Running head: DON'T KEEP IT (TOO) SIMPLE Don't ...

3 downloads 1035 Views 617KB Size Report
Journal of Language and Social Psychology. 1. University of ..... Furthermore, participants' attitudes toward general effects of violent media were measured with ...
Running head: DON’T KEEP IT (TOO) SIMPLE

Don’t Keep It (Too) Simple – How Textual Representations of Scientific Uncertainty Affect Laypersons’ Attitudes Stephan Winter1, Nicole C. Krämer1, Leonie Rösner1, and German Neubaum1 Journal of Language and Social Psychology

1

University of Duisburg-Essen, Germany

Corresponding author: Stephan Winter, University of Duisburg-Essen, Forsthausweg 2, 47057 Duisburg, Germany. Email: [email protected]

Abstract This research investigated the question of how laypersons are influenced by textual representations of scientific uncertainty. In an online experiment (N = 78), a blog article about effects of computer games on children was presented in four different versions: Each version contained three arguments on negative effects that were either phrased neutrally, contained assertive statements, or included hedges. The fourth version contained an additional argument on positive effects of computer games (two-sided). In comparison to the basic one-sided version, the two-sided text led to a more moderate attitude toward the topic. According to moderation analyses, this difference was mainly based on readers with more advanced epistemological beliefs and with a higher need for cognition, who were more strongly affected by a two-sided presentation of evidence. The assertive version was less effective than the basic version, suggesting that recipients were skeptical when statements were presented as overly certain. Keywords: attitudes, message sidedness, hedging, science communication, elaboration likelihood model, blogs

Don’t Keep It (Too) Simple – How Textual Representations of Scientific Uncertainty Affect Laypersons’ Attitudes Scientists’ answers to questions of society are often tentative. While some degree of uncertainty can be seen as a characteristic of every form of empirical results (Popper, 1968), this is particularly true in the case of current scientific debates, e.g., on the effects of new technologies (“What are the risks and benefits of stem cell research?”, “Should genetically modified food be banned?” or “Are computer games dangerous for teenagers?”). In these contentious fields, studies frequently present patterns of conflicting evidence which do not lead to straightforward conclusions on the “correct” view (Friedman, Dunwoody, & Rogers, 1999). When presenting scientific results to the general public, communicators can decide on the extent to which they want to include representations of uncertainty in their messages. On the one hand, it might be argued that it is appropriate to acknowledge limitations of current studies and present conflicting results: Such a form of two-sided argumentation (O’Keefe, 1999) or hedging (Jensen, 2008) can represent an attempt to capture the tentativeness that is connected to empirical research and might enhance laypersons’ knowledge about the nature of science (unless there is overwhelming support for the dominant position and almost no scientific disagreement anymore; Dixon & Clarke, 2013). On the other hand, research on powerful language has shown that statements which are presented as certain and in an assertive style are more influential and even increase the speaker’s credibility (Blankenship & Holtgraves, 2005; Burrell & Koper, 1998). Against this background, this paper aims to investigate the question of how laypersons deal with different depictions of scientific uncertainty. In particular, we ask which type of texts (varying from one-sided and assertive statements to a more balanced and two-sided version including hedges and limitations) has the strongest effects on readers’ attitudes.

Furthermore, we examine whether the persuasive effects of the different text types depend on characteristics of the reader: Based on the elaboration likelihood model of persuasion (Petty & Cacioppo, 1986), need for cognition (Cacioppo & Petty, 1982), the general motivation to engage in elaborate cognitive activities, and epistemological beliefs (Hofer, 2000), i.e., personal views on the structure and certainty of knowledge, are taken into account as crucial variables that may affect the process of elaborating on textual representations of uncertainty. In order to investigate these issues, we conducted an experiment in which an online science article dealing with the effects of violent computer games on children and teenagers was shown in different versions. Science blogs were chosen as the exemplary domain for our experiment since these forms of contemporary websites are frequently used channels in which laypersons encounter articles on scientific topics (Brossard & Scheufele, 2013). What Affects Attitudes toward Scientific Topics? In order to elucidate the question of which factors influence laypersons’ attitudes when reading texts on scientific topics on the Internet, research on persuasion can be used as a basic framework. Dual-process models such as the elaboration likelihood model (Petty & Cacioppo, 1986; Petty, Rucker, Bizer, & Cacioppo, 2004) are able to describe the interplay of message, source, and audience characteristics in order to predict persuasive outcomes. The model posits that recipients carefully examine the given information when they are motivated and able to do so. This kind of elaborative processing is called the central route: Here, the strength of persuasion mainly depends on the quality of the given arguments. Otherwise, when receivers lack motivation and ability (for instance, if the topic is not relevant for them or they do not have sufficient levels of prior knowledge or reading skills), they process the information superficially and are mainly influenced by peripheral aspects such as information on the source (e.g., likeability or attributed expertise). With regard to general individual differences, the ELM describes the need for cognition (NC; Cacioppo & Petty, 1982) as a personality

characteristic which influences elaboration likelihood in that people who enjoy effortful thinking (high NC) more often process information on the central route. Given that people are particularly likely to select science articles for further reading if they perceive the topic as relevant (or if they are generally interested in news and science topics), it can be assumed that the motivation of readers in the context of science blogs is relatively high or at least moderate (Winter & Krämer, 2012). Following the ELM, this should lead to central processing and a higher focus on the characteristics of the message in that higher-quality arguments should be more persuasive than lower-quality arguments, especially for readers with a higher need for cognition. What remains open, however, is the question of how different textual representations of uncertainty (as they are relevant when describing scientific evidence) relate to persuasive strength: Even when the content of the argument and its quality are basically the same, there may still be perceptual and persuasive differences due to differences in style and wording (Blankenship & Craig, 2011; Gibbons, Busch, & Bradac, 1991), and further effects can be expected when additional pieces of information are included. In the following, we will discuss different text types representing scientific uncertainty and its potential effects on readers’ attitudes. Then, with regard to individual differences, we will consider reader characteristics which may influence the effects of the different text types according to the mechanisms of the ELM. Message Type: Representations of Scientific Uncertainty Within online science articles, scientific uncertainty can be expressed through linguistic elements within the description of certain pieces of evidence (e.g., in a summary of results of an empirical study). This may arise through words like “might”, “could”, “potentially” or “possibly”, which imply the tentativeness of the presented conclusions (for instance, “The study may indicate that playing violent video games could lead to a higher level of aggression”). The use of these linguistic elements, which act as qualifiers of the given

statements, is called hedging (Blankenship & Holtgraves, 2005; Durik, Britt, Reynolds, & Storey, 2008; Markkanen & Schröder, 1997). Besides lexical hedges, the preliminary nature of the conclusions can also be indicated by additional information on the limitations of the presented study (e.g., with references to the limited generalizability of the findings), which would be a more explicit way of expressing uncertainty. Both types of hedges have been referred to as forms of powerless language (Burrell & Koper, 1998; Hosman, 2002). Based on Lakoff’s work (1975) on gender differences in speech, the concept of powerless language summarizes linguistic elements that relate to the expression of caution. Besides hedges, tag questions, hesitations, and polite forms have been mentioned as markers of powerlessness (Hosman & Siltanen, 2011). The absence of these features or the use of intensifiers (Hamilton, Hunter, & Burgoon, 1990) which express certainty (e.g., “without a doubt”) is called powerful language. In a meta-analysis of 16 studies, Burrell and Koper (1998) found that powerful language was more persuasive than powerless language and that speakers were even perceived as more credible when they used powerful expressions. Similarly, Hosman (2002) as well as Blankenship and Holtgraves (2005) described a negative relationship between powerlessness and persuasive outcomes (usually based on information processing on the central route when message relevance is high). In a more fine-grained analysis of different types of hedges, Durik et al. (2008) found detrimental effects of hedges when they refer to statements about data and of colloquial expressions such as “sort of”. In contrast, more professional expressions (e.g., “likely” or “seem to”) and hedges on interpretative statements (e.g., “this may reduce the threat…”) were not evaluated negatively by readers. One further factor may be the domain in which powerful vs. powerless language occurs: Most studies that were included in the meta-analysis by Burrell and Koper (1998) were framed in a courtroom setting (e.g., speeches of lawyers). Therefore, in the setting of

science, in which uncertainty is a basic concept (Popper, 1968), effects of hedges may be different and more positive. In this line, Jensen (2008) showed in an online experiment with science news articles that both scientists and journalists were evaluated as more credible when news coverage was hedged (and when the scientists who were responsible for the research used the hedged expression). The author explains this finding, which is contrary to prior research on powerless language, based on the characteristics of the domain of science: “Unlike many professions, science favors a more self-critical style of communication that may alter the norms of communication” (p. 362). Accordingly, hedging would be a valuable strategy of gaining credibility when reporting scientific results. However, effects on the formation of attitudes toward the topic were not investigated in the study. Furthermore, it has to be noted that the study was conducted with college students whose beliefs about the nature of science might already be different from those of average citizens, meaning that this pattern should also be tested with a more diverse sample. On a higher level, with regard to the wider composition of a message, scientific uncertainty is not only visible in the way in which scientific results are described: It is also crucial which pieces of evidence relevant to a controversial issue are chosen for presentation. In this respect, authors of science articles can choose whether only arguments and pieces of information which represent the dominant position (or the position which is evaluated as dominant in the author’s view) are mentioned, which would be a one-sided way of argumentation, or whether arguments for the opposing position are also included (two-sided argumentation; Allen, 1991; O’Keefe, 1999). Applying this to the example of science blogs, a text on violent video games could only describe negative effects (e.g., links between violent video games and aggressive behavior) or it could also refer to positive effects (e.g., links between online gaming and team skills). Even if there is no direct refutation, the juxtaposition of conflicting evidence is likely to arouse uncertainty with regard to the overall evaluation of

a topic, which may be unpleasant for readers when their prior attitudes are called into question (Hale, Mongeau, & Thomas, 1991). For the domain of advertisements, persuasion studies showed that speakers of twosided messages were perceived as more honest and more credible (Crowley & Hoyer, 1994); however, the presence of counterarguments is likely to undermine persuasive effects unless they are refuted by the speaker (O’Keefe, 1999). For the setting of online science information (where two-sided messages may be less focused on persuasive goals but can be seen rather as attempts to provide an appropriate picture of scientific domains with conflicting evidence), research showed that readers who wish to gather information more often select two-sided articles (which, according to their headlines and summaries, offer a more balanced content) than one-sided information (Winter & Krämer, 2012). However, it is not yet clear whether readers also integrate the given conflicting information into their attitudes and whether this leads to the consideration of scientific uncertainty and fragility. In order to investigate the effects of these textual representations of uncertainty, the present research aims to compare the classic way of presenting a scientific position in a relatively neutral style (one-sided basic version) with versions in which either powerful expressions (assertive style) or hedges (in terms of powerless language) are added. Furthermore, we investigate the effects of adding a counterargument, resulting in a two-sided presentation of evidence. Since reading a science blog article on a relevant topic implies moderate or high levels of elaboration, it can be assumed that these different depictions of uncertainty exert impacts on readers’ attitudes. Based on the theoretical background regarding powerful vs. powerless language and message sidedness, we expect to find the following effects in comparison to the one-sided basic version:

H1a: One-sided science articles with assertive statements will lead to a more extreme attitude toward the topic (following the position of the text). H1b: One-sided science articles with hedges will lead to a more moderate attitude toward the topic. H1c: Two-sided science articles will lead to a more moderate attitude toward the topic. With regard to evaluations of credibility and text quality, previous research showed advantages of powerful expressions (Burrell & Koper, 1998) as well as of hedges (Jensen, 2008) and two-sided messages (Crowley & Hoyer, 1994). In the light of these opposing tendencies, we posit the following research question: RQ1: How do textual representations of uncertainty affect readers’ evaluation of science articles? Effects of Reader Characteristics As a stable individual characteristic, the ELM includes the need for cognition (NC), which refers to a person’s tendency to engage in and enjoy effortful thinking (Cacioppo & Petty, 1982; Cohen, Scotland, & Wolfe, 1955). The model suggests that this personality variable increases the likelihood of elaboration: In this line, research has shown that people with higher NC are typically motivated to reflect on arguments thoroughly and are therefore mainly influenced by the quality of arguments (Cacioppo, Petty, Feinstein, & Jarvis, 1996). Transferring these assumptions to the effects of textual representations of uncertainty, rather than quality of arguments, it can be expected that people high in NC are also more willing to deal with uncertain and conflicting evidence, since this also includes elaborate and resource-demanding processing. In an experiment by Kardash and Scholes (1996) in which participants read a two-sided essay about a controversial issue, people with higher NC more

often wrote conclusions in which the situation of mixed evidence was mentioned. In accordance with this, we assume that high-NC individuals are more likely to consider conflicts and representations of uncertainty (which are particularly visible in two-sided messages), while low-NC individuals may try to ignore conflicts in order to avoid complex thinking. H2: Two-sided messages (in comparison to one-sided messages) will have a stronger influence on readers’ attitudes toward the topic when readers have a higher need for cognition. As a further personality characteristic which has not yet been focused upon in the context of attitude formation but which appears to be particularly relevant in the domain of science, we consider readers’ epistemological beliefs (Bromme, Kienhues, & Porsch, 2010; Hofer, 2000; Schommer, 1990), i.e., personal views on the nature of science and knowledge. This concept is based on work by Perry (1970) who found that upon entering college, students had a more dualistic view (knowledge is right or wrong), which changed into a more “relativistic” view during their academic career. With regard to the certainty of knowledge, naïve impressions about science are characterized by the view that knowledge in a specific domain is fixed, whereas more advanced beliefs include the view that knowledge in a specific domain is fluid and evolving (Hofer, 2000). Further dimensions refer to the source and justification of knowledge (varying from a strict reliance on experts to the view that a person has to actively construct knowledge her-/himself) or the attainability of truth (ranging from the view that there are true answers to the view that an ultimate truth is difficult to define). Research in the context of learning showed that readers with sophisticated epistemological beliefs are more successful in critically evaluating different sources of information (e.g., Mason, Ariasi, & Boldrin, 2011) and tend to employ more advanced strategies when searching for information on the Web (Kammerer & Gerjets, 2012). One goal

of the current research is to investigate whether this individual factor is also relevant for the process of attitude formation toward a scientific topic. Similar to the need for cognition, we expect that people who hold the belief that knowledge is tentative and fluid are more likely to consider representations of uncertainty and are more willing to take conflicting evidence into account. Therefore, we posit the following hypothesis: H3: Two-sided messages (in comparison to one-sided messages) will have a stronger influence on readers’ attitudes toward the topic when readers have more advanced epistemological beliefs regarding the certainty of knowledge. Method In order to test these hypotheses, we conducted an online experiment in which participants were asked to read an online science article (in different versions with different levels of expressions regarding scientific uncertainty). As an exemplary topic, the text dealt with the effects of violent video games on children and teenagers, which is discussed in the social sciences (e.g., Bushman, Rothstein, & Anderson, 2010; Ferguson & Kilburn, 2010) as well as in the media and the general public, particularly in the context of school shootings. Sample Due to the high relevance which the topic of violent video games should have for parents (who have to decide whether and to what extent their children should be allowed to play these games), the experiment was addressed at people who have children between the ages of 0 and 18. The link was distributed via mailing lists and online forums for parents. As an incentive, we ran a prize draw to give away four tokens for a large e-commerce site (125 Euro), which were distributed among the participants who entered their e-mail address in a separate form. The online questionnaire was programmed with the tool Soscisurvey (www.soscisurvey.de). Eighty-two persons (67 female, 15 male; age: M = 40.04; SD = 6.27) took part in the

experiment. Four participants were excluded from further analysis since they read the main article for less than 20 seconds. The remaining 78 respondents (65 female, 13 male) were on average 40.27 years (SD = 6.04) old. Thirty-two had a university degree, 26 had finished school with university entrance-level qualifications, and 20 had a lower level of education. Independent Variable and Stimulus Material An online article about the effects of violent computer games on children was used as stimulus material. First, a basic one-sided text about the dangers of computer games was written in a relatively neutral style; for the other versions, intensifiers, hedges, or a counterargument were added. This resulted in four experimental conditions (between-subjects factor) representing different levels of scientific uncertainty: 1) One-sided/basic: One-sided version about the dangers of computer games in a relatively neutral style (e.g., “An experimental study showed that realistic games with violent contents can raise the player’s level of aggression”, without hedges or intensifiers) 2) One-sided/assertive: One-sided version about the dangers of computer games with assertive statements (e.g., “without a doubt”, “Recent studies clearly demonstrate that…” and intensifiers such as “definitely” or “highly (dangerous)”) 3) One-sided/hedged: One-sided version about the dangers of computer games with lexical hedges (e.g., “partly”, “potential (danger)” or “could”) referring to interpretative statements (Durik et al., 2008) as well as short sentences about limitations (e.g., “It has to be mentioned that this study only compared three specific games, meaning that there are still open questions”) 4) Two-sided: A further argument on positive effects of computer games was added to the one-sided hedged version so that this version represents a two-sided way of argumentation. In order to control for the effects of order of presentation (Igou &

Bless, 2003), half of the participants in this condition received the pro argument at the beginning of the text and the other half received the additional paragraph at the end of the text. The basic content (an introduction about the rise of computer games and three arguments/paragraphs about negative effects) and the main arguments were the same in each condition. The only differences were the markers of assertiveness or hedging (placed at 15 positions in the text) and the additional argument in the two-sided condition. The final versions encompassed 551 (one-sided/basic) to 792 words (two-sided condition). Table 1 provides a paragraph from the text material in the different versions by way of example. The articles were displayed in the design of a science blog (see Figure 1). Participants were distributed almost equally across the four experimental conditions. INSERT FIGURE 1 AND TABLE 1 ABOUT HERE Pilot Study: Selection of Arguments To ensure that the arguments displayed in the article did not differ in strength, an online pilot study was conducted before the articles were composed. Nine arguments were tested, seven accentuating negative effects of computer games, and two accentuating positive effects. The arguments were presented in one or two sentences (in a random order) and were rated by 21 additional participants (14 female; age: M = 24.62, SD = 6.11) with regard to stance (positive vs. negative effects on one semantic differential from 1 to 7) and argument quality (based on three items, measuring strength, comprehensibility, and credibility, anchored from 1 to 7; scores for every argument were calculated: Cronbach’s α between .79 and .88). Based on the results, we chose the arguments with the highest scores on argument quality for the article (three with a negative stance toward computer games and one with a positive stance for the fourth condition). A repeated-measures analysis of variance (ANOVA) showed that the

quality of these arguments did not differ significantly. The chosen arguments on negative effects were as follows: Due to bonus points for violent behavior, games convey misleading values to children and teenagers; the interactivity in games particularly increases aggressive behavior; and realistic graphics strengthen identification with the action on the screen and potential effects of imitation. The positive argument stated that games can have positive effects on players’ attentiveness and coordination skills. In the final versions (see above), these arguments were outlined and explained in one paragraph each (including summaries of studies and citations of (fictitious) experts, see Table 1 for an example). Reader Characteristics Need for cognition. The readers’ tendency to enjoy and engage in complex thinking was measured with the established NC scale by Cacioppo and Petty (1982) in its German version (Bless, Wänke, Bohner, Fellhauer, & Schwarz, 1994). The scale contains 16 items such as “I only think as hard as I have to” (reverse-coded) or “I really enjoy a task that involves coming up with new solutions to problems”, anchored from 1 to 5. Items were averaged to form an NC score (Cronbach’s α = .82, M = 3.76, SD = 0.51). Epistemological beliefs. To measure participants’ views on the nature of science and knowledge, we used the questionnaire on epistemological beliefs by Hofer (2000). Participants were asked whether they agree (scale from 1 – strongly disagree – to 5 – strongly agree) with the following items with regard to knowledge in the domain of media research. Eight items belong to the dimension “Certainty of knowledge” (e.g., “Truth is unchanging in this subject”, Cronbach’s α = .76, M = 2.45, SD = 0.65). Further dimensions are justification of knowledge (4 items, e.g., “Correct answers in this field are more a matter of opinion than fact”, Cronbach’s α = .67, M = 2.88, SD = 0.73), source of knowledge (4 items, e.g. “Sometimes you just have to accept answers from the experts in this field, even if you don’t understand them”, Cronbach’s α = .60, M = 2.42, SD = 0.72) and attainability of truth (2

items, e.g., “Experts in this field can ultimately get to the truth”, Cronbach’s α = .61, M = 3.03, SD = 0.89). However, it has to be mentioned that the internal consistencies of these subdimensions were questionable. Dependent Measures As dependent measures, we assessed the participants’ evaluation of the science article and their attitudes toward the topic. The evaluation of the text was measured with eight items on semantic differentials (not credible – credible, low quality – high quality, poorly written – well-written, not useful – useful, dislikeable – likeable, not comprehensible – comprehensible, bad – good, not recommendable – recommendable, 7-point scale). The items showed a high internal consistency (Cronbach’s α = .95) and were averaged (M = 5.53, SD = 1.12). The participants’ attitudes toward the topic of the text (video games) were measured with five original items (“First-person shooter games such as Counter-Strike should be forbidden”, “Computer games have negative effects on children and teenagers”, “Violent computer games train children and teenagers for violence”, “Violent computer games are not a problem for the development of teenagers” (reverse-coded), and “Spending too much time on the computer has negative effects on children’s school performance”, Cronbach’s α = .71). For each item, participants indicated their agreement/disagreement with the presented statement on a 7-point Likert scale (M = 5.01, SD = 1.15). Furthermore, participants’ attitudes toward general effects of violent media were measured with 6 original items (“All in all, I think that violent media contents have a negative influence on children and teenagers”, “Violent media contents increase the likelihood of aggressive behavior in children and teenagers”, “To my mind, the consumption of violent media contents is harmless” (reverse-coded), “The consumption of media contents can have

negative effects on the development of children and teenagers”, “Children and teenagers should not consume violent media contents”, and “The consumption of depictions of violence in the media can be reasonable for children and teenagers to relieve aggression” (reversecoded), Cronbach’s α = .84; M = 5.73, SD = 1.11). Additionally, participants expressed their opinions about general media consumption of children and teenagers with the help of a semantic differential (items: not dangerous – dangerous, positive – negative, harmless – harmful, reasonable – not reasonable; 7-point scale; Cronbach’s α = .91, M = 3.98, SD = 1.09). Procedure The opening page displayed a text in which an online questionnaire about the topic of violence in the media and effects on children and teenagers and a test of an online information site was announced. The questionnaire was addressed to parents (of minors). After agreeing to participate, respondents were randomly assigned to one of the four conditions of text type (one-sided/basic, one-sided/assertive, one-sided/hedged, two-sided). The article was displayed (in the specific version of the condition) as a screenshot in the design of a science blog. Participants were asked to take as much time as necessary to read the article carefully. After reading, participants rated the article and filled out the questionnaire on their attitudes toward the topic and their individual characteristics. Finally, they were debriefed about the experimental manipulation of the texts and thanked for their participation. Results Preliminary analyses on the reader characteristics showed a moderate correlation between need for cognition and the domain-specific epistemological beliefs about the certainty of knowledge (r = -.362, p = .001, with a tendency for people with high NC to express less naïve epistemological beliefs).

For hypothesis testing, we conducted hierarchical regression analyses. In the first block, we entered the reader characteristics NC and epistemological beliefs about the certainty of knowledge. Given that males are more likely to be exposed to video games, especially violent ones, than females (Hartmann & Klimmt, 2006; Lucas & Sherry, 2004; Quandt, Breuer, Festl, & Scharkow, 2013), participants’ gender was also included to control for presumable differences regarding the attitudes toward video games. In the second block, we entered the experimental manipulations of the text regarding scientific uncertainty. Specifically, we compared the basic version with the assertive version (Text type: assertive vs. basic), the hedged version (Text type: hedged vs. basic) and the two-sided version (Text type: two-sided vs. basic) with the help of dummy variables (Aiken & West, 1991). To test the hypothesized moderation effects, interactions of text types and the reader characteristics NC and epistemological beliefs were entered in the final step (these interactions were computed as products of the centralized variables). For the criterion of readers’ evaluation of the text, the regression model was not significant. With regard to RQ1, this suggests that there were no significant effects of text type. According to the mean value (M = 5.53 on a 7-point scale, SD = 1.12), participants generally expressed a rather positive evaluation and rated the article as relatively credible, which was not affected by the different representations of uncertainty. H1a, b, and c addressed the effects of text type on readers’ attitudes toward the topic and H2 and H3 posited moderation effects of the reader characteristics. Table 2 summarizes the regression analyses for the criterion of readers’ attitudes toward computer games. In the first step (F (3, 74) = 3.54, p = .019, R² = .125), gender emerged as a significant predictor, indicating that men expressed a less negative attitude than women (however, it should be noted that this is based on relatively few male participants). In the second step (F (6, 71) = 2.90, p = .014, R² = .197), regression analysis showed a significant influence of the variables

regarding the two-sided and the assertive version. The negative beta coefficient shows that readers of the two-sided version expressed a less critical and more moderate view on computer games than readers of the one-sided basic version, which is in line with H1c. However, contrary to H1a, which posited a stronger persuasive effect of powerful language, readers of the assertive version also expressed a less negative attitude, indicating that the assertive version with intensifiers was less persuasive than the basic version. The hedged version (H1b) did not have a significant impact on readers’ attitudes. H2 predicted that two-sided messages would have a stronger persuasive influence on readers with a greater need for cognition. In this regard, the regression model in the third step (F (9, 68) = 2.60, p = .012, R² = .256; see Table 2) revealed that a marginally significant interaction of the text type variable concerning the two-sided message and NC added to the explanation of variance. The main effect of the text type variable concerning the two-sided version was attenuated by the interaction and was slightly beyond the conventional level of significance in the final model. In order to investigate this interaction, we conducted simple slope analyses (Aiken & West, 1991): Results showed that participants with higher NC were affected by the additional argument of the two-sided version insofar as their attitude was less negative in comparison to the one-sided basic version (b = -1,33, SE = 0.47, t = 2.85, p = .006), while participants with lower NC were not affected by the text type (see Figure 2). This pattern is consistent with the predictions of H2. INSERT TABLE 2 AND FIGURE 2 ABOUT HERE H3 addressed the effects of readers’ epistemological beliefs concerning the certainty of knowledge on the effects of one- vs. two-sided messages. To examine the hypothesis, interactions of the text type variables and epistemological beliefs were computed and entered following the same scheme as above. Table 3 summarizes the final regression model (F (9, 68) = 3.09, p = .004, R² = .290). As expected, results showed a significant influence of the

interaction between text type (two-sided vs. one-sided basic) and epistemological beliefs. Since the dummy variable is no longer a significant predictor in the final model, the interaction also appears to qualify the main effect of the two-sided version. Simple slope analyses showed that readers with sophisticated beliefs about the certainty of knowledge were affected by the two-sided version (b = -1.49, SE = 0.44, t = 3.41, p = .001), while this pattern was not found for people with naïve beliefs (see Figure 3). That is, participants who hold the view that knowledge is fluid and evolving considered the additional argument of the twosided version and had a less negative attitude than readers of the one-sided version, whereas participants who believe that knowledge is fixed did not take the argument into account, which supports H3. For the related measures, readers’ attitudes toward media violence and media consumption in general, the regression models were not significant. INSERT TABLE 3 AND FIGURE 3 ABOUT HERE Discussion In the context of current scientific debates, empirical studies often provide patterns of partly conflicting evidence. When writing about scientific findings, communicators face the question of whether or how they should refer to these levels of tentativeness or opposing findings. Against this background, the present research aimed to investigate how readers deal with different textual representations of scientific uncertainty and how attitudes toward scientific topics are formed when reading science articles on the Internet. For this purpose, a text with three arguments about negative effects of violent video games was shown in different styles of presentation (one-sided: basic neutral style vs. assertive style vs. hedged version), the fourth condition contained one additional argument on positive effects (two-sided argumentation; Allen, 1991; O’Keefe, 1999). The article was presented in the design of a

science blog, since Web 2.0 can be regarded as a popular platform for receiving information on science-related topics. Due to the presumably high relevance of the debate on media violence for this group, the experiment was addressed at parents (of minors). While readers’ evaluation of the article in terms of credibility and quality (RQ1) was not affected (participants rated each version of the blog article as relatively credible), their attitudes toward the topic were influenced by the different depictions of scientific uncertainty. Specifically, participants who received the one-sided version in the neutral mode of presentation had a more negative attitude toward the effects of violent computer games than readers of the two-sided version, who expressed a more moderate attitude (H1c). Since twosided messages can be seen as a more appropriate representation of the state of knowledge in current scientific debates, our finding suggests that this pattern of conflicting evidence is considered for evaluation by the majority of recipients who wish to reach an informed point of view on the topic. However, the hedged version (including references to limitations and words like “might” or “could” when describing conclusions) did not lead to a significantly less negative attitude (H1b). One potential conclusion might be that hedges, as a more implicit representation of uncertainty, are not salient enough to induce more moderate attitudes – however, from a more persuasion-oriented perspective, this might also suggest similarities to the study by Durik et al. (2008), which also did not find negative effects (in the sense of less agreement with the position of the text) of hedges on interpretative statements. Contrary to expectations with regard to powerful language (Burrell & Koper, 1998), the one-sided assertive version with intensifiers such as “without a doubt” was not effective in strengthening participants’ negative views (H1a). Instead, readers of the assertive version expressed a significantly less negative attitude toward computer games than readers of the neutrally presented text. This means that readers were not persuaded by powerful formulations which described scientific evidence as very certain but seemed to be skeptical

when information was presented as too simple. At first glance, this finding appears to contradict previous research which showed a negative relationship between markers of powerlessness and persuasion (Burrell & Koper, 1998; Blankenship & Holtgraves, 2005; Hosman, 2002). However, the differences may be explained through the specific context of science communication: While hedges appear to be negative in courtroom settings (the domain to which most studies on powerful language referred), markers of powerlessness are evaluated more positively in the domain of science, in which uncertainty is a basic concept (Popper, 1968). In line with findings by Jensen (2008), this suggests that these differences are also considered by laypersons. As a personality characteristic which has been shown to influence elaboration likelihood, the need for cognition (Cacioppo & Petty, 1982) was taken into account (H2). Moderation analyses revealed that participants high in NC were particularly affected by the two-sided version (compared to the one-sided neutral version) insofar as the additional argument had an effect on their attitude, while this was not the case for participants with lower NC. While classical research on persuasion in the context of the ELM showed that people who are generally motivated to engage in effortful thinking more often process information on the central route and are mainly influenced by the quality of arguments (Cacioppo et al., 1996), this finding indicates that NC also increases the receptivity to twosided information. People high in NC, who tend to enjoy complex thinking, appear to be more likely to detect conflicts and to process counterarguments. However, it has to be mentioned that the effect of the interaction between text type and NC was relatively weak. In this setting of processing a blog article on a science-related topic, a further individual characteristic was considered: readers’ personal views on the nature of science and knowledge (epistemological beliefs; Hofer, 2000; Schommer, 1990). These beliefs range from naïve views (that there are certain and unchanging answers) to advanced views that

knowledge is tentative and fluid. While this variable has been focused in research on learning (Mason et al., 2011) and strategies of Web search (Kammerer & Gerjets, 2012), the current study aimed to investigate this factor in the context of attitude formation. In line with our expectations (H3), results showed that beliefs about the certainty of knowledge in the domain had a significant influence on the way in which one-sided vs. two-sided information is processed. Participants with sophisticated beliefs expressed a less negative attitude toward the topic after reading the two-sided version, suggesting that they carefully processed the additional counter-argument and took this piece of conflicting information into account. In contrast, participants who held the belief that knowledge is fixed were not affected by the two-sided version (in comparison to the one-sided neutral text). One explanation for this could be that these readers who do not see uncertainty as a basic concept of empirical evidence were not able or willing to cope with conflicting information and therefore decided to follow the position which was presented as predominant in the text (and ignored the counterargument). Thus, epistemological beliefs about the certainty of knowledge can be regarded as an individual characteristic which leads to a stronger consideration of a two-sided presentation of evidence. In the final regression model, the interaction emerged as the most important predictor of readers’ attitudes and attenuated the main effect of text variable regarding the two-sided message. This suggests that the differences between the two-sided and the one-sided basic version that emerged in the whole sample are mainly based on readers with advanced epistemological beliefs. From a theoretical perspective, these results extend the ELM (Petty & Cacioppo, 1986; Petty et al., 2004) with regard to the effects of message sidedness and uncertainty as well as the processing of information on scientific topics. While the original model posits that people high in NC more often focus on the quality of arguments, our results rather refer to differences in the style of presentation and to the question of whether only arguments for one position or also counterarguments are presented: As mentioned above, the results showed the

tendency that readers high in NC are more likely to detect conflicts and include them in their attitudes. However, while this effect was relatively weak, epistemological beliefs about the certainty of knowledge were shown to be an individual characteristic which is even more relevant in this setting than NC and should therefore be included in analyses of the formation of attitudes toward science-related topics. In terms of its practical implications, the study indicates that laypersons who read science-related texts online engage in relatively elaborate ways of processing information and are somewhat sensitive to (relatively small) differences in style and content. When authors of science texts aim to inform the public on patterns of conflicting evidence in a current scientific debate, the two-sided way of argumentation including hedges can be interpreted as the most appropriate mode of presentation (as a normative consideration). From the results of the study, we can derive that readers (at least those with a higher need for cognition and advanced epistemological beliefs) pay attention to relevant information on scientific uncertainty and that a two-sided blog article leads to a more moderate and relativistic attitude (that is, such a representation of uncertainty can be seen as successful in influencing laypersons’ attitude toward the inclusion of conflicting evidence). Therefore, one way to advance laypersons’ knowledge about science could be to emphasize limitations and patterns of conflicting evidence, instead of playing them down. In this respect, science blogs in particular may offer possibilities to explain topics on platforms which have fewer space restrictions than traditional media. If science communicators have the contrary goal of persuading readers of a specific position (which may be justifiable in discussions about specific decisions or in scientific domains in which there is no disagreement between scientists and patterns of evidence appear to be less uncertain), a one-sided way with relatively neutral statements appears to be the most effective strategy. Assertive statements which aim to emphasize the certainty of knowledge, however, should be avoided since readers seem to be skeptical or sensitive when findings are presented as overly certain.

In terms of hedging the conclusions of the present study as well, we are aware of several limitations that deserve attention. First, the experiment was based on only one text and a specific pattern of manipulations, meaning that the conclusions should be tested with further material and other topics than the debate on violent video games. A more extensive setting could also include a simultaneous variation of text style and message sidedness, as the present research was restricted to four exemplary versions. Second, we only focus on short-term influences on attitudes (measurement directly after the reception of the text): To derive practical implications, it would also be useful to measure long-term effects or the influence on behavioral intentions. Furthermore, it has to be mentioned that readers’ attitude was only assessed after exposure to the stimulus and not before: This was done to avoid demand effects of the experimental situation but it also complicates statements about attitude change (despite the random assignment to the conditions, differences between the groups in terms of the preattitudes might have remained). Third, the experiment was limited by the small sample size and an unequal representation of male and female respondents. Although we believe that parents are a relevant target group in this context (which offers higher external validity than usual student samples), future research should try to retest the findings with a larger and more diverse sample of participants. Furthermore, the factor of source (who is the writer of the specific blog article and who hedges the statements?) was neglected in this setting, which could also be an influential determinant in the reality of the Web. Despite these limitations, the present study provides several contributions to the analysis of the formation of attitudes toward scientific topics. Results indicate that laypersons take counterarguments in two-sided messages as representations of scientific uncertainty into account when reading online information and tend to be skeptical when findings are presented as overly certain. This might encourage authors of (online) articles to explain scientific results to the public in a more complex manner. With the regard to the mechanisms of the ELM in the context of science topics, advanced epistemological beliefs and (to a lesser extent) need

for cognition strengthen the effects of conflicting information on readers’ attitudes. While these findings are based on the example of a science blog on the debate about violent video games, future research which considers further topics and takes into account the interplay of source and message factors may help to further elucidate the issue of how laypersons cope with the challenging task of making sense of scientific information. Acknowledgments The authors would like to thank Howie Giles and two anonymous reviewers for their helpful suggestions. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the authorship and/or publication of this article. Funding This research was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) in the Special Priority Program 1409 “Science and the General Public” (Kr 2240/2).

References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Allen, M. (1991). Meta-analysis comparing the persuasiveness of one-sided and two-sided messages. Western Journal of Speech Communication, 55, 390–404. Blankenship, K. L., & Craig, T. Y. (2011). Language use and persuasion: Multiple roles for linguistic styles. Social and Personality Psychology Compass, 5, 194–205. Blankenship, K. L., & Holtgraves, T. (2005). The role of different markers of linguistic powerlessness in persuasion. Journal of Language and Social Psychology, 24, 3–24. Bless, H., Wänke, M., Bohner, G., Fellhauer, R. F., & Schwarz, N. (1994). Need for Cognition: Eine Skala zur Erfassung von Engagement und Freude bei Denkaufgaben [Need for cognition: A scale for the assessment of involvement and enjoyment of cognitive tasks]. Zeitschrift für Sozialpsychologie, 25, 147–154. Bromme, R., Kienhues, D., & Porsch, T. (2010). Who knows what and who can we believe? Epistemological beliefs are beliefs about knowledge (mostly) attained from others. In L. D. Bendixen & F. C. Feucht (Eds.), Personal epistemology in the classroom: Theory, research, and implications for practice (pp. 163–193). Cambridge, England: Cambridge University Press. Brossard, D., & Scheufele, D. A. (2013). Science, new media, and the public. Science, 339, 40–41. Burrell, N. A., & Koper, R. J. (1998). The efficacy of powerful/powerless language on attitudes and source credibility. In M. Allen & R. W. Preiss (Eds.), Persuasion: Advances through meta-analysis (pp. 203–217). Cresskill, NJ: Hampton Press, Inc. Bushman, B. J., Rothstein, H. R., & Anderson, C. A. (2010). Much ado about something: Violent video game effects and a school of red herring − Reply to Ferguson and Kilburn (2010). Psychological Bulletin, 136, 182–187.

Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of Personality and Social Psychology, 42, 116–131. Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin, 119, 197–253. Cohen, A. R., Scotland, E., & Wolfe, D. M. (1955). An experimental investigation of need for cognition. Journal of Abnormal and Social Psychology, 51, 291–294. Crowley, A. E., & Hoyer, W. D. (1994). An integrative framework for understanding twosided persuasion. Journal of Consumer Research, 20, 561–574. Dixon, G., & Clarke, C. (2013). Heightening uncertainty around certain science: Media coverage, false balance, and the autism-vaccine controversy. Science Communication, 35, 358–382. Durik, A. M., Britt, M. A., Reynolds, R., & Storey, J. (2008). The effects of hedges in persuasive arguments: A nuanced analysis of language. Journal of Language and Social Psychology, 27, 217–234. Ferguson, C. J., & Kilburn, J. (2010). Much ado about nothing: The misestimation and overinterpretation of violent video game effects in Eastern and Western nations: Comment on Anderson et al. (2010). Psychological Bulletin, 136, 174–178. Friedman, S. M., Dunwoody, S., & Rogers, C. L. (Eds.). (1999). Communicating uncertainty: Media coverage of new and controversial science. Mahwah, NJ: Lawrence Erlbaum Gibbons, P., Busch, J., & Bradac, J. J. (1991). Powerful versus powerless language: Consequences for persuasion, impression formation, and cognitive response. Journal of Language and Social Psychology, 10, 115–133. Hale, J. L., Mongeau, P. A., & Thomas, R. M. (1991). Cognitive processing of one- and twosided persuasive messages. Western Journal of Speech Communication, 55, 380–389.

Hamilton, M. A., Hunter, J. E., & Burgoon, M. (1990). An empirical test of an axiomatic model of the relationship between language intensity and persuasion. Journal of Language and Social Psychology, 9, 235–255. Hartmann, T., & Klimmt, C. (2006). Gender and computer games: Exploring females’ dislikes. Journal of Computer-Mediated Communication, 11, 910-931. Hofer, B. K. (2000). Dimensionality and disciplinary differences in personal epistemology. Contemporary Educational Psychology, 25, 378–405. Hosman, L. A. (2002). Language and persuasion. In J. P. Dillard & M. Pfau (Eds.), The persuasion handbook: Theory and practice (pp. 371–390). Thousand Oaks: Sage. Hosman, L. A., & Siltanen, S. A. (2011). Hedges, tag questions, message processing, and persuasion. Journal of Language and Social Psychology, 30, 341-349. Igou, E. R., & Bless, H. (2003). Inferring the importance of arguments: Order effects and conversational rules. Journal of Experimental Social Psychology, 39, 91–99. Jensen, J. D. (2008). Scientific uncertainty in news coverage of cancer research: Effects of hedging on scientists’ and journalists’ credibility. Human Communication Research, 34, 347–369. Kammerer, Y., & Gerjets, P. (2012). Effects of search interface and internet-specific epistemic beliefs on source evaluations during web search for medical information: An eye-tracking study. Behaviour and Information Technology, 31, 83–97. Kardash, C. A. M., & Scholes, R. J. (1996). Effects of pre-existing beliefs, epistemological beliefs, and need for cognition on interpretation of controversial issues. Journal of Educational Psychology, 88, 260–271. Lakoff, R. (1975). Language and woman's place. New York, NY: Harper & Row. Lucas, K., & Sherry, J. L. (2004). Sex differences in video game play: A communicationbased explanation. Communication Research, 31, 499-523.

Markkanen, R., & Schröder, H. (1997). Hedging: A challenge for pragmatics and discourse analysis. In R. Markkanen & H. Schröder (Eds.), Hedging and discourse: Approaches to the analysis of a pragmatic phenomenon in academic texts (pp. 3–18). Berlin, Germany: De Gruyter. Mason, L., Ariasi, N. & Boldrin, A. (2011). Epistemic beliefs in action: Spontaneous reflections about knowledge and knowing during online information searching and their influence on learning. Learning and Instruction, 21, 137–151. O’Keefe, D. J. (1999). How to handle opposing arguments in persuasive messages: A metaanalytic review of the effects of one-sided and two-sided messages. In M. E. Roloff (Ed.), Communication yearbook 22 (pp. 209–249). Mahwah, NJ: Lawrence Erlbaum. Perry, W. G., Jr. (1970). Forms of intellectual and ethical development in the college years: A scheme. New York, NY: Holt, Rinehart, and Winston. Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion. Central and peripheral routes to attitude change. New York, NY: Springer. Petty, R. E., Rucker, D. D., Bizer G. Y., & Cacioppo, J. T. (2004). The elaboration likelihood model of persuasion. In J. S. Seiter & R. H. Gass (Eds.), Perspectives on persuasion, social influence, and compliance gaining (pp. 65–89). Boston, MA: Allyn & Bacon. Popper, K. R. (1968). The logic of scientific discovery. New York, NY : Harper & Row. Quandt, T., Breuer, J., Festl, R., & Scharkow, M. (2013). Digitale Spiele: Stabile Nutzung in einem dynamischen Markt [Digital games: Stable use in a dynamic market]. Media Perspektiven, 10, 483-492. Schommer, M. (1990). Effects of beliefs about the nature of knowledge on comprehension. Journal of Educational Psychology, 82, 498–504. Winter, S., & Krämer, N. C. (2012). Selecting science information in Web 2.0: How source cues, message sidedness, and need for cognition influence users' exposure to blog posts. Journal of Computer-Mediated Communication, 18, 80–96.

Author Biographies Stephan Winter is a research associate at the University of Duisburg-Essen, Germany. He earned his PhD in 2012 with a thesis on selective exposure and information processing on online news sites. His research interests include persuasion and credibility in online contexts, science communication, and self-presentation in social media applications. Nicole C. Krämer is professor for social psychology: media and communication at the University of Duisburg-Essen. She finished her PhD in 2001 and received the venia legendi for psychology in 2006 with a thesis on social effects of embodied conversational agents. Her current work focuses on social psychological aspects of Web 2.0 applications as well as on social effects of human-technology interaction. Leonie Rösner is a research assistant and doctoral student in social and media psychology at the University of Duisburg-Essen. She is interested in social influence in Web 2.0 settings as well as in crisis and risk communication. German Neubaum is a PhD student at the University of Duisburg-Essen and a research assistant in the department of social psychology. His research interests include online crisis and science communication as well as processes of online persuasion and opinion expression.

Table 1. Examples of the stimulus material (translated from the original German): One paragraph about the effects of violent video games in four versions (one-sided/basic, onesided/assertive, one-sided/hedged, two-sided). Experimental

Excerpt from the text (variations are highlighted)

condition One-sided/

“In many computer games, aggressive behavior and violence is rewarded by

basic

points and progress in the game. Players use violence as a method to gain points, to get rich, or to solve quests and problems. For instance, actions such as theft, burglary, and robbery are presented as legitimate means to acquire money, or killing the enemy player through a perfect shot in the head is rewarded with lots of points. According to Joachim Zeis, media psychologist and expert on media violence, “encouraging robbing and killing in the virtual world due to the lack of sanctions is problematic and has a dangerous influence on the development of young people. Through this, today’s children and teenagers are unaware of the fact that violence is a bad thing that is punished in real life.” […]

One-sided/

“In many computer games, aggressive behavior and violence is rewarded by

assertive

points and progress in the game. Players use violence as a key method to gain points, to get rich, or to solve quests and problems. For instance, actions such as theft, burglary, and robbery are presented as completely legitimate means to acquire money, or killing the enemy player through a perfect shot in the head is rewarded with lots of points. According to Joachim Zeis, media psychologist and expert on media violence, it is unequivocally clear that “encouraging robbing and killing in the virtual world due to the lack of sanctions is problematic without a doubt and has an extremely dangerous influence on the development of young people. Through this, today’s children and teenagers are totally unaware of the fact that violence is a bad thing that is punished in real life.” […]

One-sided/

“In some computer games, aggressive behavior and violence is rewarded by

hedged

points and progress in the game. In some cases, players use violence as a method to gain points, to get rich, or to solve quests and problems. For instance, actions such as theft, burglary, and robbery are presented as legitimate means to acquire

money, or killing the enemy player through a perfect shot in the head is rewarded with lots of points. However, this needs to be qualified by stating that not every action game automatically rewards players with higher scores for using more violence. Nevertheless, the use of force in games remains a serious issue. According to Joachim Zeis, media psychologist and expert on media violence, “encouraging robbing and killing in the virtual world due to the lack of sanctions can be problematic and might have a dangerous influence on the development of young people. Through this, today’s children and teenagers might be unaware of the fact that violence is a bad thing that is punished in real life.” two-sided

[hedged version + the following counterargument] However, it should be noted that computer games are not an inevitable cause of violence and antisocial behavior. On the contrary, they sometimes have a high educational potential and can have a positive impact. In a study conducted by the Munich Institute of Media Education it was shown that computer games promote feelings of competence and positively affect the ability to make moral judgments, perception and attention, sensorimotor coordination, emotional selfcontrol as well as players’ critical self-reflection. The games tested in the study also included so-called “shoot-‘em-ups” such as Counter-Strike: In the interviews, the players reported that it is less about killing other players, but rather a question of strategic aspects and team play. Compared to educational games, which explicitly pursue pedagogical objectives, classical games have the advantage that players’ motivation is higher, which is a prerequisite for successful learning. Thus, these findings indicate that violent games are not invariably problematic.

Note. Full versions of the material are available from the authors.

Table 2. Hierarchical regression analysis: Effects of control variables, text types, and interactions with need for cognition on readers’ attitudes toward the topic (in the final model, coefficients for all predictors are presented). Attitudes Toward Computer Games β

p

Gender

–.324

.005

Need for Cognition (NC)

.052

.656

Epistemological beliefs (EB)

.097

.422

Assertive vs. basic text

–.308

.028

Hedged vs. basic text

–.155

.245

Two-sided vs. basic text

–.280

.042

Gender

–.310

.007

Need for Cognition (NC)

.091

.453

Epistemological beliefs (EB)

.153

.223

Assertive vs. basic text

–.317

.025

Hedged vs. basic text

–.169

.213

Two-sided vs. basic text

–.269

.051

Interaction: Assertive * NC

–.193

.116

Interaction: Hedged * NC

–.025

.836

Interaction: Two-sided * NC

–.237

.060

Predictor Step 1

Step 2

Step 3

R² .125

.197

.256

Table 3. Hierarchical regression analysis: Effects of control variables, text types, and interactions with epistemological beliefs (certainty of knowledge) on readers’ attitudes toward the topic (Final model in the third step, for steps 1 and 2 see Table 2). Attitudes Toward Computer Games β

p

Gender

–.334

.003

Need for Cognition (NC)

.096

.413

Epistemological beliefs (EB)

.208

.100

Assertive vs. basic text

–.269

.049

Hedged vs. basic text

–.141

.271

Two-sided vs. basic text

–.143

.304

Interaction: Assertive * EB

.097

.481

Interaction: Hedged * EB

.116

.363

Interaction: Two-sided * EB

.380

.005

Predictor



.290

Figure 1. Stimulus material: Screenshot of the blog article.

Figure 2. Simple slopes: Interaction of text type (one-sided basic version vs. two-sided version) and need for cognition on readers’ attitudes toward the topic. Estimations are based on the regression’s coefficients (for subjects with one standard deviation below and above the mean of the group’s NC).

Figure 3. Simple slopes: Interaction of text type (one-sided basic version vs. two-sided version) and epistemological beliefs (certainty of knowledge) on readers’ attitudes toward the topic. Estimations are based on the regression’s coefficients (for subjects with one standard deviation below and above the mean of the group’s EBs).