similar age (i.e. 14 years old) with a similar sex and ethnicity, or by showing them a source that was much older (i.e. 43 years old), had a different ethnicity, and was of the opposite sex (Chen, Bell & Taylor, 2016). A photo of the source was shown on the HLB’s homepage, next to a short description of the blogger. This description contained the blogger’s interests (i.e. running and watching cooking shows) and his or her age. Bloggers could either be Caucasian, Middle Eastern or African American. Age, sex and ethnicity were conjointly used to manipulate demographic homophily, since their independent effects were not relevant to this study.
To manipulate source credibility, two cues were added to the blog post. Firstly, the beginning of the blog post stated that the information was written together with a (fictitious)
nutritionist from the Dutch Nutrition Centre, who was displayed in an image underneath this statement. Secondly, at the bottom of the blog post the statement: “this blog was written in collaboration with the Dutch Nutrition Centre and nutritionist Chantal de Bruijn”, and the logo of this organization were presented. Participants exposed to the low credibility conditions saw none of these cues.
In the end, twenty-four versions of a HLB were created. Blogs with a Caucasian, Middle Eastern, or African American ethnicity could all be written by a male or female blogger, of fourteen or forty-three years old, with or without credibility cues present. Hence, eight different versions for each ethnicity were created. Participants were randomly assigned to an ethnically different source in the low demographic homophily conditions, and were assigned to a blogger with the same ethnicity in the high demographic homophily conditions.
To ensure the quality of the different HLBs, an extensive pre-test was run (N = 23, Mage = 16, SDage = 1.73; see Appendix A).
Manipulation Check
Participants’ recall of the two credibility cues was checked by (1) asking participants to indicate whether or not the blogger collaborated with an organisation, and if so, which organisation (0 = no or I don’t know, 1 = yes, the organisation is named …); and (2) by asking them whether or not there was a collaboration with a certain person, and who this person was (0 = no or I don’t know, 1 = yes, the person is …).
Measures
Demographic homophily. To assess demographic homophily, Brown and Reingen’s (1987) method was used. Firstly, demographic questions were asked to assess respondents’ age, sex, country of origin, and their parents’ country of origin. Country of origin and
participants’ parents’ country of origin were used to measure participant ethnicity. When one of the parents was born in a country different from the Netherlands, the adolescent’s ethnicity
is based on the foreign-born parent. When parents were born in different foreign countries, the adolescent’s ethnicity is based on the mother’s country of origin (Geven, Johnsson, van Tube, 2017). Source–receiver similarity ratings were calculated for each of the three demographic variables. For age homophily, sex homophily and ethnic homophily, a score of 1 reflected similar source-receiver age groups, similar source-receiver sexes, and similar source-receiver ethnicities respectively. A score of 0 represented a strong difference in age, different sexes, and different source-receiver ethnicities respectively. Scores for age, sex and ethnic
homophily were combined in a summative scale (M = 1.25, SD = 1.31), with higher scores reflecting greater demographic homophily.
Attitudinal homophily. Attitudinal homophily was assessed using five five-point bipolar adjective scale items from McCroskey, McCroskey and Richmond (2006). Two sample items are: “. . . has thoughts and ideas that are similar to mine/has thoughts and ideas that are different from mine” and “. . . thinks like me/doesn’t think like me”. Since an initial reliability analysis indicated questionable reliability (α = .48), two items were removed from the scale (i.e. “does not care about the same things as I do/cares about the same things as I do” and “Is not like me/Is like me”).1 A mean score was calculated (α = .79, M = 2.66, SD = 0.84), with a higher score indicating greater attitudinal homophily.
Physical attractiveness. Four items from McCroskey and colleagues’ (2006)
interpersonal attraction scale were adopted to measure physical attraction of the blogger. This scale included items such as “I think the blogger is handsome” and “I think the blogger is good looking” (1 = strongly disagree, 5 = strongly agree). A mean score was calculated (α = .88, M = 2.60, SD = 0.99), with a higher score indicating higher perceived attractiveness of the source.
Likeability. Perceptions of source likeability were measured with three five-point bipolar adjective scale items adopted from Yilmaz and colleagues (2011), “not friendly – friendly”, “not likeable – likeable”, “cold – warm”). A mean score was calculated (α = .91, M = 3.85, SD = 0.84), with higher scores indicating a higher perception of source likeability.
Perceived source credibility. To assess perceived source credibility, three five-point bipolar adjective scale items from Ohanian’s (1990) celebrity endorser credibility scale were used. One adjective from the trustworthiness dimension was used (i.e. “untrustworthy – trustworthy”), in combination with two adjectives from the expertise dimension (i.e. “not an expert – expert”, “unqualified – qualified”). This combination was chosen, since three items from Ohanian’s (1990) trustworthiness dimension all translated to the same word in Dutch (i.e. betrouwbaar). A mean score was calculated (α = .82, M = 3.16, SD = 0.88), with a higher score indicating a higher perception of source credibility.
Health literacy. Norman and Skinner’s (2006) Electronic Health Literacy Scale (eHEALS) was used to assess participants’ (e-)health literacy levels. The eHEALS consists of eight items measured on a five-point Likert scale. Two sample items are: “I know what health resources are available on the Internet” and “I know how to use the health information I find on the Internet to help me” (1 = strongly disagree, 5 = strongly agree). A mean score was calculated (α = .84, M = 3.54, SD = 0.60), with a higher score indicating better health literacy.
Personal relevance. Personal relevance was measured with three items on a five-point Likert scale, based on Wilson’s (2007) description of the concept, yet adapted to represent the subject of the current study. Two sample items are: “How important is eating healthily to you?” (1 = very unimportant, 5 = very important), “On a daily basis, how much attention do you pay to eating healthily?” (1 = very little attention, 5 = very much attention). A mean score
was calculated (α = .74, M = 3.44, SD = 0.68), with a higher score indicating greater personal relevance.2
Sex. Participants’ sex was measured with one item: “What is your sex?” (0 = male, 1 = female, missing = I do not wish to answer).
Evaluation of health information. Evaluation of health information was measured based on Wang and colleagues (2008) five-item scale. Four items were used, including: “I believe that what the blogger said is true” and “The blogger gives good nutritional advice” (1 = strongly disagree, 5 = strongly agree). A higher score on evaluation of health information indicates a more positive evaluation of the source’s advice (α = .77, M = 3.84, SD = 0.68).3
Control variables. Age in years, and sex were assessed with one question and used as control variables. Furthermore, adolescents’ country of birth was measured (1 = Netherlands, 2 = Belgium, 3 = Turkey, 4 = Morocco, 5 = Germany, 6 = other, namely …), along with their father’s and mother’s country of birth (same answer categories). General social media usage was measured using two items from Rosen and colleagues (2013): “How often do you use SNSs?” (1 = never, 10 = all the time) and “On a typical day, how often do you use SNSs?” (1 = 1 – 30 minutes, 7 = more than 8 hours). Health source usage was measured by asking participants “Which sources do you use for diet, nutrition and exercise information?” (e.g. 1 = my brother(s)/sister(s)/niece(s)/nephew(s), 2 = friends; Borzekowski & Rickert, 2001; see Appendix B). Lastly, online following of health-influencers was measured by asking
participants “Do you follow one or more vloggers, bloggers or Instagrammers online? If so, who do you follow?” (see Appendix B).
2 After performing a CFA in AMOS, the last item “How relevant is the information in the blogpost to you?”, was
removed due to its low proportion of explained variance (R2 = .28). The reported mean and standard deviation
reflect the values of the new scale without this item.
3 A similar procedure as the one described above, was performed for the items “The information provided by the
blogger is useful” (R2 = .42), and " I would not recommend the advice of the blogger to a friend" (R2 = .32). The
Analytic Strategy
To test the specified model (Figure 1), a structural equation modelling (SEM) approach was adopted. SEM is known to be a technique which uses multivariate statistical analysis to evaluate the goodness of fit for complete models (Kline, 2011). Rather than focusing on individual effects, SEM can be used to test construct validity as well as the strength and directionality of the hypothesized relations (Anderson & Gerbing, 1988). Three frequently recommended fit indexes were used: the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the chi-square. Model fit is considered good with a CFI of at least .95, a RMSEA of .05 or lower, and a non-significant chi-square value (Kline, 2011).
First, a confirmatory factor analysis (CFA) was performed to test both convergent and discriminant validity of the measurement model. A path analysis was performed instead of a structural regression model, considering the complexity of the model and the small sample size (Wolf, Harrington, Clark & Miller, 2013). Each manifest factor represented in the model has been constructed by parcelling all indicators and creating a mean score for each factor, except demographic homophily. For this factor, a sum score was used, since each indicator’s score needs to be tallied to reflect the total amount of demographic homophily per participant. The first path analyses will not yet include the hypothesized moderation effects related to adolescents’ sex. For these moderation effects, a multi-group analysis will be performed on a path model (PM) that contains only those variables relevant to H4d: demographic homophily, attitudinal homophily, source likeability, perceived source credibility and evaluation of health information (Kline, 2011). Age and sex were included as control variables in the PMs, and age was included in the multi-group models (MGMs), since both variables correlated significantly with a predictor and an outcome variable (see Appendix C; Anderson et al., 1980).
Structural equation models were developed and analysed with AMOS 22, using Maximum Likelihood estimation. Since the data was not normally distributed, a bootstrap method was adopted to obtain a better approximation of the sampling distributions (Erceg-Hurn & Mirosevich, 2008). This method adhered 1000 bootstrap samples and a 95% bias-corrected confidence interval.
Results Manipulation Check
To assess whether adolescents perceived the presence or absence of credibility cues in the HLB as was intended, a Fisher-exact test of independence was performed with the recall of credibility cues (1 = yes, 0 = no or I don’t know) as the dependent variable and the type of HLB (low credibility vs. high credibility) as the independent variable.
Results indicated a significant difference in recall of credibility cues between the different HLBs (p < .001). Adolescents exposed to a HLB with credibility cues more often answered that the HLB featured a collaboration with a nutritionist, or with Chantal de Bruijn (n = 68), than not (n = 53). Similarly, adolescents exposed to a HLB without credibility cues more often answered to not recall any collaboration with a person (n = 94) or an organization (n = 94), than to recall such collaborations (n = 2, n = 2 respectively).
However, adolescents exposed to a HLB with credibility cues more frequently answered to not recall a collaboration with an organisation (n = 82), instead of naming a collaboration with the Dutch Nutrition Centre (n = 39). The organizational logo was thus not noticed as intended. Hence, only the nutritionist-manipulation was successful. Considering this result, only the nutritionist-cue will be used in the analyses.
Model Specification
The initial CFA model proved to have insufficient fit due to its highly significant chi-square. CFI almost reaches the recommended threshold, and RMSEA is good (see Table 3).
The model has good discriminant validity, such that there are no correlations higher than .85 between the different factors. To improve model fit, three indicators with low convergent validity (i.e. R2 < .50) were removed from the model (Kline, 2011). These indicators are: “How relevant is the information in the blogpost to you?” (R2 = .28), “The information provided by the blogger is useful” (R2 = .42), and “I would not recommend the advice of the blogger to a friend” (R2 = .32). Indicators from well-established scales, such as the
interpersonal attraction scale (McCroskey et al., 2006), were not discarded if their proportion of explained variance was below .50, to avoid a misrepresentation of the underlying
construct.4 Seeing that the second measurement model is a complex model with seven factors and twenty-five indicators, and the sample size is small (N = 217), chances of a significant chi-square are strong (Wolf et al., 2013). Since CFI and RMSEA values were good, further analyses were conducted with the model as specified in the second CFA.
4 A secondary analysis lacking the indicators with values below the recommended R2 threshold was run to test
for differences. In this analyses, all specified PMs were tested. No significant differences in model fit, nor in path estimates were found. PM2, for instance, proved to have good model fit without the badly fitting indicators,
χ2 (39) = 43.37, p = .291; CFI = 1; RMSEA = 0.02, 95% CI [0.00, 0.05]. All path estimates that were significant in PM2, were still significant in this new PM, and their standardized regression coefficients did not differ more than 0.02. Hence, the model including the indicators of the well-established scales was used in all analyses.
Firstly, the specified model was tested, and fitted the data well (see Table 3).
of demographic homophily and attitudinal homophily on perceived credibility and source likeability (i.e. MGM 1). The initial MGM had good fit for both females and males, and was hence used for further moderation analyses.
To maximize parsimony, an alternative model was nested under PM 2, in which all non-significant free parameters, except three, are constrained to zero (i.e. PM 3; Kline, 2011).5 After constraining these effects, model fit did not worsen significantly, χ2diff (18) = 23.45, p = .174. Hence, the more parsimonious model was retained.
Final Model Estimation
The effect of heuristic cues on source likeability, source credibility, and message evaluation.
H1, stating that adolescents will have a stronger perception of source credibility when being exposed to a HLB that contains credibility cues, compared to a HLB without credibility cues, is rejected. No significant effect of the presence of credibility cues on adolescents’ perceived source credibility was found (b* = 0.03, p = .709; see Appendix D for an overview of all standardized effects).
As predicted, increased source liking resulted in greater perceptions of source
credibility (b* = 0.37, p < .001). H2b is thus accepted. On the contrary, no effect of physical attractiveness on source likeability was found (b* = 0.44, p = .324); thereby rejecting H2a.
Similarly, no effect of demographic homophily on perceived source credibility (H3b) was found (b* = -0.30, p = .510); thereby rejecting H3b. Surprisingly, a negative effect of demographic homophily on source likeability (H3a) was found (b* = -0.30, p < .001). This
5 Three non-significant pathways were retained: demographic homophily on source likeability; health
literacy*physical attractiveness on source likeability; and personal relevance*demographic homophily on perceived source credibility, since model fit worsened significantly upon constraining all of these pathways to zero, χ2diff(24) = 71.48, p < .001. These non-significant pathways were all significant in the final model.
indicates that the more demographic similarities between source and receiver were present, the less the receiver liked the source. Since this effect is negative, H3a was rejected.
Answering RQ3, no significant indirect effect of attitudinal homophily on the evaluation of health information was found (b* = -0.37, n.s.), since not all unstandardized direct effects specified in RQ3 were non-significant (Cohen & Cohen, 1983). Significant direct effects were, however, found from both source likeability (b* = 0.30, p < .001) and perceived source credibility (b* = 0.32, p < .001) on adolescents’ evaluation of the HLB. We may thus conclude that the more positive adolescents were about the source’s character and perceived credibility, the more likely they were to positively evaluate the online dietary information in the HLB.
Sex and message evaluation.
Surprisingly, sex also had a significant effect on the evaluation of the health information (b* = -0.15, p < .05), such that on average, females evaluated the nutritional information less positively compared to males.
The moderating effects of health literacy, personal relevance and sex. Health literacy.
No interaction effect between health literacy and credibility cues on perceived credibility was found (b* = 0.01, p = .924). Hence, H4a which predicted that lower health literates would have a greater perception of source credibility when no credibility cues were present, was rejected.
H4b posited that the indirect effects of heuristic cues on message evaluation, through source liking and perceived source credibility, would be stronger for low health literates, compared to high health literates. Again, health literacy did not moderate the indirect effect of demographic homophily on message evaluation through the mediators source likeability (b* = -0.10, p = .781) or perceived source credibility (b* = -0.26, p = .466); nor did it moderate the indirect effects of attitudinal homophily on message evaluation through the mediators source likeability (b* = 0.30, p = .980), and source credibility (b* = 0.15, p = .619).
At the same time, a significant interaction effect between health literacy and physical attractiveness on source likeability was found (b* = 0.36, p < .001). That is to say, for adolescents with high health literacy a distinctly more positive effect of physical
attractiveness on source likeability is found, compared to adolescents with low health literacy. Nevertheless, this finding contradicts H4b, since the effect was predicted to be stronger for low health literates.
Personal relevance.
H4c predicted that stronger effects of the heuristic cues on source likeability and source credibility would occur for adolescents with little personal relevance. A significant and moderate interaction effect was found between personal relevance and attitudinal homophily on source likeability (b* = 0.21, p < .001), such that more attitudinal homophily resulted in a decrease in source likeability, specifically for adolescents who ascribe little personal
relevance to a healthy diet (see Figure 3). For those who ascribed great personal relevance to a healthy diet, this effect was less pronounced. The directionality of this effect, contradicts H4c.
Figure 3. Interaction effects between personal relevance and attitudinal homophily on source likeability.
1 1,5 2 2,5 3 3,5 4 4,5 5 Low attitudinal
homophily High attitudinalhomophily
Sourc e li kea bi li ty Low personal relevance High personal relevance
Moreover, a significant and moderate interaction between personal relevance and demographic homophily on perceived source credibility (b* = -0.21, p < .001) was found. Contrary to H4c, results show that adolescents with great personal relevance experienced a stronger decrease in perceived source credibility, after having been exposed to
demographically similar a source. For adolescents with little interest in healthy nutrition, the negative effect of demographic homophily on perceived source credibility was less apparent.
Contrary to what was hypothesized, personal relevance did not moderate the effect of physical attractiveness on source likeability (b* = 0.73, p = .056), nor the effect of attitudinal homophily on perceived source credibility (b* = 0.42, p = .227). Furthermore, no moderating effect of personal relevance and demographic homophily on source likeability was found (b* = 0.23, p = .490). H4c was thus rejected.
Sex.
No hypothesized moderation effects regarding sex were found, since a MGM with constrained paths for the specified moderations (i.e. MGM 2) resulted in a non-significant difference in model fit, χ2diff (3) = 3.71, p = .294. In addition, pairwise parameter
comparisons showed that none of the critical ratio values for the constrained effects exceeded 2.0. Hence, H4d is rejected.
Final model.
The final, parsimonious, model (PM 2; see Figure 4) explained moderate to strong proportions of variance for the mediating and endogenous variables (Cohen, 1988); source likeability (R2 = .21), perceived source credibility (R2 = .26), evaluation of health information (R2 = .29).
Discussion
This study aimed (1) to shed light on the key constructs that affect adolescents’ evaluation of peer-sourced dietary information in HLBs, and (2) to uncover different receiver characteristics that influence adolescents’ susceptibility to positively evaluate dietary
information on HLBs.
Firstly, two constructs were identified as the prime entities affecting adolescents’ evaluation of online dietary information: source likeability and perceived source credibility. This finding is in line with both the source-credibility model (Hovland et al., 1953) and the source-attractiveness model (McGuire, 1985). Moreover, adolescents seem to experience a halo-effect, for increased source liking resulted in a greater perceptions of source credibility (Erdogan, 1999; Ohanian, 1990).
On the contrary, neither demographic, nor attitudinal homophily could be identified as a driving force of positive message evaluation. Homophily seems less prominent as proposed Figure 4. Final model with all significant path coefficients.
Note. Numbers represent the standardized coefficient (b*). Continuous lines represent significant main
effects and significant moderation effects (i.e. for personal relevance and health literacy), whereas dashed lines represent non-significant main effects. * p < .05, ** p < .001.
in the literature (e.g. McGuire, 1985; McCroskey et al., 1975). For young adolescents, the homophily heuristic may be less effective, compared to other heuristics, such as authority. As Eysenbach (2007) explains, pre-adolescents are “more dependent on parents and authority figures than an older teenager, who deliberately seeks autonomy and questions authority” (p. 164). Still, further research has to be conducted to confirm these findings.
Rogers and Shoemaker (1971) provide a detailed explanation of homophily’s lack of persuasiveness, by describing the ‘optimal heterophily’-phenomenon. According to several studies, certain moderate dissimilarities between homophilic communicators enhance, rather than diminish, the effectiveness of communication (Alpert & Anderson, 1973; McCroskey et al., 1975). One such dissimilarity can be expertise, or the opinion leader-follower relationship (Alpert & Anderson, 1973; McCroskey et al., 1975). Even though the majority of adolescents accurately recalled the presence or absence of the nutritionist-credibility cue, this did not seem to affect their perceptions of blogger credibility. Instead, older bloggers could have been perceived as more authoritative sources with greater expertise on the topic of healthy
nutrition. Hence, age could thus be considered a demographic dissimilarity that promotes optimal heterophily, and subsequently enhances a health message’s effectiveness. This age-dissimilarity could have been so effective, it overshadowed the other homophily features. Future research should investigate whether this is actually the case, by making use of several bloggers that represent different age groups (e.g. teenager, emerging adult, adult, elderly).
Similarly surprising, references to a registered nutritionist did not influence
adolescents’ perceived credibility of the healthy-living blogger. Still, this finding is in line with previous research, which indicates that the overall effect of credibility cues on perceived source credibility is weak and non-significant (Ma & Atkin, 2017). The current study showed that few adolescents recognized the collaboration with the Dutch Nutrition Centre. This cue was most likely too inconspicuous, being at the bottom of the blogpost. Nevertheless, the
nutritionist-cue neither influenced adolescents’ credibility perceptions. As indicated by Gray and colleagues (2005a, 2005b), the majority of adolescents experiences difficulties in
accessing, understanding and evaluating the trustworthiness of online health information. Our findings show a similar trend, with adolescents having high perceptions of source credibility regardless of the presence of credibility cues. More importantly, adolescents’ health literacy levels were of no influence in the evaluation process. Hence, it is imperative that adolescents are taught how to evaluate online health information. If not, this could negatively affect their health (Gray, 2005a). Future research should test these assumptions by employing
conspicuous credibility cues—perhaps by mentioning a collaboration with a well-known health organization and showing the organization’s logo at the top of the blog page.
Physical attractiveness of the source did not seem to influence adolescents’ perceptions of source likeability. Overall, little research has been done on the effects of physical attractiveness on source likeability and communication effectiveness for adolescents (e.g. Horai, Naccari & Fatoullah, 1974; Snyder & Rothbart, 1971). Even though these studies indicate greater persuasive effects for more attractive source, a similar effect was not found in this study. Male adolescents’ normative perceptions on the attractiveness of other males, may have accounted for the non-significant effect of physical attractiveness on source likeability. During the study, several males commented that they were not attracted to males during answering of the physical attractiveness items. In an attempt to protect their masculinity, the majority of males could have negatively assessed same-sex bloggers’ attractiveness (Glick et al., 2007). Physical attractiveness therefore seems to be a measure that must be interpreted with caution when male adolescents are part of the sample.
Another surprising finding concerns the relationship between sex and message evaluation. On average, females were more critical of the information’s truthfulness and the quality of advice. Female adolescents’ more frequent use of HLBs as health information
sources (Bissonnette-Maheux et al., 2015), could have acted as a confounding factor, explaining this relation. All in all, those who more often use the internet for health information seeking, are more critical of its quality than less experienced online health information seekers (Cline & Haynes, 2001).
Moreover, adolescents’ health literacy did not influence the effects of demographic, nor attitudinal homophily on source likeability and perceived source credibility. Seeing that the source’s perceived authority or expertise could be a strong driver of these main effects, and that adolescents’ health literacy had no effect on perceptions of source credibility, this is no strange finding. However, the finding that high health literate adolescents experienced a more positive effect of physical attractiveness on source likeability, is surprising. This contradicts previous findings, which suggest that lower health literates depend more on heuristic cues, than on critical processing of the message, to evaluate health information (Chaiken & Eagly, 1993; Petty & Cacioppo, 1986). This result could be an indication of adolescents’ overestimation of their health literacy. Schmidt and colleagues (2010) found that children often claim to have very positive attitudes towards health, but nonetheless score badly on their health knowledge. Instead of self-report measures, future studies should make use of different measures that focus more on adolescents’ functional health literacy (e.g. the Newest Vital Sign; Weiss et al., 2005; or the REALM-Teen; Davis et al., 2006) to see if these results can be corroborated, or should be dismissed.
Furthermore, none of the other moderation hypothesis were supported. This lack of support was partly based on the directionality of main effects that were found (i.e. negative, instead of positive relations from demographic homophily and attitudinal homophily on the mediating variables), and partly based on the directionality of moderation effects.
Practical Implications
The findings above have important implications for health organizations that wish to utilize HLBs to promote awareness and, possibly, positive behaviour change regarding a healthy diet. For instance, when recruiting bloggers to promote a healthy diet, the expertise and likeability of the blogger, as perceived by the target audience, should factor into the design of entertainment-education campaigns. Our findings imply that demographically or attitudinally similar bloggers do not enhance message persuasiveness, since they are less liked by adolescents, and not perceived as more credible. Instead, health organizations should focus on finding bloggers who are thought to possess expertise; either through their age, or their professional appearance. Moreover, if health authorities are mentioned, this should be done noticeably to obtain the desired effect.
Strengths and Limitations
An important strength of this study, are its stimuli. Twenty-four realistic blog pages were used, thereby enhancing ecological validity, and increasing the generalizability of results (Bryman, 2015).
However, a number of limitations should be taken into account while evaluating the findings and implications of this study. Firstly, a convenience sample of secondary school pupils was used. Hence, the current sample cannot be seen as being representative of actual HLB-readers, thereby diminishing the generalizability of results. Future research should replicate this study in a representative panel of adolescent HLB-readers, to see if the current findings can be supported.
Additionally, this study utilized a fictitious blogger. Influencers could have had a greater impact on the evaluation of the online health information. For one, they most likely meet all three requirements of the source-attractiveness model: familiarity, likeability, and similarity (McGuire’s, 1985). These persons are well liked, according to their online
following, and message-receivers will be more familiar with them, compared to a fictitious blogger. Future studies should test the specified model of this study, to find out whether these relations will uphold when adolescents are exposed to a well-known influencer.
Moreover, although the perceived credibility of the different stimuli was pretested, no pre-test of adolescents’ recall of the two credibility cues was performed. Adolescents’ low recall rate of the collaboration with the Dutch Nutrition Centre, signifies a different approach is needed. More noticeable credibility cues are needed to test the influence of such cues on adolescents’ perceptions of source credibility. A way to improve the original credibility manipulation, is to include a reference to such a collaboration at the top of the HLB, in combination with a conspicuous organizational logo.
Since no manipulation check on demographic homophily was performed, no
indication of this manipulation’s effectiveness can be given. Future research should include one or more items for participants’ perceived levels of demographic homophily. For instance, McCroskey and colleagues’ (1975) measures for appearance and background homophily can be used to assess demographic homophily as experienced by the participants.
Lastly, a small sample size was used (N = 217), which does not adhere to Jackson’s (2003) N:q rule. In other words, no sample size-to-parameters ratio of 10:1 could be produced, due to the model’s complexity. The final model adhered a 1.64:1 ratio. Hence, results must be interpreted with caution. Future studies should try to replicate the current evaluative model of influencers’ HLBs by adhering a greater sample. Nevertheless, SEM’s ability to test one integrative model, rather than having to conduct several tests of the hypothesized relations, enhances the validity of the current results (Kline, 2011). Conclusion
In sum, source likeability and perceived source credibility are identified as the main drivers of message evaluation. Adolescents further experience a halo-effect, such that
increased source liking results in greater perceptions of source credibility. Homophily,
however, did not have a positive effect on source likeability, nor on source credibility. Young adolescents likely depend on authorities, rather than on peers, for health information; making authority a heuristic driver of optimal heterophily (Rogers & Shoemaker, 1971). Lastly, female adolescents were found to be more critical of dietary information in a HLB. This result can be explained by female adolescents’ more frequent usage of HLBs, compared to male adolescents. Overall, greater usage of a channel for health information, results in greater scrutiny of those channels (Cline & Haynes, 2001).
References
“Pew Research Center”, (2000). The online health care revolution: How the Web helps Americans take better care of themselves. Retrieved from
https://www.pewinternet.org/2000/11/26/the-online-health-care-revolution/
“Soa Aids Nederland”, (2016). Favoriete jongerenwebsite over seks scoort - Het succes van Sense.info. Retrieved from https://www.soaaids.nl/nl/item/news-public-prof/favoriete-jongerenwebsite-over-seks-scoort-het-succes-van-senseinfo
“Voedingscentrum”, (2018). Voedingscentrum lanceert ‘Er is meer dan vlees’ en geeft 100 shirts weg. Retrieved from
https://www.voedingscentrum.nl/nl/nieuws/voedingscentrum-lanceert-er-is-meer-dan-vlees-en-geeft-100-shirts-weg.aspx
Adams, S. A. (2010). Blog-based applications and health information: two case studies that illustrate important questions for Consumer Health Informatics (CHI) research. International journal of medical informatics, 79(6), e89-e96.
Alpert, M. I., & Anderson Jr, W. T. (1973). Optimal heterophily and communication effectiveness: Some empirical findings. Journal of Communication, 23(3), 328-343.
Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between celebrity endorser effects and advertising effectiveness: A quantitative synthesis of effect size. International Journal of Advertising, 27(2), 209-234.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411-423.
Anderson, S. R., Auquier, A., Hauck, W. W., Oakes, D., Vandaele, W., Weisberg, H. I., Bryk, A. S., & Kleinmann, J., (2009). Statistical methods for comparative studies. New York: Wiley.
Bandura, A. (1969). Social learning theory of identificatory processes. In D. A. Goslin (Ed.), Handbook of socialization theory and research (pp. 213-262). Chicago: McNally. Bandura, A. (1986). Social foundations of thought and action. Englewood Cliffs, NJ: Prentice
Hall.
Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3, 265–299.
Bandura, A. (2004). Social cognitive theory for personal and social change by enabling media. In A. Singhal, M. J. Cody, E. M. Rogers, & M. Sabido (Eds.), Entertainment-education and social change: History, research, and practice (pp. 75–96). Mahwah, NJ: Lawrence Erlbaum.
Barnes, C. (2017). Mediating good food and moments of possibility with Jamie Oliver: Problematising celebrity chefs as talking labels. Geoforum, 84, 169-178.
Bissonnette-Maheux, V., Provencher, V., Lapointe, A., Dugrenier, M., Dumas, A. A., Pluye, P., ... & Desroches, S. (2015). Exploring women’s beliefs and perceptions about healthy eating blogs: A qualitative study. Journal of medical Internet research, 17(4), e87.
Boepple, L., & Thompson, J. K. (2014). A content analysis of healthy living blogs: Evidence of content thematically consistent with dysfunctional eating attitudes and behaviors. International Journal of Eating Disorders, 47(4), 362-367.
Boon, S. D., & Lomore, C. D. (2001). Admirer-celebrity relationships among young adults: Explaining perceptions of celebrity influence on identity. Human Communication Research, 27, 432–465.
Borzekowski, D. L., & Rickert, V. I. (2001). Adolescent cybersurfing for health information: a new resource that crosses barriers. Archives of pediatrics & adolescent medicine, 155(7), 813-817.
Boyer, C., Selby, M., Scherrer, J. R., & Appel, R. D. (1998). The health on the net code of conduct for medical and health websites. Computers in biology and medicine, 28(5), 603-610.
Brown, J. J., & Reingen, P. H. (1987). Social ties and word-of-mouth referral behavior. Journal of Consumer research, 14(3), 350-362.
Brown, W. J., Basil, M. D., & Bocarnea, M. C. (2003). The influence of famous athletes on health beliefs practices: Mark McGwire, child abuse prevention, and androstenedione. Journal of Health Communication, 8(1), 41-57.
Bryman, A. (2015). Social research methods. Oxford university press.
Burger, J. M., Messian, N., Patel, S., Del Prado, A., & Anderson, C. (2004). What a coincidence! The effects of incidental similarity on compliance. Personality and Social Psychology Bulletin, 30(1), 35-43.
Carrotte, E. R., Vella, A. M., & Lim, M. S. (2015). Predictors of “liking” three types of health and fitness-related content on social media: a cross-sectional study. Journal of medical Internet research, 17(8), e205.
Chaiken, S. (1986). Physical appearance and social influence. In C.P. Herman, M.P. Zanna, E.T. Higgins (Eds.), Physical Appearance, Stigma, and Social Behavior: The Ontario Symposium. Hillsdale, NJ: Lawrence Erlbaum.
Chang, F.-C., Chiu, C.-H., Chen, P.-H., Miao, N.-F., Lee, C.-M., Chiang, J.-T., et al. (2015). Relationship between parental and adolescent ehealth literacy and online health information seeking in Taiwan. Cyberpsychology, Behavior, and Social Networking, 18(10), 618–624.
Chen, M., Bell, R. A., & Taylor, L. D. (2016). Narrator point of view and persuasion in health narratives: The role of protagonist–reader similarity, identification, and
self-referencing. Journal of Health Communication, 21(8), 908-918.
Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual. Rev. Psychol., 55, 591-621.
Cline, R. J., & Haynes, K. M. (2001). Consumer health information seeking on the Internet: the state of the art. Health education research, 16(6), 671-692.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nded.). Hillsdale, NJ: Lawrence Erlbaum Associates.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Colliander, J., & Dahlén, M. (2011). Following the fashionable friend: The power of social media. Journal of advertising research, 51(1), 313-320.
Cross, S. E., & Madson, L. (1997). Models of the self: self-construals and gender. Psychological bulletin, 122(1), 5-37.
Davis, T. C., Wolf, M. S., Arnold, C. L., Byrd, R. S., Long, S. W., Springer, T., et al. (2006). Development and validation of the Rapid estimate of adolescent literacy in medicine
(REALM-Teen): A tool to screen adolescents for below-grade reading in health care settings. Pediatrics, 118(6), e1707–e1714.
Dellande, S., Gilly, M. C., & Graham, J. L. (2004). Gaining compliance and losing weight: the role of the service provider in health care services. Journal of Marketing, 68(3), 78-91.
Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social influences upon individual judgment. The Journal of Abnormal and Social Psychology, 51(3), 629–636.
Dholakia. R.R., & Sternthal, B. (1977). Highly credible sources: Persuasive facilitators or persuasive liabilities. Journal of Consumer Research, 3, 223-232.
Diviani, N., van den Putte, B., Giani, S., & van Weert, J. C. (2015). Low health literacy and evaluation of online health information: a systematic review of the literature. Journal of medical Internet research, 17(5). Retrieved from
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4468598/
Dodd, L. (2018). An analysis of the nature and role of social influencer trustworthiness within the health and fitness sector and how it facilitates brand trust. Journal of Promotional Communications, 6(2), 184-197.
Domingues Aguiar, T., & Van Reijmersdal, E. (2018). What is influencer marketing? Stichting Wetenschappelijk Onderzoek Commerciële Communicatie, 76, 13-24. Duggan, M., & Brenner, J. (2013). The demographics of social media users, 2012 (Vol. 14).
Washington, DC: Pew Research Center's Internet & American Life Project.
Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Orlando, FL: Harcourt Brace Jovanovich College Publishers.
Eastin, M. S. (2001). Credibility assessments of online health information: The effects of source expertise and knowledge of content. Journal of Computer-Mediated Communication, 6(4), JCMC643.
Erceg-Hurn, D. M., & Mirosevich, V. M. (2008). Modern robust statistical methods: an easy way to maximize the accuracy and power of your research. American Psychologist, 63(7), 591-601.
Eysenbach, G. (2007). From intermediation to disintermediation and apomediation: new models for consumers to access and assess the credibility of health information in the age of Web2. 0. In Medinfo 2007: Proceedings of the 12th World Congress on Health (Medical) Informatics; Building Sustainable Health Systems (p. 162). IOS Press. Eysenbach, G. (2008). Medicine 2.0: social networking, collaboration, participation,
apomediation, and openness. Journal of medical Internet research, 10(3).
Fischer, E., Gainer, B., Bristor, J., (1997). The sex of the service provider: does it influence perceptions of service quality? Journal of Retail, 73(3), 361-382.
Geven, S., Jonsson, J. O., & van Tubergen, F. (2017). Gender differences in resistance to schooling: The role of dynamic peer-influence and selection processes. Journal of Youth and Adolescence, 46(12), 2421-2445.
Glick, P., Gangl, C., Gibb, S., Klumpner, S., & Weinberg, E. (2007). Defensive reactions to masculinity threat: More negative affect toward effeminate (but not masculine) gay men. Sex Roles, 57(1-2), 55-59.
Gray, N. J., Klein, J. D., Noyce, P. R., Sesselberg, T. S., & Cantrill, J. A. (2005a). The Internet: a window on adolescent health literacy. Journal of adolescent health, 37(3), 243-e1.
Gray, N. J., Klein, J. D., Noyce, P. R., Sesselberg, T. S., & Cantrill, J. A. (2005b). Health information-seeking behaviour in adolescence: the place of the internet. Social science & medicine, 60(7), 1467-1478.
Holden, G. (1992). The relationship of self-efficacy appraisals to subsequent health related outcomes: A meta-analysis. Social work in health care, 16(1), 53-93.
Holland, G., & Tiggemann, M. (2017). “Strong beats skinny every time”: Disordered eating and compulsive exercise in women who post fitspiration on Instagram. International Journal of Eating Disorders, 50(1), 76-79.
Horai, J., Naccari, N., & Fatoullah, E. (1974). The effects of expertise and physical attractiveness upon opinion agreement and liking. Sociometry, 37(4), 601-606. Hovland, C. I., Irving K. J., & Harold H. K. (1953). Communication and persuasion. New
Haven, CT: Yale University Press.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Hu, Y., & Shyam Sundar, S. (2010). Effects of online health sources on credibility and behavioral intentions. Communication Research, 37(1), 105-132.
Huh, J., Liu, L. S., Neogi, T., Inkpen, K., & Pratt, W. (2014). Health vlogs as social support for chronic illness management. ACM Transactions on Computer-Human Interaction (TOCHI), 21(4), 23.
Ibarra, H. (1992). Homophily and differential returns: Sex differences in network structure and access in an advertising firm. Administrative science quarterly, 422-447. Jackson, D. L. (2003). Revisiting sample size and number of parameter estimates: Some
Joseph, W. B. (1982). The credibility of physically attractive communicators: A review. Journal of advertising, 11(3), 15-24.
Kelly, J. A., St Lawrence, J. S., Diaz, Y. E., Stevenson, L. Y., Hauth, A. C., Brasfield, T. L., ... & Andrew, M. E. (1991). HIV risk behavior reduction following intervention with key opinion leaders of population: an experimental analysis. American journal of public health, 81(2), 168-171.
La Greca, A. M., Prinstein, M. J., & Fetter, M. D. (2001). Adolescent peer crowd affiliation: Linkages with health-risk behaviors and close friendships. Journal of Pediatric Psychology, 26(3), 131-143.
Lattimore, P. J., & Halford, J. C. (2003). Adolescence and the diet‐dieting disparity: healthy food choice or risky health behaviour?. British journal of health psychology, 8(4), 451-463.
Levine, M. P., & Murnen, S. K. (2009). “Everybody knows that mass media are/are not [pick one] a cause of eating disorders”: A critical review of evidence for a causal link between media, negative body image, and disordered eating in females. Journal of Social and Clinical Psychology, 28(1), 9-42.
Lewis, T. (2006). Seeking health information on the internet: lifestyle choice or bad attack of cyberchondria?. Media, Culture & Society, 28(4), 521-539.
Lieberman, M., Gauvin, L., Bukowski, W. M., & White, D. R. (2001). Interpersonal influence and disordered eating behaviors in adolescent girls: The role of peer modeling, social reinforcement, and body-related teasing. Eating behaviors, 2(3), 215-236.
Liu, W., Sidhu, A., Beacom, A. M., & Valente, T. W. (2017). Social network theory. The international encyclopedia of media effects, 1-12.
Lu, A. S. (2013). An experimental test of the persuasive effect of source similarity in narrative and nonnarrative health blogs. Journal of Medical Internet Research, 15(7), e142.
