Introduction

What factors contribute to the rejection of well-founded messages about environmental issues and climate change by certain social groups? What strategies might be employed to encourage these groups to adopt more environmentally friendly attitudes and behaviours through more effective communication? Despite extensive research on environmental communication, these remain some of the most pressing yet largely unanswered questions in the field. This article argues that a significant challenge lies in the tendency of many studies in environmental communication to adopt a predominantly linear approach when examining the media’s role in shaping public representations.

In response, it presents an alternative approach that emphasizes the social and cultural cognitive dynamics influencing how diverse individuals interpret and respond to mediated environmental information. It examines the formation of individual mental representations about environmental issues, their transformation into cultural representations, and their intersubjective sharing with other members of a social group. Additionally, it demonstrates how the interplay between these cultural representations and media messages generates images of reality that are conducive to different types of environmental behaviour.

The research proposed is influenced by connectionist neuropsychology, cognitivist sociology, anthropology and cultural sociology. It draws on Bourdieu’s concept of practical reason to examine the possibility of combining different aspects of cognitive theories that highlight idiosyncratic thought processes with social theories aimed at understanding stable, collective socio-cultural representations. The operationalisation of this approach may be described as a cognitivist-inspired form of media reception analysis (see Livingstone and Das, 2013 for background). A combination of qualitative and quantitative methods was employed to conduct this analysis in which subsets of participants (n = 1440) responded to three different self-constructed news messages and a minimal information condition. The messages pertain to a genuine pilot project concerning the utilisation of treated municipal wastewater for the irrigation of agricultural land in northern Belgium.

The objective of this research was twofold. The first objective was to determine whether and to what extent cognitive and affective associations formed prior to exposure to the media messages influenced participants’ perceptions and interpretations of these messages; additionally, by experimentally manipulating appeals to fear and disgust, the research allowed for proposing general recommendations on the effective use of threat in environmental communication.

Secondly, the study aimed to explore the extent to which these cognitive and affective associations are idiosyncratic or socio-cultural.

Theory and literature

This section presents a three-pronged analysis. The initial section offers a critique of environmental communication approaches that draw upon media effects and framing theories. The subsequent section outlines the potential contributions of connectionism to an alternative understanding of the problem. The third section introduces elements of a connectionist-inspired sociology of media reception. The final section proposes an operationalisation of the theoretical concepts presented based on the concept of resonance.

The effects problem

The classical transmission model of media effects is certainly not as prevalent as it was in the early days of communication studies. Nevertheless, linear thinking about media effects still largely prevails. This is particularly evident in environmental communication, where three groups of theories dominate the literature.

The first body of literature comprises information deficit theories. The information deficit model posits that public inaction or misunderstanding of environmental issues is primarily due to a lack of information. It suggests that increased information and education will raise public awareness and promote behavioural change if information is carefully vetted and meticulously processed to fill knowledge gaps and guard against misinformation (Diamond, Bernauer, and Mayer, 2020; Hart, 2013; Yao et al., 2024). Although these theories are gradually being discarded (Priest, 2022; Suldovsky, 2017), they remain prevalent, either explicitly or implicitly, in studies of climate change, the environment, sustainability, and science communication (Grant, 2023; Lindenfeld et al., 2012). This is also evident in research on public support for water reuse (Barnes, Krishen, and Hu, 2021). Information deficit theory has been widely criticised for neglecting factors such as beliefs, values, attitudes and emotional dispositions that significantly influence cognition and behaviour (Lindenfeld et al., 2012; Nisbet and Scheufele, 2009).

Furthermore, the analytical model of information deficit theories does not help to understand the social dimension of knowledge construction and, in particular, the interaction involved in any form of communication (Harris and Sanborn, 2014). It thus remains essentially a linear approach, proposing that a discrete external cause, namely information, acts directly on the individual, leading to the accumulation of knowledge and subsequent behavioural change, without considering contextual or cognitive factors (Leiserowitz, 2006; Schultz and Zelezny, 2003; van den Broek, Bolderdijk, and Steg, 2017; Whitmarsh, 2015).

The second body of literature builds on congruence or consistency theories (such as cognitive dissonance theory) and the related idea of congruent persuasion in strategic communication. This group of psychological theories addresses the issue of effective environmental messaging by emphasising that the mobilising effect of communication efforts depends largely on the consistency between the message’s appeal and the recipient’s pre-existing beliefs, values and attitudes (Leiserowitz, 2006; Schultz and Zelezny, 2003; van den Broek, Bolderdijk and Steg, 2017; Whitmarsh, 2015). It is also generally accepted that these factors cannot be separated from the different social contexts to which recipients belong (Kahan et al., 2012; Norton and Grecu, 2015; Whitmarsh, 2015).

The core analytical perspective of congruence theories thus revolves around the notion of ‘differential’ effects (Kuhn 2000). Media effects can vary in strength according to the attitudes, value orientations and dispositions of recipients, depending on how strongly they are ‘moderated’ or ‘mediated’ by social and dispositional factors (Nilsson et al., 2016). However, the extent to which media messages are perceived and interpreted ‘differently’, i.e. with significantly different meanings for different recipients, remains largely unexplored. Thus, while congruence theories place media effects in a broader context, their analytical model also remains largely linear. The message is seen as a distinct external cause that can be moderated or mediated by external factors, but still directly affects cognition and influences public action.

The third body of literature, which has grown considerably in the field of environmental communication, is based on framing theory. While framing theory is operationalised in different ways, the basic idea is that through exposure to media frames, or ‘interpretative stories’ (Nisbet and Scheufele, 2009), which consistently and repeatedly select, emphasise and articulate information about social issues, people become familiarised with frame arguments (Lecheler and De Vreese, 2019).

Framing theory generally pays attention to both the socio-cultural context in which frames emerge and persist, and the cognitive factors that influence the reception of media frames as mediating factors. These include the availability of belief content (individuals’ existing knowledge and beliefs), the accessibility of belief content (the ease with which beliefs or ideas can be recalled), and the salience of belief content (the importance individuals attach to particular beliefs or ideas) (Lecheler and De Vreese, 2012; 2019; Schuck and Feinholdt, 2015). The analytical model of framing research is thus quite similar to that of congruence theories. As discrete causes, frames have a linear effect on recipients’ views, which may be differentially mediated by external social and cognitive factors. In a number of studies of effective environmental communication, this starting point is translated into even more rudimentary experimental research designs that view media messages as the cause and cognition as a self-evident mediating variable or effect (Dasandi et al., 2022; Greenaway and Fielding, 2020). Again, the role of prior mental representations in constructing different meanings from the same message is not addressed.

Cultural schemas and connectionism

An alternative approach to understanding how cognition affects our understanding of frames is proposed by D. Scheufele and Tewksbury (2007), who state that the term framing refers to ‘modes of presentation that communicators use to present information in a way that resonates with the underlying schemas of their audiences’ (own emphasis). Schemas are then defined as ‘templates of reality’ that people routinely use to decode media frames (B. Scheufele, 2004).

Schema theory thus helps to explain, at least in part, how existing representations retroactively influence interpretations of reality, including the media messages that represent that reality. As pre-existing ‘templates’ of reality, schemas are also thought to allow people to ‘fill in the blanks’ and make sense of new experiences even when they have only partial information about them. However, for the same reason, schemas are also thought to distort perception, either by filtering out information or by projecting schematic information onto reality (Cerulo, 2010). In this way, the seeds of a conceptualisation emerge in which the effects of frames are not ‘differential’, but in which a message can be understood in substantially different ways.

Schemas are thought to operate at the level of ‘heuristic’ or ‘automatic’ thinking, i.e., a rapid and largely unconscious way of thinking applicable to most media consumption (Boutyline and Soter, 2021; Shrum, 2009). The application of schema theory to communication, nevertheless, presents both theoretical and operational challenges (Nisbet and Newman, 2015). One of the primary challenges associated with the schema concept, as elucidated by Rumelhart et al. (1986b), is that schemas are not ‘things’ (p. 20). Indeed, neuropsychological research has not identified any specific neurological functions that can be attributed to the existence of schemas (Ghosh and Gilboa, 2014). The observation of schemas is contingent upon the examination of their consequences; that is to say, the tangible forms in which individual mental representations are communicated into ‘public representations’ (Sperber, 1996).

A second issue is that schema theory does not elucidate how schemas can be at the same time sufficiently robust and sufficiently adaptive to be applicable across different contexts while simultaneously remaining flexible and dynamic enough to accommodate new information and provide new meanings (Rumelhart et al., 1986b). It is therefore necessary to consider how to develop an alternative concept to schema that is sufficiently detailed to capture the observed regularities in responses to reality or media content, while at the same time being sufficiently flexible to account for the mind’s plasticity when encountering media content that addresses new situations, events, and solutions, such as wastewater reuse.

Connectionist neuropsychology addresses this issue by providing a conceptualisation of cognition as a network of units of information connected by weighted links stored in memory. Upon activation of the network by an external input, such as a media message, specific information units are triggered. These units will interact and exert influence over one another, contingent on their degree of activation (Rumelhart et al., 1986b). It is this pattern of connectivity that determines the system’s knowledge and determines how it will respond to any arbitrary inputs (Rumelhart et al., 1986a). These patterns of activation thus constitute a mental representation of the reality that precedes and conditions the way people perceive and respond to inputs from reality, including media messages as illustrated in Fig. 1.

Fig. 1
figure 1

Individual representation cycle.

Full size image

Although connectionism is clearly relevant to communication theory, one of its shortcomings is a notable absence of consideration by its founding fathers of the role of emotions, particularly considering the increasing recognition of the importance of affect in cognitive processes (Pessoa, 2008; 2010). However, Suri and Gross (2022) have demonstrated how emotions can be integrated as ‘nodes’ within connectionist models. Earlier studies also indicated how connectionist models could simulate cognitive dissonance by ascribing positive or negative emotional valence to the links in an activation pattern (e.g., Van Overwalle & Siebler, 2005). As the social science literature on climate change increasingly recognises the importance of emotions in raising awareness and garnering public support (Leiserowitz, 2006; Lockwood, 2016), further integration of emotion as a component of cognition will also enhance the applicability of connectionist theory to this field of study.

Social representations

At the time of writing, only one study employing connectionist modelling techniques was identified that provided a psychological account of the reasons why individuals perceive media messages differently (Ewoldsen, Hoewe, and Grady, 2022). In view of the present study’s focus on the shared perceptions of media representations and their role in influencing environmental engagement across diverse social groups, an alternative sociological approach is proposed. This approach seeks to address the question of why groups of people interpret media messages (and reality) differently.

In essence, connectionism may be regarded as a theory of representation. Mental representations may be defined as the internalisation of inputs from the external reality to which they refer. Learning is an indispensable factor in the formation of these representations. In connectionist terms, learning is defined as changes in the weights between information links that occur as different inputs are processed. Consequently, the process of building on an existing knowledge structure is generally a relatively straightforward and rapid one, whereas the integration of new information that is inconsistent with the existing structure is a slower and less impactful process, resulting in minimal changes to the existing connections (McClelland, 2013).

Given that learning is not an individual but rather a social process that is implicated in structures of social stratification and differentiation, it is possible to outline the tenets of a connectionist-inspired sociology of communication effects. The foundations of this approach can be situated in cognitivist anthropology and sociology, where a ‘cultural’ understanding of connectionist representations was first introduced through the lens of schema theory. In sociology, cultural schemas are defined as ‘socially shared representations deployable in automatic cognition’ (Boutyline and Soter, 2021). In the field of anthropology, Strauss and Quin (1997) have presented a connectionist analysis of the concept of cultural schemas, aligning it with the aforementioned connectionist critique of schema theory. They therefore propose that cultural representations are constructs that ‘vary in their schematicity, depending on the strength and density of the interconnections among the units of which they are composed’ (p. 52).

In sociology, Vaisey (2009) argues that a connectionist approach to schemata aligns with Bourdieu’s (1990, p. 53) concept of habitus, which refers to ‘durable and transposable dispositions’ that ‘generate and organise practices and representations’. Bourdieu (1984) suggests that habitus acts as both a ‘structuring structure’ and a ‘structured structure’ (p. 170). As a ‘structuring structure’, habitus shapes individuals’ social practices and their understanding of others’ practices. As a ‘structured structure’, it is influenced by the internalisation of socially differentiated representations and practices of others.

In sum, while connectionism provides a neuropsychological basis for the sociological theory of social representations, anthropology and sociology provide a contextualisation that explains how cognitive patterns are shaped and differentiated by social-cultural practices. Applied to communication, this means that the perception and interpretation of media messages are always conditioned by preceding, socially shaped mental representations or cognitive structures, as illustrated in Fig. 2. It follows logically that different (media) representations that inspire different practices can coexist in society, shaped by the patterns of social stratification and differentiation that they help to maintain, change or reinforce.

Fig. 2
figure 2

Social representation cycle.

Full size image

Operationalisation: resonance analysis

It has been argued that the continued prevalence of linear transmission thinking in environmental communication represents a significant challenge. To address this issue, Hansen (2018) suggests that communication studies should adopt a distinct ‘vocabulary’ that encompasses terms such as ‘dialectic, interaction, reinforcement, engagement, information loops, neural networks, multidirectional, resonance and parallel forms of meaning-making.’ (p. 173). These concepts give rise to a particularly challenging and perhaps incommensurable research agenda. Any empirical research is necessarily reductionist, and the complex realities of media, communication, and social interaction must be cast into simpler models to be researchable. This may account for the dearth of studies that have succeeded in empirically capturing the intrinsic social and cultural dimensions of meaning-making as a recursive cognitive process.

In light of the aforementioned limitations, this article puts forth an operational framework that regards communication as a socially situated process, whereby meaning is collectively constructed through the interplay between the information conveyed in a message and the pre-existing representations stored in memory. In alignment with connectionist theories, these representations can be operationalised as weighted connections between information units, whereas cultural representation can be operationalised as socially shared connections between information units. If a given media input is repeatedly associated with the same content categories by a statistically significant number of research participants, it may be posited that these categories are indicative of a socially shared pattern of connections.

Given our interest in people’s automatic, intuitive and effortless reasoning, i.e., their unconscious mental representations, we employ thought elicitation as a means of inferring people’s associations with our research topic. This technique involves the prompting of participants to express their thoughts, feelings and associations on topics or messages spontaneously and rapidly, i.e., how these topics or messages resonate with people. The thought-listing technique is based on the premise that the psychological significance of an individual’s thoughts and feelings, as well as the underlying cognitive processes, can be examined by analysing the individual’s reported concepts (Cacioppo et al., 1997).

Semantic coding is employed to regroup the shared concepts reported by participants into coherent semantic categories. This process constitutes the qualitative core of data processing, as it interprets and provides meaning to the raw data by capturing underlying patterns and nuances in the participants’ responses. It is important to execute this step with great care in order to guarantee that the derived categories accurately reflect the participants’ intentions, experiences and perspectives. Any misinterpretation or overgeneralisation at this stage has the potential to distort the meaning of the data and lead to conclusions that fail to accurately represent the essence of the participants’ input.

Subsequently, the connections between the shared concepts are quantified using multiple correspondence analysis (MCA). MCA is a statistical technique that is appropriate for the analysis of categorical data, such as elicited thoughts and feelings. MCA is a so-called geographical method, which entails the visualisation of relationships between variables in a low-dimensional space (Le Roux and Rouanet, 2010).

The proximity of points in this space represents the average strength of relationships between categories, based on the frequency with which they are mentioned together. We call these visual clusters of related concepts resonance schemes. Resonance schemes are thus the patterns of verbalised association produced by a group of study participants in response to an input, in this case, an article or some basic information.

The dimensions of the MCA plot show the lines along which these categories and schemes can be situated. These dimensions suggest the existence of some underlying factors, such as social values or dispositions. The further a group of concepts is from the origin axis representing the dimensions, the more it has contributed to the construction of that dimension. The dimensions are often antagonistic, meaning that the poles of the dimension indicate social tension. Consequently, the further apart the resonance schemes are, the more likely they are to be antagonistic.

A subsequent step involved in MCA is the inclusion of supplementary variables, such as emotions, behaviours, socio-demographics, and attitudes in the MCA plot. These secondary or passive variables do not influence the construction of the MCA plot but provide a broader framework for understanding how the observed conceptual patterns intersect with the social and personal characteristics of participants.

Methods

Resonance analysis can be used to analyse the processing of ideas, situations, people, objects, etc. In this study, a resonance analysis was set up to analyse the reception of differently framed newspaper-like articles on wastewater reuse in agriculture. The following sections explain the design of the reception analysis and sampling, the construction of the accompanying survey, the data collection and processing procedures, and how MCA was used to analyse the data. The survey questionnaire, articles and datasets are available in the Supplemental Materials.

The objective of the study was to investigate how people respond cognitively and affectively to environmental messages. Participants were confronted with differently framed newspaper-like articles about a field experiment using treated municipal wastewater for groundwater recharge through subirrigation of farmland. At the time of data collection, the project was still ongoing and had only received marginal media attention. In April 2021, there was one local radio clip and one newspaper article, and in September 2022, there was one local radio clip. It can thus be reasonably assumed that the attitudes of the participants had not been influenced by media coverage prior to the commencement of the experiment. Unlike in other countries, wastewater reuse is not a public controversy in Belgium, minimising potential political bias and polarisation of opinions (Duong and Saphores, 2015).

The reception analysis consisted of four conditions designed to explore how participants interpreted media messages about wastewater reuse in agriculture.

Condition 1: Minimal Information—This condition provided participants with only three factual sentences offering basic information about the field experiment. The minimal approach aimed to assess how participants relied on their prior knowledge, beliefs or biases without any influence from media priming or framing.

Condition 2: Advocacy—In this condition, participants were presented with a newspaper-like media message that favoured wastewater reuse. It expanded on the minimal information from Condition 1 by providing arguments in support of wastewater reuse, while deliberately avoiding any framing through fear or urgency.

Condition 3: Contamination Frame—This condition added a preemptive warning to the article from Condition 2, cautioning against misinformation about the risks of wastewater reuse. It explicitly highlighted concerns about contamination, using evocative language like “dark water” to frame the issue negatively.

Condition 4: Drought Frame—In this condition, the preemptive warning from Condition 3 was replaced by a paragraph discussing the consequences of recurring droughts. Wastewater reuse was positioned as a solution to address water scarcity, rising water costs, and the increasing impacts of drought.

Participants (n = 360 per condition) read one of these four articles and subsequently completed a thought elicitation test. They were prompted with the open-ended question: “What thoughts come to mind when you think about eating vegetables grown with treated wastewater? Write what comes to mind in the boxes below. Your answers can be a single word, a phrase, or a sentence.” To maintain the quality of responses and ensure that they reflected participants’ spontaneous associations, no responses were forced. Additionally, participants were not informed about the topic of the articles prior to the test.

All verbatim responses were coded in vivo using MAXQDA software and then re-coded into broader semantic categories as shown in Table 1. A triple coding procedure was used to ensure stability. Disagreements were resolved through discussion and consensus. Whenever one of the three coders expressed doubt about the assignment of a concept to a category, the concept was not included.

Table 1 Overview of categories and concepts mentioned by participants (n per sub-sample).
Full size table

Two variables were incorporated into an accompanying survey to conduct a manipulation test. Given that the second newspaper-like article in our experiment refers to contamination risks, the Fear of Contamination subscales of the Padua Inventory on Contamination Obsessions and Washing Compulsions (Burns et al., 1996) were deemed the most appropriate for this study. The selected Padua scales comprise items pertaining to both fear and disgust and has been demonstrated to be a highly robust indicator of acceptance of agricultural wastewater reuse (Verhoest et al., 2022). A Categorical Principal Component Analysis (CATPCA) on the data collected in this survey revealed a single component with a Cronbach’s alpha coefficient of 0.90, an eigenvalue of 5.09 and a total variance explained of 56.6% (see Table 3, Appendix A).

To provide a counterweight to the Padua scales, and given our interest in science and environmental communication, a scale on trust in science was incorporated. This scale was developed based on previous work by Hartman et al. (2017) and Pechar, Bernauer, and Mayer (2018). A six-item scale was constructed, comprising three items pertaining to science and three items pertaining to scientists. It was anticipated that the scale would yield two components, based on previous testing conducted in the post-Covid period, when scientific experts were highly visible in the media and often generated controversy. However, only two items pertaining to trust in scientists exhibited differential loading. Accordingly, a single-component scale was employed. The Cronbach’s alpha coefficient for this scale was 0.88, with an eigenvalue of 3.73 and a total variance explained of 62.1% (see Table 4, Appendix A).

Additional parameters

Given our interest in the relationship between people’s representations and their practices and emotions, our questionnaire included four additional scale-based questions. To accurately capture the underlying structures and fully utilise the available data, we also processed all these multi-item scales using CATPCA with Oblimin Rotation. Missing values were treated as active in the analysis (i.e., they were included in the computations as an additional category).

The current water conservation behaviour of the participants was evaluated through a behavioural measure. Previous studies on the subject have employed a variety of scales, yet they were either based on an insufficient number of items to be deemed reliable, required familiarity with information that is not necessarily common knowledge, or consisted of ambiguous prescriptive statements (Aprile and Fiorillo, 2017; Corral-Verdugo, Fraijo-Sing, and Pinheiro, 2006; Dascher, Kang, and Hustvedt, 2014). As an alternative, a seven-point scale (ranging from always to never) was employed to assess an eight-item list based on water-saving behaviours that are relatively commonly known and promoted on public websites. The CATPCA revealed a single dimension with a Cronbach’s alpha of 0.80, an eigenvalue of 3.37, and a total variance explained of 42.13 (see Table 2 and Appendix A). Intended consumption of vegetables grown with treated wastewater was assessed by two complementary questions. The first question aimed to assess likely (passive) acceptance. The second was aimed at likely (active) resistance. Using a 5-point Likert scale, participants were asked to agree or disagree with two statements: ‘I am willing to eat vegetables grown with wastewater’ and ‘I will avoid eating vegetables grown with wastewater’ (see Table 1a, b, Appendix A).

Data analysis

As previously stated, MCA was employed to analyse the responses to the thought elicitation test. The application of MCA with Burt-adjusted inertias enables the mapping and interpretation of the verbalised associations of participants in a two-dimensional space. Fourteen categories were selected for this analysis. These are categories that were referenced by a minimum of 8% of respondents in at least one of the four conditions. The rationale for this restriction is that the number of participants who collectively reference the same concepts rapidly diminishes after a specific tipping point. This point was identified to be approximately 8% of participants across all four conditions. Given our primary focus on group representation and the objective of maintaining a clear and concise analysis, only concepts mentioned by more than 8% of participants in any of the four conditions were retained.

One disadvantage of MCA is its inability to provide information regarding the statistical significance of relationships between categories. To address this limitation, Fisher’s Exact Test was employed to ascertain whether any two categories exhibited a significant association (see Appendix C). The combination of Fisher’s Exact Test with adjusted standardised residuals offers a significant advantage in terms of statistical validation of the relationships and antagonisms identified in the MCA. This test allows us to determine not only whether the associations were statistically significant, but also whether they were positive or negative. Positive associations indicate that concepts were often mentioned together, while negative associations indicate that they were rarely mentioned together.

Results

This section describes and interprets the data as they become visible in the four MCA plots, one per test condition. The primary variables are shown in black. These are the concepts mentioned by the participants. The supplementary variables are shown in colour. The socio-demographic variables of age, gender and education are in blue. Intended behaviours (‘Willingness’ to accept and ‘Avoidance’ of treated wastewater) and current water consumption behaviour (‘Conservation’) are in purple. Fear of contamination (‘Fear’), trust in science (‘Trust’) and water-saving behaviour (‘Water’) are in lavender. The original MCA coordinates, including the missing values, which are left out here for layout purposes, are available in the Supplemental Material.

Condition 1: Minimal information

The MCA plot for the minimal information condition, shown in Fig. 3a, provides a two-dimensional representation. The vertical axis corresponds to Dimension 1, accounting for 58.8% of the inertia of the conceptual categories. The horizontal axis represents Dimension 2, which explains 11.5% of the inertia and is considered significant as it exceeds the Kaiser threshold of 7.14 (calculated as 100% divided by 14 concepts). Together, these dimensions account for a cumulative inertia of 70.3%, indicating a robust explanatory result.

Fig. 3: MCA of condition 1: Minimal information.
figure 3

a MCA of condition 1 categories (n = 360). b MCA of condition 1 supplementary variables (n = 360).

Full size image

As a two-dimensional MCA plot, this model is comprised of four quadrants. The bottom-right quadrant contains three categories. Those who referenced ’pollutants’ and expressed ‘health concerns’ also reported feelings of ‘repulsion’. Considering the pivotal role that fear plays in the associations held by this group of individuals, we may refer to this cluster of concepts as the phobic resonance scheme.

The secondary scores of the fear of contamination and trust in science variables on the MCA plot indicate that, on average, participants in this quadrant exhibit the highest levels of fear of contamination and the lowest levels of trust in science. The other supplementary variables indicate that participants in this quadrant are the least willing to utilise wastewater reuse and report the greatest efforts to avoid its use. Their reported current water-saving behaviour is, on average, the lowest of all participants. Participants in this quadrant additionally exhibit lower levels of trust in science and water-saving behaviour, while exhibiting higher levels of fear of contamination (Fig. 3b).

The upper right quadrant, in turn, is comprised of three categories. Participants expressed ‘doubts’ about the quality of the water or crops but stated that they would have ‘no problem’ consuming them if adequate ‘control’ measures were implemented. The acceptance of wastewater reuse by these participants may be characterised as conditional. It can therefore be proposed that this conceptual system be termed the ambivalent resonance scheme.

Turning to the left side of the model, the upper left quadrant comprises three primary categories. Participants in this quadrant expressed ‘appreciation’ about the proposed solution, with several describing it as ‘innovative’. Two more categories are distinctly positioned at the apex of this quadrant. These categories pertain to participants who express a sense of ‘urgency’ and further justify their stance by asserting that the quality of treated wastewater is essentially ‘equivalent’ to other water used for irrigation. All these categories appear to be unified by pragmatism and realism. The conceptual system collectively demonstrated by the participants in this quadrant can therefore be characterised as a utilitarian pragmatist scheme.

On average, these participants score highest on the trust in science and lowest on fear of contamination variables. These attitudes are mirrored in their intended behaviours, as evidenced by high scores for intended acceptance and low scores for active avoidance. They also exhibit the highest scores in actual water-saving behaviours.

Finally, a coherent set of categories unified by ‘pro-environmental’ statements is observed in the bottom left quadrant. These categories are supported by references to the role of water reuse for irrigation in enhancing ‘sustainability’, mentions of ‘water conservation’, and the classification of wastewater reuse as ‘ecological’. Collectively, these categories present a clear depiction of what can be characterised as an ecological resonance scheme.

With regard to socio-demographic variables, gender does not serve as a differentiating factor, and age appears to be relevant only for the second dimension. There is a moderate correlation with education. Participants who display increased avoidance behaviour tend to have a lower educational profile, whereas those who make ecological and utilitarian associations generally have a higher educational background. To avoid repetition, it can be noted here that these relationships remain stable across all test conditions.

Condition 2: Advocacy

As previously stated, the advocacy article (Fig. 4a) comprised the complete version of the self-constructed news message, omitting any references to fear or drought. The total inertia of the analysis is thus 74.1%. However, in contrast to the two-dimensional MCA plot that was previously presented, the current plot is essentially one-dimensional. Dimension 1 accounts for 68.8% of the inertia between the categorised concepts, which is approximately equal to the sum of the two significant dimensions of the minimal information MCA. Dimension 2, with a value of 5.3%, is below the Kaiser criterion of 7.14%. Therefore, it is advisable to exercise caution when interpreting the vertical displacement of the variables. The significance and direction of the associations between the concepts were assessed using through the Fischer Exact Test (see Appendix C).

Fig. 4: MCA condition 2: Advocacy.
figure 4

a MCA of condition 2 categories (n = 360). b MCA condition 2 supplementary variables (n = 360).

Full size image

This MCA plot being one-dimensional is a noteworthy finding in and of itself, warranting further investigation. The previously observed association between the ‘doubt’ and ‘control’ categories is no longer evident. In contrast, the concept of ‘doubts’ is now found to be closely associated with the negative concepts of ‘health concerns’, ‘pollutants’ and ‘repulsion’. Conversely, the ‘control’ category is now closely linked with the positive concepts of ‘innovation’, ‘ecology’, ‘sustainability’ and others. The pragmatist utilitarian scheme displayed in the top left of the plot has thus become considerably more conditional in comparison with the previous one. Respondents who were previously characterised as pragmatist utilitarians seem to have become more ambivalent.

On the right-hand side of the model, the distance between ‘pollutants’, ‘health concerns’ and ‘repulsion’ has decreased considerably, indicating a strengthening of the relationships between these concepts. Except for the bottom left quadrant, which still comprises the most strongly ecologically motivated respondents (for further details, the results of the Fisher’s Exact Test, Table B in Appendix C), all other quadrants appear to be affected by the addition of the fear appeal. The plotting of the supplementary variables is largely consistent with these findings. The impact of gender and age is still negligible.

The discrepancy in resonance between the minimal information condition and the advocacy message is an unexpected result. This effect is most likely due to the second paragraph of the article (see Supplemental Material). This paragraph lists several pollutants and describes the chemical tests and checks carried out by scientists to monitor water quality. This suggests that, despite the message’s intention to reassure readers, the mention of pollutants is sufficient to elicit a fear response. This conclusion is supported by the subsequent manipulation of the contamination condition (Fig. 4b).

Condition 3: Contamination frame

Figure 5a shows the MCA of the news story framed around the fear of contamination. The core text is identical to that of the advocacy article. However, an introductory paragraph was added that strongly implied the risk of contamination and used vocabulary that evoked disgust. Similar to the advocacy article, the MCA of this message is one-dimensional. Dimension 1 accounts for 57.6% of the inertia. Dimension 2 accounts for 5.6%. The total inertia of this model is 62,2%, which is slightly over 10% less power than the control article. The second dimension falls below the Kaiser criterion of 7.14%. Dimension 2 should therefore be interpreted with great caution. Reference to Fisher’s Exact Test (see Appendix C) will be made to compensate for this.

Fig. 5: MCA of condition 3: Contamination frame.
figure 5

a MCA of condition 3 categories (n = 360). b MCA of condition 3 supplementary variables (n = 360).

Full size image

A discernible pattern emerges when the MCA results of the advocacy article are subjected to comparison. A further decrease in the distance between the concepts of ‘doubts’, ‘pollutants’, ‘repulsion’ and ‘health concerns’ is observed. The results of the Fisher test conducted on these variables (see Appendix C) corroborate this finding, demonstrating a notable strengthening of the associations due to the evocation of disgust and fear. The organisation of categories and the positioning of the control and other supplementary variables remain largely unchanged compared to the control article and consistent with the findings (Fig. 5b).

Condition 4: Drought frame

The first dimension of this MCA (Fig. 6a) accounts for 65.8% of the inertia between the terms mentioned, while the second accounts for 6.3%. The total inertia of this MCA is therefore 72.1%. The second dimension is just below the Kaiser criterion of 7.14. This means that the inertia explained by the second dimension should be interpreted with caution. However, Fisher’s exact test (Appendix C) provides some additional information that supports a two-dimensional interpretation.

Fig. 6: MCA of condition 4: Drought frame.
figure 6

a MCA of condition 4 categories (n = 360). b MCA of condition 4 supplementary variables (n = 360).

Full size image

For memory, the drought frame article retains the identical text used in the advocacy article but begins with a drought warning. Notwithstanding minor differences, both articles elicit similar responses. Indeed, Fig. 6a closely resembles the MCA plot of the minimal information message shown in Fig. 3a. In particular, the relationship between participants expressing ‘doubt’ and those expressing ‘control’ has been restored. As detailed in the following section, this restored relationship is statistically significant, suggesting that the inclusion of the drought argument as a motivating factor has indeed positively influenced some participants’ associations.

Fisher’s Exact Test

Fisher’s Exact Tests were used to assess both the direction and significance of the relationships among the concepts. This test not only validates the results of the MCA but also enhances the accuracy of our findings’ description and interpretation. It highlights two overarching patterns consistent across all four MCA plots. First, relationships among clustered categories within the same quadrant are generally positively significant, meaning these concepts were frequently mentioned together. Second, relationships between concepts located at opposite ends of the main dimension are generally negatively significant, indicating these concepts were rarely co-occurring.

However, there are two notable exceptions to the second pattern. Specifically, in the prior condition, the concepts ‘control’ and ‘doubts’, which are situated within the same quadrant, exhibit a positive significant association, as expected. This relationship shifts when fear and disgust-inducing information are introduced as input stimuli.

In the minimal information condition, ‘doubts’, central to the ambivalent resonance scheme (top-right quadrant), demonstrates a positive association (p < 0.01) with ‘health concerns’, a core concept in the phobic resonance scheme (bottom-right quadrant). Additionally, ‘doubts’ also shows a positive association (p < 0.05) with ‘equivalence’, a key concept in the opposing utilitarian pragmatist resonance scheme (top-left quadrant). However, in the two conditions where fear and disgust are most clearly triggered, the positive association between ‘doubt’ and ‘equivalence’ becomes insignificant, reflecting a shift in cognitive connections.

Similarly, the concept ‘control’ shows a significant positive association with ‘sustainability’ in the utilitarian pragmatist resonance scheme (p < 0.01) and a negative association with ‘repulsion’ in the phobic resonance scheme (p < 0.05). Nevertheless, when fear or disgust are prevalent, associations with both ‘control’ and ‘equivalence’ become insignificant, while the positive association with ‘health concerns’ remains significant in both the advocacy article and contamination frame condition (p < 0.05).

Discussion

After receiving minimal information about a field trial of wastewater reuse in agriculture, but prior to exposure to newspaper-like media messages, four distinct resonance schemes emerged from participants’ responses: a phobic and ambivalent scheme dominated by fear and disgust; and a utilitarian pragmatist and ecological scheme characterised by a relative trust in science. Taken together, these resonance schemes comprised 14 conceptual categories. All 14 categories mentioned by participants also appeared in subsequent manipulations in which participants were exposed to fully elaborated and differently framed articles on the same topic. Even the relative frequency of concepts mentioned across conditions remained relatively stable.

These findings support our proposition that the interpretation of media messages is constructed on the basis of associations made prior to exposure to the message. Thus, while acknowledging that individuals consuming the same media content may draw different conclusions about it, it is equally important to recognise substantial similarities and differences in these representations at the group level. In our resonance analysis, the main difference observed was between participants categorised as ‘phobic’ and ‘ambivalent’ and those categorised as ‘utilitarian pragmatist’ and ‘ecological’. This finding suggests that there is a primary distinction between participants driven by fear and disgust and those motivated by trust and social motives.

An explanation for this finding can be found in Schwartz’s theory of self-enhancement and self-transcendence values (Lee et al., 2019; Schwartz, 2010). According to this theory, individuals who prioritise self-enhancement values tend to experience higher levels of personal anxiety, often manifesting in concerns about their personal health and safety. In contrast, individuals who prioritise self-transcendence values are less likely to experience the same level of personal anxiety. Schwartz further proposes that personal anxiety reduces a person’s capacity to engage in pro-social behaviour, whereas those with self-transcendent values have a greater capacity for such behaviour by managing their anxiety and focusing on benefiting others.

A second tension, between ‘utilitarian pragmatists’ and ‘ambivalent’ participants versus ‘ecological’ and ‘phobic’ participants, could also be identified. This tension can be explained in classical Weberian terms. The pragmatic and ambivalent group can be said to rely on ‘utilitarian rationality’, whereas the phobic and ecological groups rely on ‘affective or value rationality’ (Weber, 1978, pp. 24–25). This finding further emphasises that different social groups are unlikely to be equally receptive to the same arguments. As the advocacy condition unexpectedly revealed, even conventional scientific arguments may fail to persuade certain audiences if they are predisposed to react negatively to fear and disgust. This result was unintended, but is not surprising in hindsight, given the powerful and mutually reinforcing nature of emotions such as fear of contamination and disgust (Tracy and Randles, 2011; Verwoerd et al., 2013). Certain arguments, especially those based on fear, may cognitively backfire with some groups, triggering resistance and fostering oppositional attitudes (O’Neill and Nicholson-Cole, 2009).

In sum, our findings are consistent with the theoretical perspective of the connectionist sociology of communication, which postulates that the interpretation of media messages depends on pre-existing representations stored in memory. These stored representations actively influence new perceptions in real time. This process is influenced by cognitive and affective structures that have been formed as a result of past experiences and that are shared by people with similar backgrounds. While these cognitive and affective structures can be modified by new experiences, their ‘selective power’ imposes limits on the extent of such modifications (Bourdieu, 1990, p. 60). Because these structures are largely shaped by past experiences, they tend to maintain their consistency and resist the incorporation of information that is inconsistent with the accumulated experiences of the past.

This theoretical perspective has implications for the field of environmental communication. Most people have only recently become aware of environmental degradation and climate change. This means that the majority of people continue to perceive environmental issues and their consequences through the prism of social perspectives that predate the popularisation of the issue. Furthermore, adapting to climate change poses challenges to established lifestyles, consumption and mobility patterns. The process of learning about and adapting to climate change can therefore be expected to require significant cognitive and affective effort. It is therefore not surprising that individuals tend to resist environmental messages that challenge their long-held beliefs and perspectives.

In this context, the present study contains some general recommendations for environmental communicators. Fear appeals are a common feature of environmental communication, although the emotional intensity of such appeals is not always as high as in the framing conditions presented in this study. Some studies have shown that the use of threat or fear, when used in moderation and in an appropriate format, can serve as an effective motivator in the context of environmental communication (Compton et al., 2021; Van Der Linden et al., 2017). However, as our experiment shows, pointing out risks can also prove counterproductive for certain groups, potentially generating resistance to the proposed arguments and fostering oppositional attitudes.

The impact of threat can vary depending on the issue and the nature of the threat itself. Consequently, there can be no standard format for formulating effective environmental messages. Our experiment suggests that environmental communicators must strive to construct messages that resonate with people’s preconceptions while avoiding defensiveness, which can lead to disengagement or resistance and even reinforce oppositional views. Rather than directly challenging deeply held beliefs, communicators should cultivate empathy to gain an understanding of the audience’s worldview and emotional landscape. This will enable them to identify cognitive anchors—pre-existing ideas, values or beliefs that can be linked to new information.

Conclusion

This study departed from a critique of linear models of environmental communication, exploring the influence of prior representations, operationalised as patterns of information stored in memory. The analysis presented was based on a self-developed method of ‘resonance analysis’, which combines both interpretive techniques and statistical methods. Further testing and refinement of the methodology may enhance its reliability and validity. A key area for consideration is the integration of affective factors. In addition, the results obtained through thought elicitation could be validated through the use of alternative methods that are indicative of ‘automatic thinking’ and/or using complementary methods to assess possible effects on processes of deliberative thinking.

At the level of substance, the experiment demonstrated that the interpretation of media messages cannot be reduced to the content or frame to which recipients have been exposed, nor to their prior mental representation of the subject. Rather, it is through the interaction between the two that meaning is generated. Moreover, the study illustrated that the cognitive and affective predispositions of social groups exert a considerable influence on the interpretation of environmental media messages. It is evident that further generalisations and recommendations can be formulated only with the addition of further case studies on environmental issues that are urgent and require changing attitudes and behaviours, such as fast fashion, food and forms of transport with high carbon footprints, etc.

In order to facilitate a better integration with communication research, it may be necessary to further develop the theoretical apparatus. This could entail, for instance, the integration of elements drawn from cultivation theory, suggesting that radical shifts in attitudes or behaviours are not realised by a few manipulations in messages, but by long-term interaction between what media tell and how this resonates with what people know and believe. A deeper insight into the underlying social processes, such as those involved in social stratification and differentiation, will also contribute to a more comprehensive sociological understanding of the formation of social representations. Despite its limitations, this case study provides valuable recommendations for environmental communication.

The effectiveness of a message depends on many factors, including the issue at hand, the characteristics of the audience and the context in which the communication takes place. What resonates with one group may fail with another due to social, cultural, or psychological differences. Moreover, audience receptivity can vary significantly depending on the issue. For instance, while psychological factors like fear play a key role in wastewater reuse communication, other topics, such as fast fashion, high-carbon-footprint food or transport, may involve very different dynamics. Each context requires tailored research to identify psychological and social barriers to sustainable behaviour, as there is no one-size-fits-all approach.

Because there is no universal strategy—whether fear-based, hope-based or otherwise—that will work in all cases, environmental communicators should always carefully balance negative and positive messages. While threats can motivate action, appeals to fear must be used judiciously and coupled with empowering arguments. Although more research is needed on positive messaging, this study suggests that such arguments should highlight actionable steps and tangible benefits—especially those with personal relevance—rather than abstract environmental or societal benefits.

In addition, scientific consensus on the need for and benefits of environmental action can increase resistance, especially among ambivalent people who are sceptical about environmentalism. Those who are wary of environmental innovation or sustainable lifestyles often also distrust science, rendering purely scientific arguments ineffective or even counterproductive. Paradoxically, established scientific consensus can thus reinforce the scepticism of distrustful groups, deepening their uncertainty rather than resolving it.

In summary, environmental communication should move beyond the oversimplified fear-versus-hope dichotomy and adopt a more nuanced, audience-centred approach. By blending negative and positive arguments—and carefully tailoring content to the target audience—communicators can increase campaign effectiveness. Future research should further explore the interplay between these messaging strategies, particularly how positive arguments can overcome resistance and encourage proactive engagement on different issues.