Introduction

The post-truth era and influence of emotion in social information

Emotions, particularly negative emotions and counterarguing (Lang et al., 1988; Nabi and Myrick, 2018; Lillie et al., 2021), are constantly used in persuasive communication and have been equated with information (Kunda, 1987, 1990; Schwarz, 2010), which influences audiences’ attitudes toward news reports at different cognitive stages.

The post-truth era is characterized by valuing emotions over facts, bias over objectivity, and attitudes over truth (Higgins, 2016; Zhang, 2017; Rochlin, 2017; Gewin et al., 2017), suggesting that emotional information has a greater influence than objective facts on personal attitudes, networked subjectivity, and public opinion (Keyes, 2004; Harsin, 2015; Midgley, 2016; Boler and Davis, 2018), thereby enhancing existing dogmatic beliefs (Ashton and Craft, 2021), leading to increasing fractionation of the online public sphere (Lewandowsky et al., 2017; Thorson, 2016), and a precipitous decline in media credibility (Keyes, 2004; Roberts, 2010; Bounegru, 2017).

In the post-truth communication environment, emotion is not only personal information but also social information (Van Kleef, 2009; Liu and Fu, 2022), with audiences influenced by external emotional input; information processing relies on the comprehensiveness and depth of processing of emotional claims. On the social level, emotional expressions represent and reproduce social and structural bias (Zembylas, 2020) and have a visibly determinant role in the media influence of individuals (Boler and Davis, 2018). Recent neurocognitive studies have also revealed the influence of effect on complex cognitive tasks such as truth discernment (Van Bavel and Pereira, 2018; Paul et al., 2020).

As mentioned above, affective responses and cognitive processes interact, and to a large extent, the processing of emotional information occurs before and influences factual information in the post-truth communication environment. While a wide range of studies has utilized the term “post-truth era”, the precise nature of this era, including its advent and formation, is still a matter of debate, and the mechanism influencing audience cognitive processing has yet to be explored by communication researchers. A few studies have suggested that these two kinds of information exhibit opposite patterns in terms of their time course and neurocognitive basis of influences on audiences’ attitudes. To probe this question, the study utilized the implicit association test (IAT) and event-related potentials (ERPs) to elucidate the characteristics of the post-truth era and refine the related theories.

Measurement of ERPs and audience implicit attitudes

Audience attitude is a critical concept in media psychology that generally includes three components: affect, behavior, and cognition. In cognitive psychology, the IAT is commonly used to measure the audience’s implicit attitudes, which are unconscious and accumulated from one’s past experiences and present feelings; these attitudes, to some extent, can affect one’s predispositions, emotional tendencies, and behaviors toward social matters (Greenwald and Banaji, 1995; Luo et al., 2006; Xu and Zhang, 2009, 2011).

Event-related potentials (ERPs), such as the N2 and N400 (negative peaks that occur 200–350 ms and approximately 400 ms after stimulus onset) and the late positive potential (LPP), are observed in electroencephalography (EEG) data, which have a high temporal resolution; these ERPs represent an ideal cortical activity measurement for investigating cognitive processing stages (Martin-Loeches et al., 2001; Hinojosa et al., 2001; Hauk et al., 2006). Previous research has revealed that emotion can enhance attention and medial frontal negativity changes in association with emotion-influenced decision-making (Huang and Luo, 2004). However, there is a paucity of research on the speed of negative bias development and the stages of information processing that these biases manifest (Smith et al., 2003).

To this end, the present study applied cognitive neuroscience research methods to analyze the implicit attitudes of audiences and influencing factors; specifically, we collected EPR data while participants read emotionally charged articles, which are prevalent in the post-truth era. This study addressed the research question of whether there are differences in audience information-processing mechanisms between reading neutral news articles and news articles with emotional content. If so, then this difference should be reflected in brain activity during the IAT (Friese et al., 2011). We hypothesize that in the post-truth era, audience attentional resources are depleted by the emotional content in news reports that affect implicit attitudes; if so, the N2 and N400 amplitudes will be larger. Furthermore, the collection of ERP data during the IAT was used to disentangle the two stages of emotional information processing. The aim of this study was to explore audience cognitive conflict monitoring and evaluation of emotional information to develop a two-stage model and examine the mechanism by which emotion (as social information) influences the audience in the post-truth communication environment.

Methods

Participants

The experiment recruited 44 undergraduate and graduate students from local universities and randomly divided them into two groups: an experimental group (24 participants, 12 males, Mage = 22.13 ± 1.94 years) and a control group (20 participants, 9 males, Mage = 23.20 ± 2.19 years). All the participants were right-handed, had normal or corrected-to-normal vision, and majored in fields not related to psychology. No participants had a history of neurological or psychiatric illness or took medication that would affect their nervous system. The participants completed the Chinese version of the Beck Anxiety Inventory (BAI), the Beck Depression Inventory (BDI), and the Positive and Negative Affect Scale (PANAS) (as shown in Table 1); there was not a significant difference in positive [t (42) = 1.314, p = 0.196] or negative affect [t (42) = 1.224, p = 0.228] between the two groups before the experiment. All participants signed informed consent before the experiments and were paid after participation. The experimental procedures were approved by the Institutional Review Board of the School of Journalism and Communication, Beijing Normal University and were performed in compliance with the principles of the Declaration of Helsinki.

Table 1 Age and BAI, BDI and PANAS scores of the experimental and control groups.
Full size table

Materials

A variety of articles on popular media topics were presented to participants in a pilot test, including one on genetically modified foods (GMFs). This topic was eventually selected, as scientific debate can trigger emotions, and personal feelings come into play when discussing controversial scientific issues (Scheufele and Krause, 2019; Neil et al., 2018; Gewin et al., 2017). The news topic selected was GMFs, which have long been at the center of the “post-science” (Iyengar and Massey, 2019; George, 2013) controversy in China and worldwide and is an example case in which truth and facts are obscured by controversy and strong emotional reactions.

In the present study, the two groups were asked to read two stimuli materials: a news article with neutral content and a news article with emotional content. The neutral news report was an unbiased article introducing the national compulsory labeling of GMFs, which contained 1369 Chinese characters, while the emotional news article was consistent with the argument in the neutral article but modified to contain more emotional expressions, with a total of 1343 characters. All the materials were printed on A4 paper, with a standard size of 210 mm × 297 mm.

Procedure

The whole experiment lasted approximately 10 min. Before the experiment, the participants were given 3 min to relax (resting state). In the experiment, they were instructed to read the news article in detail and complete a questionnaire related to its content as a manipulation check. The reading time was approximately 4 min. After the experiment, participants were given another 3-min break to return to baseline emotional status.

IAT

After reading the materials, the participants completed the IAT, which was divided into five phases (Huijding et al., 2005). First, attribute words with positive and negative meanings were presented on the screen, and participants pressed keyboard buttons in response. For positive words, they pressed “F”, and for negative words, they pressed “J”. Second, there was a practice phase in which target words and attribute words were both presented on the screen. For positive attribute words, the participants pressed “F”, and for negative attribute words and target words, they pressed “J”. Third, in the sorting phase, the target words and attribute words were presented again, and the participants sorted them as in the second phase. Fourth, in the reverse practice phase, the target words and attribute words were both presented, but (in contrast to the second and third phases), the participants were instructed to press “F” for target words and positive attribute words and to press “J” for negative attribute words. Fifth, in the reverse sorting phase, the target words and attribute words were presented again, and the participants sorted them according to the rules in the fourth phase. The third and fifth phases were the formal experimental phases, with all types of words appearing at random for a total of 60 times, while in the practice phases, all types of words appeared at random for a total of 20 times.

To offset the influence of the word order, the order of phases with the first rule (second and third phases) and the second rule (fourth and fifth phases) was counterbalanced among the participants (Huijding et al., 2005; Xu and Zhang, 2009). In the IAT, the differences in reaction times (RTs) to the target words and the ERP data in the third and fifth phases were taken to indicate the participants’ attitude bias toward the stimuli. For instance, RTs are shorter for congruent (strongly associated) words (hereafter, positive association) than for incongruent (not strongly associated) words (hereafter, negative association), which can reveal participants’ unconscious attitudes or preferences.

In the experiment, 10 target words and 20 attribute words (with equal numbers of positive and negative words) were presented. The valence of the attribute words was recorded on a 7-point Likert scale (ranging from 1=negative to 7=positive). Twenty participants were asked to evaluate the material. Positive attribute words and negative attribute words differed significantly in valence [t (38) = 18.36, p < 0.001]. The font used was Microsoft YaHei, and the font size was 34. Words were shown in the center of the computer screen, with a size of 13 inches, horizontal angle of 3.8° and vertical angle of 1.0°, and screen resolution of 1920 × 1280.

The IAT was administered with E-prime 3.0 (Psychology software tools for solutions for research, assessment and education). In each trial, a white fixation cross on a black screen was shown for 500 ms, and then the words were presented at a duration of 1000 ms each. During the experiment, the participants sat quietly in the laboratory, staring at the center of the screen (approximately 60 cm away from their eyes), and were instructed to move as little as possible and to respond via key presses as quickly and accurately as possible when words were presented.

Recordings and analysis

RTs to target words in the third phase (negative association) and the fifth phase (positive association) were recorded for the two groups of participants, and statistical analysis was then carried out using SPSS 22.0. RTs less than 100 ms or more than 1000 ms were removed as outlier values.

The EEG data were recorded using a wireless Cognionics Quick-30 32-channel amplifier and sampled at 1000 Hz with a 0–100 Hz bandpass. The reference electrodes were the left mastoid during recording. EEG Lab 14.1.1 was used to conduct an independent component analysis (ICA), and the averaged EEG signals were low-pass filtered at 30 Hz and segregated every 2 s. The ERP amplitudes at the Fz, Cz, and Pz electrodes were analyzed using SPSS 22.0.

Results

Reaction time

The RT to target words after positive words and negative words was analyzed by repeated-measures analysis of variance (ANOVA) in a 2 (group: emotional vs. neutral) × 2 (association: positive vs. negative) design; group was the between-subject factor, and association was the within-subject variable. The results showed a significant main effect of association on RTs [F (1, 42) = 80.467, p < 0.001, η2P = 0.646]. RTs for target words presented after positive words (MPA = 576.61 ± 66.27) was shorter than those for target words presented after negative words (MNA 637.16 ± 58.81). The main effect of the group was not significant [F (1, 42) = 0.200, p = 0.657, η2P = 0.005], with no significant difference in RTs between the two groups. There was no significant effect of a group × association interaction [F (1, 42) = 1.741, p = 0.194, η2P = 0.038]. These results (shown in Table 2) revealed that RTs for positive associations (congruent condition) were shorter than those for negative associations (incongruent condition) in both the emotional and neutral groups; this indicated that the participants all had a positive implicit attitude toward GMF issues.

Table 2 Reaction times on the IAT of the experimental and control groups (ms) (M ± SD).
Full size table

ERP data

The average amplitudes of the ERP data from all participants are shown in Table 3, and the ERP waveforms and brain topographic maps are shown in Fig. 1.

Table 3 Average ERP amplitudes of participants (M ± SD).
Full size table
Fig. 1: N2, N400 and LPP components of the experimental group and the control group as well as topographic maps.
figure 1

IE incompatible condition of the experimental group, CE compatible condition of the experimental group, IC incompatible condition of the control group, CC compatible condition of the control group.

Full size image

The ERPs measured in this study were as follows: the N2 (a negative peak that occurs at 200–350 ms after stimulus onset in the frontocentral region; indicates cognitive processing, such as discrimination, attention and target selection), the N400 (a negative peak occurring approximately 400 ms after stimulus onset in the right parietal and temporal regions; often associated with higher cognitive processes, such as language cognition and cognitive conflict) and the LPP (a late positive peak occurring approximately 300 ms after stimulus onset; often associated with higher cognition, such as continuous processing of attention and emotional conflict and subsequent effects).

The average amplitude of the N2 was analyzed by repeated-measures ANOVA with a 2 (association: positive vs. negative) × 2 (group: experimental vs. control) × 3 (electrode site: Fz vs. Cz vs. Pz) design. The results revealed a significant main effect of association on N2 amplitudes [F(1, 42) = 4.244, p = 0.046, η2P = 0.092], with N2 amplitudes significantly larger for target words with negative associations (the incongruent condition) than those with positive associations (the congruent condition). The main effects of group [F (1, 42) = 0.371, p = 0.546, η2P = 0.009] and electrode position [F (2, 84) = 3.262, p = 0.054, η2P = 0.072] were not significant. The effects of the association by electrode interaction [F (2, 84) = 0.153, p = 0.830, η2P = 0.004], the association by group interaction [F (1, 42) = 2.780, p = 0.103, η2P = 0.062], and the electrode by group interaction on N2 amplitudes [F (2, 84) = 1.463, p = 0.239, η2P = 0.034] were not significant. There was no significant effect of the three-way interaction of association, group, and electrode on N2 amplitudes [F (2, 84) = 0.339, p = 0.684, η2P = 0.008].

The average N400 amplitude was analyzed by repeated-measures ANOVA with a 2 (association: positive vs. negative) × 2 (group: experimental vs. control) × 3 (electrode position: Fz vs. Cz vs. Pz) design. The results revealed that the main effect of association on N400 amplitudes was significant [F (1, 42) = 7.490, p = 0.009, η2P = 0.151], with significantly larger N400 amplitudes for target words with negative associations (incongruent) than target words with positive associations (congruent). The main effect of group on N400 amplitudes was also significant [F (1, 42) = 13.846, p = 0.001, η2P = 0.248], with N400 amplitudes significantly smaller for the emotional-information group than the neutral-information group. The main effect of electrode position on N400 amplitudes was not significant [F (2, 84) = 0.052, p = 0.888, η2P = 0.001]. There were no significant effects of the association by electrode interaction [F (2, 84) = 0.221, p = 0.778, η2P = 0.005], the association by group interaction [F (1, 42) = 1.168, p = 0.286, η2P = 0.027], or the electrode by group interaction on N400 amplitudes [F (2, 84) = 2.874, p = 0.083, η2P = 0.064]. There was no significant effect of the three-way interaction of association, group, and electrode position [F (2, 84) = 0.534, p = 0.569, η2P = 0.013].

The average amplitude of the LPP was analyzed by repeated-measures ANOVA with a 2 (association: positive vs. negative) × 2 (group: experimental vs. control) × 3 (electrode position: Fz vs. Cz vs. Pz) design. The results revealed that the main effect of association on LPP amplitudes was significant [F (1, 42) = 5.565, p = 0.023, η2P = 0.117], with significantly larger LPP amplitudes for target words with positive associations (congruent) than for those with negative associations (incongruent). The main effect of group on LPP amplitudes was also significant [F (1, 42) = 4.438 p = 0.041, η2P = 0.096], with significantly smaller LPP amplitudes in the emotional-information group than in the neutral-information group. The main effect of electrode position on LPP amplitudes was not significant [F (2, 84) = 0.853, p = 0.401, η2P = 0.020]. There were no significant effects of the association by electrode interaction [F (2, 84) = 0.028, p = 0.970, η2P = 0.001], association by group interaction [F (1, 42) = 0.816, p = 0.371, η2P = 0.019], or electrode by group interaction on LPP amplitudes [F (2, 84) = 1.211, p = 0.294, η2P = 0.028]. There was no significant effect of the three-way interaction of association, group, and electrode position [F (2, 84) = 0.097, p = 0.904, η2P = 0.002].

Conclusions

This study used IAT and ERP data to elucidate the mechanism by which emotional information influences the implicit attitudes of individuals. The main findings were as follows. First, regarding the behavioral results, all participants had positive implicit attitudes toward the selected news topic (i.e., GMFs). Second, the ERP data were influenced by emotional information. Third, compared with the neutral news article, the emotional news article elicited an earlier N2, and the amplitudes of the N400 and the LPP were reduced. These findings suggest that news articles with emotional content may engage in different information-processing mechanisms than those with neutral content.

The early N2 and reduced N400 amplitudes reveal conflict monitoring in the emotional-information group

On the one hand, the amplitudes of the N2 and N400 were significantly larger for target words with positive associations (congruent) than those for target words with negative associations (incongruent), which is consistent with previous studies (De Houwer et al., 2006; Hilgard et al., 2015; Xu and Zhang, 2009, 2011). These results showed that the incongruent condition triggered stronger cognitive conflicts than the congruent condition. Previous studies have shown that the N400 reflects conflict and integration during semantic information processing and that the amplitudes of the N400 are positively related to task difficulty (Coch and Holcomb, 2010).

Hence, if individuals had positive implicit attitudes toward GMFs, they would experience greater cognitive conflict when processing negative words. However, after participants read the emotional information, this conflict was weakened, which made it impossible for them to effectively process and integrate the core factual information contained in the news articles.

This is primarily because emotional media discourse can prioritize the effects of emotional arousal, leading to a greater allocation of attentional resources to information distinction and categorization. In the incongruent condition, the increased N2 amplitude indicated decreased conflict monitoring for implicit attitudes. Furthermore, the N400 amplitude decreased with decreases in conflict in advanced semantic processes. Previous studies have found that N2 or N400 amplitudes are associated with the incongruent condition and that P300 amplitudes are associated with the congruent condition (Coates and Campbell, 2010; Hilgard et al., 2015; Xu and Zhang, 2009, 2011)—the performance of the neutral-information group in this study is congruent with those results.

At 300–450 ms, the N400 amplitude in the incongruent condition was larger than that in the congruent condition, which suggests that there was greater cognitive conflict due to semantic associations inconsistent with participants’ implicit attitudes. However, the emotional-information group had an obviously larger N2 amplitude in the early 200–300 ms period. Although the average amplitude of the N2 in the neutral-information group was not significant, the semantic-associated N400 had a longer duration. The different N2 responses in these two groups indicate a difference between conflict monitoring and information classification and discrimination (Donkers and Boxtel, 2004; Xiao et al., 2015) when reading emotional information, which further affected the amplitudes of the LPP.

Decreased LPP amplitudes reveal depleted attentional resources and reduced conflict evaluation in the emotional-information group

The LPP component represents later evaluation of the target tasks (Hajcak et al., 2010; Hajcak and Nieuwenhuis, 2006; Parvaz et al., 2014). Previous studies have mostly found that after 500 ms, the amplitude of this component is larger under the congruent condition (in which stimuli are consistent with implicit attitudes) (Xu and Zhang, 2009, 2011; Coates and Campbell, 2010; He et al., 2009). In this study, the results of the neutral-information group were basically consistent with these earlier findings.

However, in the emotional-information group, the LPP amplitude decreased significantly. In this field, cognitive conflict (and subsequent judgment and decision-making) is thought to be reflected by negative ERPs; therefore, the LPP appears in the later stage, at which time information processing slows. Late positive components indicate continuous processing of certain stimuli, which is closely related to the allocation of attentional resources, arousal by external stimuli, and participative motivation (Lang et al., 1998; Ding et al., 2007). Compared with neutral stimuli, emotional stimuli are more engaging for participants, resulting in greater motivation to participate in reading tasks and thus larger LPP amplitudes (Brown et al., 2012; Dunning and Hajcak, 2010).

We speculate that this may be related to the fact that the early conflict monitoring and semantic processing (the N2 and N400) mentioned above consumed more cognitive resources, leading to an inability to effectively distinguish target stimuli in the late evaluation stage. In the early stage of conflict monitoring, emotional content consumed more attentional resources, which reduced the later evaluation of the target tasks. This means that in post-truth communication, individuals devote very little attention to the late-stage assessment of emotional information compared with objective and neutral information. Information processing and implicit attitudes were affected to a considerable extent; when cognitive conflict was weakened, negative associations were used as the basis of judgment and decision-making, and individuals did not fully evaluate the core factual information.

Discussion and corollaries

The nonevidential elements of post-truth media have long been considered controversial in the public sphere, as facts and evidence are neglected (Ashton and Craft, 2021); in contrast, social media platforms allow the public to make a claim and voice their feelings as information (Schwarz, 2010), encounter an echo-chamber of similar opinions, and be influenced by “emotion as social information” (Van Kleef, 2009; Van Kleef et al., 2015). Some scholars have pointed out that the post-truth era does not involve the denial of truth but the implication of “plural epistemological compasses” (Fuller, 2022). Therefore, the mechanism by which emotional information in media influences attitudes must be re-evaluated.

This study used neurocognitive experiments to assess media influences on audiences’ attitudes, which provided more precise evidence-based inference than previous self-report methods (Liu and Fu, 2022). The importance of these findings lies in the requirement of “developing multidisciplinary and meta-cognitive skills to help people distinguish between beliefs and facts” (Glăveanu, 2017) and providing solid academic evidence for the notion of post-truth communication as an “alternative epistemology not conforming to conventional standards of evidentiary support” (Lewandowsky et al., 2017).

In summary, the methodological value of this research was its innovative application of the IAT paradigm and ERP data in the communication field to explore emotional priming and attitude changes, which are especially relevant in the post-truth era. The pattern of influence on audience attitudes and behaviors can be summarized as a cognitive conflict monitoring and evaluation (CCME) model, which divides information processing into two stages. First, in the early conflict-monitoring stage, emotional information was prioritized due to emotional arousal, and more attentional resources were allocated to message distinction and semantic processing, reflected by stronger N2 amplitudes, an earlier N400, and weaker N400 amplitudes. Second, in the late evaluation stage, the depletion of attentional resources led to reduced late-stage evaluation of the target stimuli by the audience, reflected by reduced LPP amplitudes in the emotional-information group.

Thus, the results and conclusions of this study preliminarily support our hypothesis that in the post-truth communication environment, audiences allocate excessive attentional resources to the cognitive conflict stage due to emotional discourse, which reduces the late-stage evaluation of core factual information and alters implicit attitudes. More importantly, it may indicate that individuals are more likely to agree with (but not actually think deeply about) information in the news or rely on heuristics or biases, thereby preventing audiences from expanding beyond their bubble of existing beliefs (Wood and Porter, 2016; Kunda, 1990; Taber and Lodge, 2006). In addition, belief echoes theory (Thorson, 2016) suggests that once the negative image or biased beliefs are formed and crystalized, they continuously influence the audience even though fact-checking or the truth is provided immediately afterward. The practical implications of these findings are to promote awareness of public media literacy and cultivate individual initiative and critical thinking (Zembylas, 2020) to discern factual information in the post-truth era.

Finally, there are some limitations to this study that merit further exploration. For instance, emotion and objectivity in the post-truth communication environment could be affected by framing and confirmation bias and the reward pathway, which involves brain regions such as the mPFC and NAcc (Kappes et al., 2020). In addition to text information, video-based emotional framing in the post-truth era also shapes our truth discernment (Duncombe, 2019). Thus, further fMRI or functional near-infrared spectroscopy (fNIRS) experiments with more precise spatial resolution and multimodal news information should be considered in the future.