Developing A Model of News Literacy in Early Adolescents: A Survey Study

ABSTRACT Early adolescents’ insufficient critical engagement with (online) news demands increased application of news literacy, but it remains unclear which factors increase news literacy application. To provide more insights, this survey study develops and tests a model of news literacy application in early adolescents (12‒15 y/o, N = 492). The more comprehensive model looks at the relationship between the application of news literacy and news consumption, knowledge of news media production, news literacy skills, value for (news) media literacy, news literacy and news consumption motivation, social norms, and demographics. Most importantly, the model shows the interconnectedness of news literacy application and news consumption. Furthermore, it shows the importance of early adolescents’ value for (news) media literacy, skills, and motivation to increase their news literacy application, and the lesser importance of news production knowledge. To stimulate the application of news literacy in early adolescents, researchers and practitioners could take a stepwise approach: (1) stimulate intrinsic news consumption motivation and, with that, news consumption, and then (2) build value for (news) media literacy, skills, and peer social norms to stimulate the application of news literacy. In precisely targeting these factors of news literacy application, interventions can empower early adolescents to become more critical news consumers.

2016). This is problematic because-especially in times of information overload and informed-lacking critical engagement with news can lead to youths being wrongly informed or even uninformed. Being wellinformed is a prerequisite to form opinions and to be an active citizen and is thus an important requirement for functioning in a democracy (McCombs et al., 2011). Therefore, it is crucial for young people to critically engage with news more often, an important application of their news literacy.
News literacy, stemming from the larger field of media literacy, empowers citizens in their engagement with news (Kajimoto & Fleming, 2019;Malik et al., 2013). Scholars have not yet reached a consensus on how news literacy should be defined, implemented, and measured. However, they generally agree that news literacy should prepare and empower people to become critical news consumers who can read, understand, analyze, deconstruct, develop, share, and sometimes even create news (e.g., Ashley et al., 2013;Aufderheide, 1993;Malik et al., 2013;Vraga et al., 2015Vraga et al., , 2020. Most recently, news literacy has been defined as "knowledge around the personal and social processes by which news is produced, distributed and consumed, and skills that allow users some control over these processes" (Tully et al., 2021, p. 5). News literacy knowledge and skills, among others, cover the role news plays in society, being able to distinguish between reliable news and disinformation, to understand (personal) bias, and to decide what information is trustworthy and actionable (Klurfeld & Schneider, 2014;Malik et al., 2013).
However, research has shown that being knowledgeable and skilled does not equal behaving in a news literate manner, such as recognizing misinformation and verifying news (Jeong et al., 2012;Vraga etal., 2020). Although previous research has looked at news literacy knowledge and skills (e.g., Ashley et al., 2013;Kleemans & Eggink, 2016), there is not much work yet on how people actually employ their news literacy and which mechanisms explain the extent to which people critically engage with news (Swart, 2021;Tully et al., 2021). Factors that might stimulate the application of news literacy have been described and tested (e.g., Ku et al., 2019;Swart, 2021;Tamboer et al., 2020;Vraga et al., 2015Vraga et al., , 2020, but one of the problems with the news literacy literature is the failure to include the multiple influences that might stimulate people to critically consume news in one model (Vraga et al., 2020). Therefore, the current study aims to explore which combination of news literacy factors leads to the application of news literacy in young people, and how.
The application of news literacy is of particular importance for youths between twelve and fifteen years old: early adolescents. Early adolescents are relatively passive in their news consumption, consume news mostly online, particularly via social media, and often feel that news will "just find them" (Craft et al., 2016;Tamboer et al., 2020). They have knowledge of news content and news effects, and are aware of the need to be critical news consumers, but this does not lead to the successful critical evaluation of news (Ku et al., 2019;Tamboer et al., 2020). At the same time, early adolescents are starting to develop lasting news consumption habits and are highly susceptible to increases in political and societal interest during this developmental phase (Marchi, 2012;Russo & Stattin, 2017;York & Scholl, 2015). Although they are an important target group, early adolescents are underrepresented in news literacy research (Marchi, 2012), presumably because of the challenges of studying minors.
Early adolescents particularly lack the application of news literacy when it comes to analyzing and evaluating, or validating, information, both when they have to distinguish between reliable news and misinformation, and when they have to evaluate the reliability of sources and evidence (Ku et al., 2019;Wineburg et al., 2016). Therefore, the current survey study looks at the extent to which various factors-news consumption, news production knowledge, value for (news) media literacy, motivation, skills, social norms, and individual factorsgo together with early adolescents' application of their news literacy. More insight into these factors is a necessary step in the development of more effective initiatives to stimulate early adolescents' critical engagement with news. Building on models of media literacy (Potter, 2004;Rozendaal, 2017), news literacy (Vraga et al., 2020), and behavioral change (Fisher & Fisher, 2002), this study empirically tests a model of news literacy application (see, Figure 1). The central question of this study is: Which factors go together with early adolescents' application of news literacy, and how?

News consumption
An important first step in determining the factors that influence early adolescents' application of news literacy is their news consumption. Although some define news consumption as part of the application of news literacy (Vraga et al., 2020), it is regarded as a separate factor in this study. News consumption is the extent to which early adolescents read, listen to, or watch news. It coincides with being well-informed on what is going on in the world, which is crucial to participate in a democracy (McCombs et al., 2011). Moreover, news consumption is an important factor in news literacy because part of the goal of news literacy is the stimulation of and value for news consumption (Ashley et al., 2013;Vraga & Tully, 2016). Therefore, theoretically, more news literacy should go together with more news consumption. This relationship is expected to be reciprocal. Consuming news can lead to more reflection on news, which can in turn increase news literacy (experience-based learning, Kolb et al., 2001).
Everyday news experiences can contribute to young people's news literacy via learning-by-doing (Swart, 2021). The first factor that is expected to go together with early adolescents' application of news literacy is thus news consumption (H1, see, Figure 1).

Knowledge and skills
In line with current definitions of news literacy, being knowledgeable is at the core of news literacy (e.g., Tully et al., 2021;Vraga etal., 2020). Early adolescents have to be knowledgeable about news media and the news production process (Amazeen & Bucy, 2019;Ashley et al., 2013;Tully et al., 2021). The need for knowledge is in line with an often-cited model in the domain of media literacy: the cognitive model of media literacy by Potter (2004). This information-processing model posits that knowledge structures, skills, and strong personal goals are fundamental to the acquisition of media literacy (Potter, 2004). Therefore, it is expected that early adolescents who are more knowledgeable in this regard about the news production process score higher on the application of news literacy than early adolescents who are less knowledgeable (H2).
In addition, early adolescents should be able to validate news (e.g., Amazeen & Bucy, 2019;Vraga etal., 2020). Or they should at least have the feeling that they are capable of doing this, i.e., have self-efficacy-the belief that someone can successfully apply their news literacy (Fisher & Fisher, 2002). They should have the agency and ability to exercise control over their relationship with news (e.g., Potter, 2004;Tully et al., 2021). It is expected that early adolescents who are more skilled in validating news engage in more application of their news literacy than early adolescents who are less skilled (H3).

Value for media literacy and motivation
In order to apply their news literacy, people need more than just knowledge and skills. Efforts to increase people's news literacy should include a combination of knowledge, skills, and various other key factors, such as attitudes, motivation, and perceived norms, to be able to increase people's actual critical engagement with news (Tully, 2021). Over time, various theoretical models have been applied to and developed for (news) media literacy in general (e.g., Potter, 2004), and more recently particularly for its application (e.g., Rozendaal, 2017;Vraga etal., 2020). News literacy theory and research points specifically at the need for cognition, self-perceived media literacy, the value for (news) media literacy (e.g., Vraga et al., 2015Vraga et al., , 2020, motivation (Rozendaal, 2017;Tamboer et al., 2020;Vraga etal., 2020), and social norms (Swart, 2021;Tamboer et al., 2020;Vraga etal., 2020) in relation to news literacy and its application. However, relevant factors in news literacy may depend on the outcome that is being considered (Vraga et al., 2015) and on the specific target group.
To further build and test a model that can explain the factors in the application of news literacy in early adolescents, we mainly draw from the Information-Motivation-Behavioral Skills model (IMB-model; Fisher & Fisher, 2002). This model was specifically designed to easily translate to interventions and has been proven successful in various health behavior change interventions (Fisher & Fisher, 2002;John et al., 2017). Therefore, this model-supplemented by research and theory on news literacy (application)-will be used as the base of our model of news literacy application in early adolescents (see, Figure 1).
Next to the aforementioned need for knowledge and the skills to engage in the target behavior, the IMB-model posits that motivation is key to understanding and promoting behaviors (Fisher & Fisher, 2002). As a starting point in this regard, adolescents should value news literacy for society (value for media literacy, or VML; Vraga et al., 2015). This construct, which is specific to news literacy, has been found to be a consistent predictor of media knowledge, current events knowledge, and of skepticism toward the press Vraga et al., 2015). Moreover, in a recent study among teenagers, higher VML has been shown to go together with higher levels of self-perceived news literacy and the application of news literacy: checking the credibility and trustworthiness of news items more often (Vandamme & Van Leuven, 2019). It is thus expected that early adolescents who value news literacy also apply their news literacy more often (H4).
Next to their value for (news) media literacy for society, people can differ in their more specific motivation to engage in the application of news literacy and in their motivation for news consumption. Previous studies have shown early adolescents' lack of news literacy motivation and speculated on its relationship to the application of news literacy (e.g., Tamboer et al., 2020). Therefore, it is expected that early adolescents who are more strongly motivated to engage in news literate behaviors engage in more news literacy application than others. Intrinsic motivation -meaning that people validate news because they believe it is important or because they like to-is expected to relate most strongly relate to the application of news literacy and to news consumption (Ku et al., 2019;Ryan & Deci, 2000). However, early adolescents have various motives for applying their news literacy and for consuming news (Tamboer et al., 2020). While some consume and validate news because they intrinsically like to do it, others engage in these behaviors because of extrinsic pressure from their parents or teachers. Therefore, both intrinsic and extrinsic motivations to apply news literacy and to consume news are taken into account. More specifically, it is expected that early adolescents who are more intrinsically (H5a) and extrinsically (H5b) motivated to validate news engage in more news literacy application. Furthermore, it is expected that early adolescents who are more intrinsically (H6a) and extrinsically (H6b) motivated to consume news engage in more news consumption. Finally, previous research has shown that being motivated to consume news can also lead to higher levels of news literacy (Maksl et al., 2015). Therefore, we expect that higher news consumption motivation (intrinsic and extrinsic) goes together with more news literacy application (H7a-b; see, Figure 1).

Social context and demographics
It is important to put early adolescents' application of news literacy in a social context, because the application of news literacy is an inherently social and connective act (Swart, 2021). As with other behaviors (Fisher & Fisher, 2002), critically engaging with news does not happen in a vacuum without influences from the social context and demographic characteristics of the individual (e.g., Vraga et al., 2020). Behavior is more likely to happen when supported by perceived social norms (John et al., 2017). This is particularly so for early adolescents, who are highly susceptible to the norms of the people around them (Duncan et al., 2007). Early adolescents tend to be susceptible to both parental and peer norms and modeling, at least regarding their news consumption (Edgerly et al., 2017;Tamboer et al., 2020). Although the home often remains the most central sphere for socialization, other socializing agents should be considered. Previous studies youths' have shown the significant (but not always consistent) role of peers on youths' news consumption (Edgerly et al., 2017;Lee et al., 2012). Less well understood is the role of teacher norms, although studies have shown the positive role of news-based school curricula (Lee et al., 2012). Therefore, we hypothesize that positive social norms from peers (H8), parents (H9), and teachers (H10), and news coverage in class (H11) go together with higher levels of news literacy application and also with news consumption (H12-15).
Finally, empirical literature shows the importance of demographic differences, although the research is mixed (Craft et al., 2016;Kleemans & Eggink, 2016;Maksl et al., 2015). In particular, older and more highly educated youths tend to have higher levels of news literacy and consumption, but the role of sex seems to be questionable (Edgerly et al., 2017;Kleemans & Eggink, 2016;Ku et al., 2019;Maksl et al., 2015). We expect positive relations between age and educational level and both news literacy application (H16-17) and consumption (H18-19) and question whether there are differences in news literacy application and consumption based on sex (RQ1-2).

The current study
This survey study thus explores how various factors-news consumption, news production knowledge, value for (news) media literacy, motivation, skills, social norms, and individual factors (see, Figure 1)-go together with early adolescents' application of their news literacy. Therewith, it contributes to existing news literacy literature in three ways. First, our test of factors in the application of news literacy considers news literacy application, but also early adolescents' news consumption, and the relationship between the two. Although the literature expects these to be related, there is, to our knowledge, not yet an empirical test of this. Besides, this study goes beyond most of the literature by focusing on the application of news literacy, instead of only news literacy, often conceptualized as news media knowledge. Second, this study goes beyond current research by offering a more comprehensive overview and test of multiple factors in early adolescents' application of news literacy (and news consumption) in one model, fulfilling the need to consider multiple factors in the application of news literacy (Vraga et al., 2020). Third, we include early adolescents' social context, which is of great importance during the developmental phase of adolescence (Duncan et al., 2007). This sheds more light on the influence of the social context on the application of news literacy, and on how the influence of the social context relates to other factors in early adolescents' news literacy application.

Method
To test the hypothesized model, an online survey was conducted in Dutch secondary schools in May and June of 2019. The study received the approval of the Ethics Committee of the Faculty of Social Sciences at Radboud University (ECSW-2019-068). The study was pre-registered at the Open Science Framework (osf.io/32mxt).

Participants and procedure
The study consisted of an online survey conducted in secondary schools. Schools were requested to participate in the study, which resulted in three participating schools across the Netherlands (all urban public schools). After approval from the principal, parents or caretakers received an information letter on the study, and adolescents were informed in class. Parents or caretakers gave passive consent, except for the parents of three children. The researchers went to the schools to administer the online survey on laptops. At the start of the survey, the adolescents were informed again and decided whether to participate. Those who did not want to participate were immediately sent to the end of the survey (n = 16). The participants worked independently on the survey. After filling out the survey, they were thanked and compensated with a short fake news lesson.
A total of 528 early adolescents participated, in line with our a priori power analysis for SEM (d = .20, power = .80, 10 latent and 6 observed variables, alpha = .05 leading to 475 participants needed; Soper, 2019). The pre-registration states that participants would be excluded based on outliers in the total reaction time per educational level. However, this turned out to be inadequate because it did not exclude those participants who clearly did not take the survey seriously. In total 36 participants were excluded. A total of 26 participants gave clearly impossible answers on their educational level (e.g., attending the highest level of education while their school only offers lower levels) or age, such as "87." In addition, 10 participants were excluded because of total reaction times below 5 minutes and/or careless answers, meaning that they gave identical answers in a sequence more than the mean identical sequence plus three times the standard deviation (Meade & Craig, 2012).
The final sample consisted of 492 early adolescents. Their age ranged between 12 and 17 years (484 between 12 and 15 1 ) with a mean age of 13.57 years (SD = 0.95). Boys and girls were represented almost equally, with 45.7% girls. Of the total sample, 77.4% had a lower level of secondary education, 22.6% had a medium to high level of education 1 . Most participants were in their first year of secondary school (60.0%), 17.5% were in their second year, and 22.6% were in their third year.

Measures
Questions based on existing scales were translated using back-translation. All the final scales had sufficient scores on the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy, a significant Bartlett's test of sphericity, and a determinant score indicating the absence of multicollinearity.

Dependent variables
News Literacy Application (NLA): Analysis and Evaluation. For news literacy application (NLA), participants indicated in eight questions to what extent they performed actions that indicate a critical engagement with news, or more specifically, the validation of news. These questions were constructed based on the definition of news literacy (Craft et al., 2016;Livingstone, 2004) and answered on a six-point Likert scale (1 = never to 6 = always). For an overview of the items, see the Appendix. A principal component analysis (PCA) of the eight items yielded two factors with eigenvalues larger than one and items loading >.5 (Cronbach's α = .73; .82).
Conceptually, these factors cover two different parts of NLA: analyzing and evaluating news. Because the factors emerged as statistically and 1 In the Dutch school system learning processes are adapted to the needs of individuals by dividing pupils into different streams, based on their achievement levels at the end of primary education (around the age of 12). Ordered from lowest to highest achievement, pupils go to pre-vocational secondary education (called VMBO), higher general secondary education (HAVO), or pre-university level (VWO). For this study, these are considered low, medium, and high levels of education. This survey was administered in classes in the first three grades of secondary education, in which most adolescents are between 12 and 15 years old, but some are slightly older.
theoretically different, they were both included in the analyses and treated as two separate parts of news literacy. News analysis means taking a step back when consuming news and considering criteria such as trustworthiness, framing and bias, and (commercial) media motives. News evaluation is more actively checking sources, seeking further information, and crosschecking (visual) information. We calculated two mean scores, for analysis and for evaluation (all descriptive statistics, see, Table 1). News Consumption (NC). For news consumption (NC), participants indicated how often they consume news per medium (print, TV, radio, news via the mobile application/website of news organizations, social media news, other people, and other (to be specified by the participant)) ranging from 1 = never to 6 = always (for all items, see OSF project). We calculated a mean score combining all items in one variable measuring NC. Some participants used the open-answer box to mention other ways they get news. However, answers were mostly specifications of the other options they had already filled out, such as the internet, TikTok, Instagram, at school, when parents are watching news, WhatsApp group chats, and being at the news event itself.

Independent variables
Knowledge NL knowledge was measured by the fifteen statements from the NL scale by Ashley et al. (2013), containing knowledge on authors and audiences (AA), messages and meanings (MM), and representation and reality (RR; for all items, see OSF project). In line with Kleemans and Eggink (2016), statements were judged on whether they are generally right or wrong, so 1 = true and 2 = false. One point was awarded for each right answer, leading to a sum score ranging from zero to fifteen.

Skills
For NL skills, we focused on self-efficacy of NL skills. Self-efficacy is an often-used proxy for behavioral skills in research that focuses on developing models for interventions (e.g., John et al., 2017). Self-efficacy was based on guidelines by Bandura (2006) and on the definition of NL by Craft et al. (2016). The seven statements included "I am able to find news that is trustworthy" and "I am able to judge whether a news message is correct." Answers ranged from 1 = not at all true to 6 = completely true. For an overview of the items, see the Appendix. A PCA of the seven items showed two factors with eigenvalues larger than one. Upon inspection of the factors, the item "I am able to find news that does not correspond with my own ideas" has the lowest factor loading (α = .43) on the main factor, most likely because of misinterpretation of the negative wording. After deleting this item, the PCA yielded one factor with all items loading > .5 on this factor (α = .70). Therefore, we calculated a mean score combining these items in one variable measuring NL skills.

Value for media literacy (VML)
VML was measured by a scale developed by Vraga et al. (2015). Participants answered eight statements on the importance of NL, on a scale from 1 = not at all true to 6 = completely true (for all items, see OSF project). A PCA of the eight items showed two factors with an eigenvalue larger than one. Upon inspection of the factors, the items that loaded onto the second factor seem to have some problems in comprehension. The two items that were reversed were most likely misinterpreted, as happened before (Vandamme & Van Leuven, 2019). Furthermore, the translation of the item stating that "It is the job of citizens to overcome their own biases in consuming news" was most likely confusing to participants. After these three items were removed, the PCA yielded one factor with factor loadings > .6 (α = .72). We calculated a mean score combining the items in one variable measuring VML.

News Literacy (NL) motivation: intrinsic and extrinsic NL motivation was measured with four items developed for this study, based on Maksl et al. (2015) and Vandamme and Van Leuven (2019).
Participants judged the statements such as "If I critically engage with news, it is for my own benefit" and "If I critically engage with news, it is because I like it" ranging from 1 = not at all true to 6 = completely true (for all items, see OSF project). A PCA did lead to two meaningful factors representing clearly intrinsic and extrinsic motivation, but with insufficient Cronbach's Alpha's (α = .59; .36). We, therefore, decided to include only the most intrinsic item, "If I critically engage with news, it is because I like it," for intrinsic motivation, and the most extrinsic item, "If I critically engage with news, it is because I'm supposed to," for extrinsic motivation. Because these variables now consist of only one item, they were changed from latent ( Figure 1) to observed variables ( Figure 2).

News consumption (NC) motivation: intrinsic and extrinsic
NC motivation was measured in line with NL motivation, replacing critically engaging with news by consuming news (for all items, see OSF project). A PCA of the four items together resulted in two components with an eigenvalue higher than one. However, these did not clearly represent intrinsic and extrinsic motivation, but showed overlap because of the statement "I don't see what consuming news does for me." Therefore, as with NL motivation, we decided to include only the most intrinsic item, "I consume news because I like it," for intrinsic motivation, and the most extrinsic item, "I consume news because I'm supposed to," for extrinsic motivation. Because these variables now consist of only one item, they were changed from latent ( Figure 1) to observed variables ( Figure 2).

Peer/parental/teacher social norms
The measurement of social norms was roughly based on earlier work by Edgerly et al. (2017). In total, participants answered three sets of four statements on social norms such as "My peers believe it is important to know what is going on in the world" and "My parents believe it is important to check whether news is correct or not" (for all items, see OSF project). Statements were answered on a scale from 1 = not at all true to 6 = completely true. Three separate PCAs were conducted. The PCAs of the four items for peer, parental, and teacher norms yielded a factor with all items loading > .6 (α = .82; α = .79; α = .86). Therefore, we calculated mean scores combining all the items in variables measuring peer, parental, and teacher norms.

News in school
Participants were asked to indicate how often they discuss news in class, on a six-point scale ranging from 1 = never to 6 = always (for all items, see OSF project).

Demographics
Participants indicated their educational level. Educational level was recoded as 0 = low, and 1 = medium to high. Age was reported and sex was indicated by choosing 0 = male or 1 = female.

Strategy of analyses
Descriptive statistics were calculated to examine standard deviations and means for all variables in the model (see , Table 1). Subsequently, correlations between all model items were performed to assess bivariate associations (for the full correlation table, see OSF project). The primary analysis consisted of a structural equation model (SEM) testing various predictors of NLA and NC (see, Figure 1) in R Studio (RStudio Team, 2019). NLA, VML, skills, and social norms from peers, parents, and teachers were included as latent constructs, the others (knowledge, NL and NC motivation, NC, news in school, age, educational level, and sex) as manifest variables. Parameters were estimated with a (full-information) maximum-likelihood estimator with robust standard errors (MLR), accounting for possible deviations from multivariate normality. Model fit was assessed by χ 2 , CFI (comparative fit index, with a cutoff value of .95), and RMSEA (root mean square error of approximation, with a cutoff value of .06; Hu & Bentler, 1999).

Model fit
Bivariate correlations showed a moderate correlation between the two NLA components (analysis and evaluation, r = .56, p < .001). Therefore, the covariance between analysis and evaluation was added to the model. Figure 2 presents the results of the path model. The hypothesized model did not have a good fit, χ 2 (5) = 36.91, p < .001, CFI = .958, RMSEA = .12. To better represent the data structure and to make the model more comprehensiveespecially the connection of the base model to NC-three paths (with MI > 5; cf., Gunzler & Morris, 2015) were added stepwise to the model (see dashed lines in Figure 2). These paths are the relationship between intrinsic literacy motivation and NC, the relationship between knowledge and NC, and the relationship between skills to evaluate news content and NC. Theoretically, these paths make sense because of the important connection between NC and NLA. With these additions, the model had a good fit with the data: χ 2 (2) = 3.92, p = .141, CFI = .997, RMSEA = .05.

Model findings
Because NLA turned out to consist of two components (analysis and evaluation), the hypotheses on NLA will be discussed for both parts. For both parts, our analyses revealed the expected positive covariance with NC (H1 for news analysis: β = .27, p < .001, H1 for news evaluation: β = .30, p < .001). Furthermore, analysis and evaluation were positively related (β = .38, p < .001). Differences between relations with analysis and evaluation will be briefly discussed, but as we did not a priori hypothesize differences between the two, our analyses on this should be seen as exploratory and thus interpreted with caution.

Knowledge
Contrary to our expectations in H2, the analyses revealed that knowledge (β = −.06, p = .142) was not related to news analysis, and negatively related to news evaluation (β = −.14, p = .001), rejecting H2. Participants who were more knowledgeable about news media actually evaluated news less often.

Skills
Partly supporting H3, having skills was positively related to news analysis (β = .24, p < .001), but not to news evaluation (β = .07, p = .097). Early adolescents who felt that they were able to critically engage with news, engaged in more news analysis than others who felt less able.

VML
Supporting H4, VML was positively related to news analysis (β = .14, p = .008) and to news evaluation (β = .16, p < .001). Early adolescents with higher VML engaged in more news analysis and news evaluation.

NL motivation
The analyses showed that intrinsic NL motivation did not go together with news analysis (β = .08, p = .073), but did relate to news evaluation (β = .10, p = .023). H5a can thus be rejected for news analysis, but supported the relation between intrinsic NL motivation and news evaluation. Early adolescents with higher intrinsic motivation for NL engaged in more news evaluation. Extrinsic NL motivation was not related to either news analysis (β = −.04, p = .304) or news evaluation (β = .02, p = .589), rejecting H5b.

NC motivation
Regarding NC, the analyses supported H6a and b, that higher intrinsic (β = .32, p < .001) and extrinsic NC motivation (β = .16, p < .001) goes together with more news consumption. Participants who felt more that they were supposed to consume news or who liked consuming news more than others, also consumed more news. Regarding NLA, the analyses showed that intrinsic NC motivation did go together with news analysis (β = .12, p = .018), but did not relate to news evaluation (β = .04, p = .416). H7a is thus partly supported. Participants who consumed more news because they like it, indicated analyzing news more often than others. Extrinsic NC motivation was found to be related to both news analysis (β = .10, p = .017) and evaluation (β = .09, p = .019), fully supporting H7b. Participants who consumed news because they were supposed to also analyzed and evaluated news more often.

Social norms: parents
The analyses showed that parental norms did not go together with news analysis (β = .07, p = .179), but did relate to news evaluation (β = .12, p = .017). H8 can thus be rejected for news analysis, but supported the relation between parental norms and news evaluation. Furthermore, parental norms coincided with NC (β = .12, p = .007), supporting H12. Participants who believed their parents were more positive about consuming and critically engaging with news also evaluated news more often and consumed more news.

Social norms: peers
In line with H9 and H13, peer norms had a significant positive relation with news analysis (β = .20, p < .001), news evaluation (β = .27, p < .001), and NC (H15, β = .10, p = .025). Participants who believed their peers were more positive about consuming and critically engaging with news also analyzed, evaluated, and consumed news more often.

Social norms: teachers
Based on the analysis, both H10 and H14, on the influence of teacher norms on both NLA and NC, can be rejected. Teacher norms were not related to news evaluation (β = −.08, p = .071) and NC (β = −.04, p = .391). Moreover, teacher norms even had a negative relation with news analysis (β = −.14, p = .005). Participants who felt that their teacher was very positive about NL actually engaged in less news analysis.

News in school
The extent to which news is discussed in school had the expected positive relations with news analysis (H11, β = .11, p = .013) and news evaluation (H11, β = .10, p = .039), and with NC (H15, β = .12, p = .007). Participants who indicated that news is discussed in class more often also analyzed, evaluated, and consumed news more often.

Added paths
To develop and test a model that is theoretically sound and statistically fitting, three paths were added in the analysis process (stepwise). These showed a significant negative relation between knowledge and NC (β = −.09, p = .015) and positive influences of intrinsic NL motivation (β = .15, p < .001) and skills (β = .15, p < .001) on NC. Adolescents who were more knowledgeable, with higher scores on intrinsic NL motivation and NL skills, thus consumed more news.

Discussion
Building on existing models of (news) media literacy and behavioral change, this article offers a test of a more comprehensive model of the application of news literacy in early adolescents. Within this model, we find a strong interconnectedness of news literacy application and news consumption and argue for a focus on early adolescents' value for (news) media literacy, skills, and motivation to increase their application of news literacy. This application of news literacy is manifold, and initiatives should be carefully adapted to suit the targeted outcome. Although knowledge is often heralded as an important factor in news literacy, this study shows that the social context-including peers, parents, and the classroom-is of much more importance for early adolescents' news literacy application. However, this social context should be carefully considered because teacher norms could also lead to less application of news literacy.
First, this study emphasizes the relation between news consumption and the application of news literacy. Our participants sometimes consumed news, while they engaged in less (rarely to sometimes) news analysis and evaluation. This is in line with previous qualitative research on both youths' news consumption-which is often incidental-and application of news literacy, in which youths are quite passive (Swart, 2021;Tamboer et al., 2020). Youths are thus not only insufficiently able to critically engage with news (e.g., Wineburg et al., 2016), but the current study shows that they rarely engage in news consumption and in the application of their news literacy.
News consumption and the application of news literacy are clearly related news engagements that are largely influenced by motivational constructs. This study, however, does show that the factors that influence consumption and news literacy application depend on the targeted outcome, as previously argued by Vraga et al. (2015). To stimulate the application of news literacy in early adolescents, researchers and practitioners could take a stepwise approach: (1) stimulate intrinsic news consumption motivation and, with that, news consumption, and then (2) build value for (news) media literacy, skills, and peer social norms to stimulate the application of news literacy.
The importance of the targeted outcome also became visible in the emergence of two separate forms of news literacy application: news analysis and news evaluation. Theoretically, the difference seems to imply a scale of how active someone is in their application of news literacy. Someone can apply their news literacy by taking a step back when consuming news and analyzing important news criteria, such as trustworthiness and framing. A more active approach, then, is to go one step further and to, for example, evaluate sources and seek further information. Although this distinction was not expected in advance, it is in line with the argument by Vraga et al. (2015) that the factors that influence news literate behaviors depend on the exact behavior that is targeted. The explorative finding that these two news literacy application concepts show differential patterns of influence suggests that different levels of news literacy application activity warrant different approaches. However, because there were no hypotheses regarding differences between these two forms, future research should confirm differences between them and their relations to other factors.
Second, this study shows that when considering multiple factors, not early adolescents' knowledge, but how they value news literacy was found to be a consistent factor in both news analysis and evaluation. Skills were important for both news analysis and news consumption, and intrinsic news literacy motivation was found to be important for news evaluation and news consumption. This supports qualitative and quantitative evidence of the importance of motivational constructs in early adolescents' critical engagement with news (e.g., Swart, 2021;Tamboer et al., 2020;Vandamme & Van Leuven, 2019). Moreover, it indicates that initiatives aiming to stimulate early adolescents' application of news literacy should at least focus on their value of (news) media literacy. To stimulate this, initiatives could, for example, include a clear focus on citizens' role in a society to critically engage with news content (such as shown by Tully & Vraga, 2017).
The finding that knowledge was not, or was even negatively, related to news evaluation and news consumption can be explained in various ways. More knowledgeable adolescents could underestimate their own news consumption and evaluation (a Dunning-Kruger effect, Dunning, 2011), or more knowledge could have led to news media cynicism, and as a result to less news consumption and evaluation (as has been found before in Tsfati, 2010;Vandamme & Van Leuven, 2019). It could also be that knowledge served as a base for the value for (news) media literacy, with which it correlates, and thus fueled the relation between the value for (news) media literacy and the application of news literacy.
Finally, this finding could be due to the specific application of news literacy, validating news, or to the measure that was employed. Previous use of the scale by Ashley et al. (2013) has shown mixed results: although some found that higher scores on this scale went together with the critical processing of news information and media skepticism (Ku et al., 2019;Vraga et al., 2015), others found that this scale did not relate to fake news recognition (Jones-Jang et al., 2021). Future research could also administer other knowledge scales (e.g., Amazeen & Bucy, 2019;Maksl et al., 2015) to assess a possible relation between knowledge and specific applications of news literacy.
Third, this study stresses the importance of social norms in the application of news literacy. The influences of both peers and parents support earlier findings on the susceptibility of early adolescents to people close to them (Duncan et al., 2007;Edgerly et al., 2017;Tamboer et al., 2020). This study adds that, when considered together with more substantive influences on news literacy application, peer social norms turn out to be most important, even more so than the value for (news) media literacy. A notable finding is that the extent to which news is discussed in class positively concurred with news consumption and both analysis and evaluation (in line with Edgerly et al., 2017;Lee et al., 2012), but that teacher norms did not go together with news consumption and evaluation, and negatively co-occurred with news analysis. It is thus effective to discuss news in class, but teachers should be cautious in their expressions of the need to critically engage with news. This could be a form of psychological reactance, which is not uncommon in efforts to intervene in adolescent behavior (e.g., Burgoon et al., 2002). However, future research should examine this relation further.
This study has a number of strengths, such as the relatively large sample size and the test of all influences simultaneously in one SEM model. However, there are also some limitations, related to our measures, sample, and model. First, all assessments were based on self-reports, which leaves room for over-and underestimation of frequencies or differences in the interpretation of answer categories by participants. Future research could use or develop more objective measures, such as diary studies or news evaluation tests, and further examine both the quantity and quality of news consumption. Besides, some measures use the term "news" without defining it. Although we expect participants to have conceptualized news mainly in line with traditional definitions (as found in Swart, 2021;Tamboer et al., 2020), this could have led to different interpretations by participants.
Second, although our sample is relatively large and contained a variety of early adolescents, there is only limited insight into participants' background and diversity. This study took place in the Netherlands, and most participants were in the first year of secondary school of the lower educational level. For future research, more insight into and variation in educational backgrounds, socio-economic breakdown, and parental level of education is needed. Additionally, the model should be tested in different countries and contexts.
Third, our test of the new model was relatively clean, correlational, and partly exploratory because of the added paths. We should be aware that there are other underlying factors and interactions that, for example, influence people's motivation to critically engage with news. Characteristics such as a need for cognition and media locus of control have been shown to go together with motivation for news consumption (e.g., Maksl et al., 2015), but are also likely to influence people's NL motivation and skills. Future research should further disentangle these underlying relations and their role in the application of news literacy, for example, by building on the model that we employed in this study or by empirically testing the recently proposed model by Vraga et al. (2020). Furthermore, the application of news literacy can be due to situational and personal factors, such as message and outlet characteristics and ideology (Edgerly et al., 2020;Swart, 2021), which could be considered in future research.
Ultimately this study's value lies in its test of insights into the factors in early adolescents' application of news literacy. It has combined theoretical insights from models on media literacy (Potter, 2004;Rozendaal, 2017), news literacy (Vraga et al., 2020), and behavioral change (Fisher & Fisher, 2002), and empirically tested these in a more comprehensive model. This study adds to our knowledge that the stimulation of early adolescents' news literacy application should focus on their peer norms, news consumption, value for (news) media literacy, skills, and motivation. That means that initiatives should not be strictly instructional, but should be interactive, social, and persuade early adolescents that news literacy is important for society and for themselves. Schools could play a role in that, but the line between stimulation and reactance is precarious. Critically engaging with news is not something that comes naturally to anybody, but is highly important. Future research and initiatives could use these to make that change: to empower early adolescents to become more critical news consumers.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Data Availability Statement
The anonymized data that support the findings are available in the Open Science Framework project (https://osf.io/6u2pm/).