Framing and Framing Effects

Framing is concerned with the presentation of information. Therefore, framing effects can be defined as the effects of differences in an issue’s presentation on how people perceive and understand a topic [5,9,10]. These effects are moderated by factors such as the recipients' issue involvement, perceived importance [11–13], and prior knowledge [13–16].

One classic example of framing research is the “Asian disease study” by Kahneman and Tversky [17]: Participants were much more likely to choose the variation of a program to fight an “Asian disease” naming a number of certain survivors over a factually equivalent program naming a number of certain deaths. This kind of framing is also referred to as equivalence framing, that is, framing which directs attention to different aspects in a factually identical description [18]. Equivalence framing should be distinguished from emphasis framing which suggests a certain perspective by emphasizing specific, equally true, yet different subsets of aspects of a topic without the assumption that the information presented is necessarily factually equivalent [19,20]. In other words, the main difference is that while equivalence frames present factually equivalent information from different perspectives, emphasis frames present information that suggests a certain perspective, yet is not necessarily factually equivalent between different frames. Emphasis framing is common in public media and is therefore often found in research on media effects and specifically news frames.

Effects of news frames on perception of risk, emotions, and learning

News frames exert their effects by emphasizing the “salience of different aspects of a topic” [1], p. 53 (cf. emphasis frame). Lang and colleagues [21] highlighted the importance of psychological theories for understanding cognitive processes underlying media effects in general, and framing effects in particular. Specifically, theories of mental models [22] and situation models in text comprehension [23] provide important frameworks for understanding cognitive media effects [24]. A situation model is essentially a mental model of the situation or issues described in the text, which the reader builds by enriching the information given in the text (text-base) with prior knowledge. Additionally, the construction-integration model of text comprehension proposes that linguistic cues guide the readers´ attention while reading the text, thereby influencing the focus of the situation model [25]. Despite the importance of cognitive theories to explain framing effects, there are few studies directly addressing and systematically measuring cognitive effects of news frames on information processing and learning [26,27]. The study of Berinsky and Kinder [28] is one of the rare examples in which framing effects on learning were investigated and embedded in a cognitive theory. Their results are related to psychological theories on information processing [23,29] and corroborate the assumption of framing theories that an event’s specific framing helps people understand complex sequences of events and create a coherent story [18,30,31]. Emphasizing a particular storyline may also direct readers´ attention toward certain aspects of the information while drawing it away from others, leading to an effect similar to the seductive details effect in multimedia learning [32]. However, while the term seductive details is usually used for interesting yet irrelevant informational aspects, framing draws attention to relevant aspects of information while detracting it away from other equally relevant aspects. In relation to the seductive details effect, this effect may therefore be designated as a seductive emphasis effect [33].

Studies about framing effects in the media identified the following widely used journalistic news frames: the conflict frame, attribution of responsibility frame, morality frame, economic consequences frame, and human-interest frame [34–36]. Responsibility frames emphasize the responsibility of an individual or a group for a certain event or outcome, while the morality frame presents an issue in the context of religious or ethical prescriptions. The economic consequences frame focuses on economic consequences for individuals or groups. Finally, human-interest frames use specific features, such as dramatizing or emotionalizing vocabulary to catch the audience´s attention and often lend the story an individual, personalized touch [35].

Although human-interest frames are widely used in news stories, there is little research on the specific effects of this frame on emotions, perception of risk, and learning [36]. From an experimental perspective, the human-interest frame is particularly interesting as it consists of different features, which have been studied separately before [6,21,33,37,38]. However, we were interested, if the human-interest frame as a whole elicits similar effects as its separate features. Therefore, it seems promising to manipulate the strength of these features, emphasising one or the other more, while keeping all necessary characteristics of the human-interest frame as defined by Semetko and Valkenburg [35], and to test if effects can be “generalized” over different variants of the frame. This variation of emphasis allows for controlling that it is really “the sum of its parts” and not merely one of the features of the frame that produces typical human-interest frame effects on emotions, perception of risk, and information processing.

Different journalistic news frames are known to exert differential emotional responses [6] and to alter perceived risk [33]. Appraisal theories of emotion and the model of social amplification of risk respectively, may explain framing effects on emotions and perception of risk. Within the framework of appraisal theories, appraisal is defined as the cognitive evaluation of a situation with respect to the demands, constraints, and resources in the situation on the one hand, and the individual’s goals and resources on the other [39–41], (for an overview see [42]:). Similarly, the cognitive evaluation of a situation, especially the perceived severity and probability of negative outcomes, influences emotions and perceived risk [43]. The cognitive evaluation (or appraisal) of the situation is likely to be influenced in both cases by the difference in the salience of certain aspects achieved by framing. Additionally, the model of social amplification of risk assumes a strong media influence on both the public discourse and each individual’s and the social group’s perceived risk [44]. Recent theories on the perception of risk additionally assume an interaction between the perception of risk and emotions [45,46]. For example, perceived risk may be considered a specific form of appraising a situation (as risky versus not so risky), and therefore potentially influence emotions [47].

Additionally, connecting emotions with cognitive theories, the personality systems interaction theory assumes that positive emotions such as happiness activate rather wide semantic fields and lead to holistic information processing, while negative emotions such as anger lead to a focus on rather narrow informational aspects, that is, the spread of activation is limited to close associates and dominant words [48]. Assuming that human-interest frames lead to stronger negative emotions, this could have two effects on learning: First, the higher activation level should lead to better (quantitative) learning outcome in general (cf. cognitive motivational mediation model of emotions [7]). Second, restricting the spread of activation should lead to focusing on only those aspects that seem to be particularly relevant to the frame. Both might be particularly true for complex and controversial topics that demand substantive processing [49]. Additionally, specifically moderate arousal (i.e., arousal that is neither specifically high or low) may enhance memory of frame-specific, negative information [38].

One problematic aspect in most studies investigating news frame effects is, that the systematically different information about the topics at hand - which results from the fact that the contents of emphasis frames naturally differ to some extent from one another - makes it difficult to distinguish between content and actual framing effects. For example, Nelson and colleagues [50] framed a Ku Klux Klan rally as either a free speech issue or a disruption of public order to investigate framing effects on, for example, tolerance of the Ku Klux Klan. Igartua and colleagues [15] framed the issue of an increase in immigration in Spain as either causing positive economic consequences or a rise in criminality, and measured framing effects on attitudes towards immigrants, emotions, and cognition.

Some researchers aimed to separate content from framing effects by implementing the news frames only in the headline, introduction and concluding paragraphs of newspaper articles while keeping the core component of the text the same for all conditions [3,4]. Valkenburg and colleagues [4] compared effects of different news frames (conflict, human interest, responsibility, and economic consequences) and one control condition, which consisted only of the common core text. Additionally, they used two different cover stories, namely the introduction of the Euro and increasing crime rates. They found, in accordance with Price and colleagues [3], that readers' thoughts referring to the text as a whole (differently framed parts as well as common parts) are strongly influenced by the frame. Additionally, they found a negative effect on recall measured by multiple-choice test (only items referring to the information identical for all conditions) for the human-interest frame in comparison to all other frames, specifically for the crime story but not the introduction-of-the-Euro story. They hypothesized, that it was likely the emotional nature of the topic that led to stronger emotional reactions (i.e., high emotional arousal) and therefore poorer recall.

More recently, Von Sikorski and Schierl [13] attempted to separate the confounding of frame and content by systematically varying only those sentences that contained frame-relevant information while keeping the core text constant. They compared reports on disability sports concerning their effects on the attitude towards an athlete described in the texts 1 frame placed emphasis on the athlete’s performance, while the other emphasized his dependency on public support. Additionally, they introduced a third, so called “mixed frame” providing information from both main frames. As expected, participants´ attitudes in the mixed-frame condition fell in between the attitudes reported in both main-frame conditions. This result indicates that what is sometimes presented as a “framing effect” in research employing emphasis frames might at least in part be a “content effect”.

The Present Study

Our study focuses on the effects of the human-interest frame compared to a neutral text on perceived risk, emotions, and learning. More specifically, we compared a neutral text (control condition) with three variants of the human-interest frame, which contained an emphasis on the different features constituting the human-interest frame, that is, either dramatizing or emotionalizing vocabulary, or an individualized storyline. We did not attempt to completely separate those three aspects of the human-interest frame but rather to present ecologically valid versions of the frame with specific emphases. Hence, we did not investigate specific differences among the three versions of the human-interest frame but rather used the variation to test for the "generalizability" of our results to common variants of the human-interest frame instead of comparing only one kind of human-interest frame to a neutral text. That is, the variations in the human-interest frames allowed us to control for the possibility, that merely one feature of the human-interest frame is responsible for human-interest frame effects.

In order to measure framing effects independent of content, we followed the design of Valkenburg and colleagues [4] and kept the core component of our text constant for all conditions and only varied the headline, introduction, and concluding paragraph. To avoid influences of text length, also participants in our control condition were presented with an introduction and conclusion, which, however, was written in rather neutral language not employing any frame-specific features. Additionally, we separated the differently framed introductions and concluding paragraphs thematically. More specifically, we used the topic of invasive species in the differently framed introduction and conclusion sections and the topic of the Asian ladybug, one example for invasive species, in the common core text. Reported framing effects refer to information given in the invariable core text only, unless explicitly stated otherwise. The topics of invasive species in general and the Asian ladybug in particular seem particularly suited for investigating framing effects as they are highly complex topics about which the general public has little prior knowledge and the scientific community shares conflicting opinions [51,52]. Furthermore, we controlled for specific prior knowledge concerning invasive species and the Asian ladybug, as well as prior perceived risk concerning invasive species (no prior perceived risk measures were collected for the Asian ladybug, as the issue was new to most of our participants). Specifically, we tested the following hypotheses:

First, we anticipated framing effects on perceived risk and emotions. More specifically, we assumed that perceived risk would be higher for all variants of the human-interest frame compared to the neutral-text condition (perceived-risk hypothesis). In addition, negative emotions would be stronger for all variants of the human-interest frame compared to the neutral-text condition (emotions hypothesis). Finally, we assumed that framing effects on emotions are mediated by perceived risk (mediation-by-risk hypothesis).

Second, we expected framing effects on learning outcomes. More specifically, we assumed that participants in the three human-interest frame conditions would differ in the amount of knowledge they acquire compared to those in the neutral-text condition (amount-of-learning hypothesis). Additionally, participants in the three human-interest frame conditions would prioritize negative information and disregard positive information (prioritization hypothesis). Finally, we assumed that framing effects on learning and prioritization are mediated by the emotional state (mediation-by-emotion hypotheses).

To answer the question whether effects are "generalizable" to different variants of the human-interest frame or whether it is specifically one feature that mainly exerts respective effects, we also tested for differences between the human-interest frame conditions with respect to the perceived risk and emotions hypotheses, and the amount-of-learning and prioritization hypotheses. However, we did not necessarily expect such differences.

Ethics statement

All participants volunteered and provided verbal informed consent. The study was conducted in accordance with the German Psychological Society (DGPs) ethical guidelines (2004, CIII) as well as APA ethical standards. According to the German Psychological Society´s ethical commission, approval from an institutional research board only needs to be obtained, if funding is subject to ethical approval by an Institutional Review Board. This research was reviewed and approved by the Ministry of Science, Research, and the Arts of Baden - Württemberg, Germany (grant number 33-720.281/34), which did not require additional Institutional Review Board approval. All data was collected and analyzed anonymously.

Participants and design

Participants were recruited at the main campus of a German university. Ninety-one students of law (51.6%), medical science (22%), social studies (20.9%), and economics (5.5%; age: M = 20.88 years, SD = 2.97; 30.8% male and 69.2% female) participated in our study. All participants received €10 as compensation. The participants were randomly assigned to one of the four conditions (emotionalizing, dramatizing, or individualizing human-interest frame and neutral text) in a pretest-posttest design.

Materials

Socio-demographic data and the assessment of perception of risk and emotions

The socio-demographic data questionnaire contained items on gender, age, education, and activities in environmental organizations. Perceived risk and emotions were assessed on 5-point scales. Perceived risk of invasive species (termed as “recently immigrated animals and plants”) was assessed before and after reading the text. Emotions and the perceived risk related to the Asian ladybug were assessed immediately after reading the text.

The questionnaires on perception of risk consisted of seven items based on a questionnaire developed by McDaniels and colleagues [54] and tested by Lazo and colleagues [55]. It included one general item on perceived risk and two items each on McDaniel and colleagues’ identified factors “impact on species” and “impact on humans” as well as one item each on the factors “controllability” and “knowledge of impacts”. A rating of “1” indicated low risk and a rating of “5” indicated high risk for all items. Correlations among all seven items revealed that only five (namely the general risk item and the four items on impact on species and humans) intercorrelated substantially. This finding is in accordance with results of factor analyses by Lazo and colleagues [55]. Therefore, we used a combined measure of the five highly intercorrelated items for our analyses and termed it “perceived risk” (Cronbach’s α = .74 for perceived risk of the Asian ladybug and Cronbach’s α = .81 for the perceived risk of invasive species in general). The items concerning perceived risk of invasive species in general were additionally used to measure perception of risk prior to reading the texts (“prior perceived risk”, Cronbach’s α = .76).

The questionnaire on emotions consisted of one item addressing general affectedness and four items addressing the specific emotions of anger, sadness, guilt, and fear (1 = not at all affected/angry/sad/guilty/anxious to 5 = very affected/angry/sad/guilty/anxious). The items, rated immediately after reading the text, were based on those used in the study by Nerb and Spada [56]. The reliability of the combined measure of all items was sufficient (Cronbach´s α = .70). We therefore used the combined measure “negative emotions” for our analyses.

Assessment of prior knowledge and learning outcomes

To assess prior knowledge, we asked participants to answer two open questions and six multiple-choice items. The open questions presented prior to reading the text were designed to assess prior knowledge of invasive species in general and the Asian ladybug in particular. Thus, we first asked the participants whether they knew of any invasive species in the region and second, if they had ever heard of the Asian ladybug; if yes, they should tell us what they knew about it. The answers to these open questions were rated on a 5-point scale (0 = no prior knowledge to 4 = very elaborate prior knowledge). The answers of 32 (i.e., 35.16%) of the participants were rated independently by two raters. Interrater agreement was very good, at ICC = .98 (intraclass correlation coefficient). In the rare cases of divergence, the answers were re-examined together to arrive at a final rating. Given the high interrater agreement, we had just one rater rate the remaining questions. Total prior knowledge was calculated as the sum of points earned in the two open questions (maximum of 8 points for very elaborate answers in both questions) and the six multiple-choice items (maximum of 6 points for all multiple-choice items).

After participants had read the text, and completed the questionnaires on emotions and perceived risk, we asked them to answer one open question and six multiple-choice items, which were equivalent to the multiple-choice items in the pretest. The open question posed after reading the text was designed to assess the type of information the participants prioritized [57]. We therefore asked the participants the following question: “Which do you consider the most important information about the Asian ladybug - which would you, for example, tell a friend?” To rate the answers to the open questions, we first segmented the answers into units of analysis [58]. These units were assigned to the categories for positive (e.g., “fights pests”), negative (e.g., “invades vineyards”), and neutral (e.g., “is originally from Asia”) statements and for statements referring to information about the Asian ladybug in the text’s body or about invasive species in general in the opening and concluding paragraphs 2 raters coded the answers to the open questions from 32 participants (i.e., 35.16%) independently. Interrater agreement was very good, κ = .87 (Cohen’s Kappa; [59]). In the rare cases of divergence, the answers were re-examined together to arrive at a final consensus. Again, given the high interrater agreement, only one rater rated the remaining questions. Units referring to the text’s opening or concluding paragraph (i.e., differently-framed components) were excluded from further analysis to capture “pure” framing effects (in contrast to content effects). Only units referring to information about the Asian ladybug (i.e., information that was the same across all conditions) were analyzed for prioritization effects.

Of the six multiple-choice items posed after reading the text, the first asked what invasive species are, that is, it referred to information given in the differently-framed introduction of all four texts. The remaining five multiple-choice items referred to information provided in the body of the text, that is, about the Asian ladybug, which was the same across all conditions. An example of a multiple-choice item about the Asian ladybug is: “What does the Asian ladybug feed on? a) grapes, b) vine pests, c) plant louses, d) grape juice, e) none of the options is correct, or f) I don’t know” (translated from the German). The first item about invasive species served as a control for an appropriate reading of the introductory information. It was not included in the total posttest score on multiple-choice items, resulting in a maximum of five potential points on the multiple-choice posttest. In the pretest, only 7.7% of all participants answered this first item correctly, whereas 98.9% of all participants did so in the posttest. These results reveal that: (1) participants´ prior knowledge about invasive species was very low and (2) the participants had read the opening paragraph carefully, which also enhances the probability of framing effects.

Procedure

Data collection took place in a teaching room at the Department of Psychology at the University of Freiburg, Germany. Two trained supervisors assessed groups of up to 20 participants simultaneously. Data on all conditions were collected simultaneously at each session. Materials for each condition were distributed evenly across the seats in the room, that is, each seat was assigned to one specific condition before participants arrived. Participants then selected their own seats when entering the room, which can be considered as a method of random assignment to conditions. Participants were first informed that the study aim was to examine how they perceive a newspaper article concerning invasive species. They were then given an overview of the procedure. They first completed a package consisting of the socio-demographic data questionnaire, prior knowledge test, and the questionnaires on prior risk perception. Participants were then given the respective versions for their seats (i.e., conditions) of the newspaper article, and were asked to read it attentively, as if they had selected it from a newspaper based on vested interest. After reading the text, they were asked to rate their emotions and the perceived risk of invasive species in general and the Asian ladybug in particular. They were then asked to rate the perceived informational value (“How informative do you think the text about the Asian bug was?”; 5-point scale from 1 = not informative at all to 5 = very informative) and the text's credibility (“How credible do you think the text about the Asian ladybug was?”; 5-point scale from 1 = not credible at all to 5 = very credible). Additionally, we asked participants which criteria they considered when evaluating the informative value, as well as the credibility: content of the text, presentation of information, or personal opinion on the topic. Finally, participants answered the open question about which information in the text they considered most important and again completed the same multiple-choice test as in the pretest. At the end of each session, participants were debriefed and received their compensation for participation.

Pre-analyses

Framing effects on perceived risk and emotions

Mediation-by-risk hypothesis.

Finally, we tested for a mediation effect, that is, whether framing enhanced negative emotions via the increased perceived risk of the Asian ladybug. Results of ANCOVAs of the effect of framing on perceived risk already indicated that framing significantly affected perceived risk. Regression models as calculated in mediation analyses arrived at the same result (path a, Figure 1). The same can be said about the direct effect of framing on negative emotions (path c, Figure 1). Additionally, there was a significant effect of perceived risk on negative emotions controlling for the influence of framing (path b, Figure 1). Finally, bootstrapping analysis for indirect effects indicates that perceived risk can indeed be considered a mediator between framing and negative emotions, a*b = .46, z_ab = 4.06, p<.001, η² = 0.10, SE =.11, LCL = .276, HCL = .718.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Mediation model: Mediation-by-risk hypothesis.

**p =.004, ***p < .001.

https://doi.org/10.1371/journal.pone.0079696.g001

Taken together, our results on perceived risk and emotions indicate that all aspects usually found in human-interest frames increased both perceived risk and negative emotions more than a neutral version of the text. Furthermore, our results demonstrate that effects of the human-interest frame can be "generalized" to different variants of the frame, and that the increase in negative emotions was mediated by perceived risk.

Framing effects on learning outcomes

Limitations and directions for future research

The main aim of our study was to investigate the effects of the human-interest frame compared to a neutral-text version of the same information. Additionally, we tested the "generalizability" of the findings to common variants of the human-interest frame in the field. Our attempt to provide ecologically valid material as well as other features of our research design and sample, pose some limitations to our study.

First, we did not make an experimentally sharp distinction between the features used in human-interest frames. The lack of a sharp distinction resulted from our procedure to arrive at the respective frame versions based on real-life examples. While this approach seems desirable to ensure ecological validity, it led to methodological “fuzziness” with respect to the features commonly used in human-interest frames. Therefore, our material cannot distinguish between the framing effects of particular features of the human-interest frame. However, our aim was not to investigate distinct effects of the features used in human-interest frames. Rather, we attempted to investigate whether emphases in one or the other feature lead to significant differences, that is, if effects are attributable to mainly one feature of the frame or actually to the “sum of its parts”.

Second, one might argue that despite the explicit reference to the Asian ladybug in measures of emotions and perceived risk, participants may not have been able to really separate their impression of the differently framed parts of the texts and the common core text. Also, the perceived risk of invasive species was likely affected by the core text about the Asian ladybug. This might be especially true as the Asian ladybug was portrayed as an example of an invasive species. However, we still argue that compared to most studies in the field of framing, our attempt enabled us to measure relatively “pure” farming effects.

Third, the relatively small and homogenous sample size of ninety-one university students makes generalization to a larger population difficult. There might be, for example, variance restrictions in our variables due to this sample as well as in moderator variables such as intelligence or socioeconomic status influencing our results. Also, a larger and more heterogeneous sample would allow for more in-depth analyses of the constructs measured in our study using factor analyses as well as SEM analyses for a more detailed investigation of relations between perception of risk, emotions, and learning outcomes. Therefore, future studies should include a more representative and larger sample to test for framing effects.

Finally, it would be worthwhile to employ more elaborate learning measures to better understand framing effects on learning. One possibility would be to use a posttest developed based on item response theory to be able to differentiate better between difficult and easier test items. Despite these limitations, we are confident that our findings provide valuable insights to explain framing effects on information processing and learning combining theories of media research with cognitive and educational psychology.