The significance of causal attributions for mental illness

In societies with high mental health literacy, causal explanations are a common discussion-point in conversations about mental health [5]. As proposed by attribution theory [6], this may be due to a natural tendency to search for meaning behind unusual or challenging phenomena [7]. Causal attributions have been shown to aid people with mental health difficulties in navigating their own experiences with mental illness by redirecting blame away from themselves [8], and finding meaning in their experiences [9]. Conversely, attributions that foster self-blame can impede adjustment and coping processes [10].

Moreover, the causal attributions of the general public are key influences on attitudes to and treatment of people with mental illness. In recent decades, public campaigns aimed at reducing stigma or increasing mental health literacy have promoted a medical model of mental illness, often citing genetics or brain chemistry as responsible [2, 11, 12]. However, there is clear evidence to suggest that diverse social factors are also responsible for mental illness, including experiencing or being the subject of violence, discrimination, socioeconomic disadvantage, poor social cohesion, and more [13]. Research has suggested that members of the public underestimate the social determinants of physical illness [14], but it is unknown whether similar patterns exist for mental illness. While mental health literacy research has suggested the public often favour psychosocial attributions for mental illness [5], there is evidence that biogenetic attributions are becoming more commonplace [12, 15, 16]. Yet the longitudinal rise in biogenetic attributions has not been paralleled by observed improvements in stigmatising attitudes towards people experiencing mental health difficulties, and in the case of certain mental illnesses such as schizophrenia, stigma may be increasing [12, 17, 18].

Empirical research has shown that the medicalisation of mental illness can promote unfavourable attitudes towards people with mental health difficulties. Extensive meta-analytic evidence has found that while biological attributions for mental illness are associated with reduced blame towards people with mental illness, they are also associated with greater perceived dangerousness, greater desire for social distance, and greater prognostic pessimism than social attributions for mental illness [1, 2]. This may be explained by theories of psychological essentialism, which suggest that biogenetic explanations promote the lay belief that underlying “essences” are responsible for a group’s shared characteristics, which in turn triggers the aforementioned stigma attitudes [1]. Indeed, Dittrich et al. [19] report that essentialism fully mediates the relationship between biogenetic explanations of schizophrenia and increased social distance. The relationships between biological attributions and negative social outcomes have been observed in both cross-sectional [20] and experimental [21] research, with the latter demonstrating that causal beliefs can be manipulated by exposure to experimentally-presented causal explanations.

While extensive research has investigated how biological explanations may trigger essentialist ways of thinking, much less evidence illuminates the consequences of social attributions for mental illness. The majority of previous studies have used a brief generalised measure of “social explanations” as a control or comparative point for biological explanations [21]. Minimal research has investigated the potentially variable consequences of endorsing different types of social determinants of mental illness. Social psychologists studying other forms of intergroup relations have proposed the existence of a social component of essentialism, whereby certain social explanations may be associated with similar levels of essentialist thinking as biological explanations [22]. In particular, social attributions perceived as being highly deterministic (e.g., one’s upbringing or neighbourhood environment) may be associated with a greater degree of essentialist thinking than social attributions perceived as being more malleable (e.g., a recent traumatic experience, such as a car crash). For example, people with deterministic beliefs about one’s social origin show more prejudice against other nationalities [22]. It is unknown whether different social attributions for mental illness have distinct attitudinal consequences. Even more fundamentally, no research has catalogued the range and relative prevalence of social explanations for mental illness, which circulate in the public sphere.

Variations in causal attribution patterns

A nuanced understanding of lay causal attributions for mental illness requires attention to how attributional patterns can vary across context. Beyond a tendency to collapse potentially diverse social attributions into a single unidimensional category, much previous literature on lay attributions for mental illness is also limited by a tendency to conceptualise ‘mental illness’ as one universal category [23]. This masks any potential variation in attitudes across specific diagnoses [24]. There is some evidence that members of the public hold differing causal beliefs across mental disorder type. For instance, a review by Angermeyer and Dietrich [23] found that people are more likely to endorse biological factors as being responsible for schizophrenia than for depression. Moreover, the repercussions of certain attributions can differ across disorder categories: for example, a recent study of lay attitudes to functional (psychogenic non-epileptic/dissociative) seizures links biomedical explanations with less negative attitudes [25]. Observing differences in causal representations across different types of mental disorders can thus provide crucial nuance that would otherwise be lost in studying causal explanations of a generalised concept of mental illness alone.

As well as varying across mental disorder type, causal attributions are known to vary across time and culture [2, 5, 26]. For instance, in western cultures, as biological attributions have risen in popularity over recent decades, attributions to supernatural factors have dropped significantly [2, 12]. How aetiology is conceptualised is also contingent on historical context. An experimental study investigated how the COVID-19 pandemic reshaped public beliefs about the causes of mental disorders, finding that when people read about a specific (fictional) case of anxiety symptoms that began during the pandemic, generalised attributions to biological causes weakened [27] However, such experimental designs necessarily rely on contrived scenarios that require participants to respond to predefined scenarios in predefined ways. Full understanding of the malleability and contextual contingency of lay explanations for mental disorders requires an ability to track the causal attributions that spontaneously occur in everyday discourse, across particular points in time.

Media representations and their implications

An ecologically valid exploration of attributional patterns in public discourse, which can incorporate attention to time and context, necessitates access to naturally-occurring data. The mass media are a key site at which lay publics are exposed to aetiological explanations for mental disorders. While media content does not fully determine lay understandings, it nevertheless plays an important role in shaping public opinion [28]. This is particularly the case regarding more abstract or scientific topics, such as the aetiology of mental disorders, which are not perceptible to immediate observation or experience [29, 30]. Media data offer a naturalistic opportunity to establish the frequency with which causal attributions spontaneously appear, and the range of causes that are invoked, within everyday discourse about mental disorders.

The present study uses the Republic of Ireland as a case study to spearhead an original study of naturalistic attributions for mental disorders in public discourse. The history of mental illness in Ireland is characterised by stigmatisation and incarceration, gradually ceding to increasing mental health awareness and therapeutic support [31]. However, mental health provision in Ireland remains severely under-resourced, with long waiting-lists and minimal specialist services. An OECD report [32] found Ireland had the 3^rd highest prevalence of mental illness in Europe, with nearly one-in-five residents qualifying for a psychiatric diagnosis. Demand for mental health services, and the salience of mental illness as a public health issue, appear to have increased since the COVID-19 pandemic [33], making this a timely context to appraise how public understandings of mental disorders are evolving.

This study uses content analysis to identify the causal explanations for various mental disorders, and their prevalence, in major print and online news media in Ireland. The paper builds on previous literature by revealing the ways in which the public are exposed to causal explanations of mental disorders, paying particular attention to representations of social determinants, which are under-researched in previous literature. It also illuminates differences in news representations across mental disorders type by comparing across six categories of mental disorders. Finally, it investigates the malleability of attributions by exploring whether they shifted following the onset of the COVID-19 pandemic. In line with a ‘cultural meanings’ rather than ‘deficit model’ approach to lay understandings [34], the concern is not with evaluating the accuracy or ‘truth’ of public attributions, but documenting their range, prevalence, and distribution across time and mental disorder category.

Primary research question

• How are causal explanations for mental disorders represented in Irish news media?

Secondary research questions

• What proportion of articles that mention mental disorders contain causal explanations?
• Are causal explanations represented differently across different categories of mental disorders?
• Did patterns of causal explanations change during the COVID-19 pandemic?

Data collection

News articles were identified using the Nexis Advance database. Search terms were devised based on DSM-5 diagnostic terms and categorisations [35; S1 Appendix]. Six categories of mental disorders were chosen based on global prevalence rates [35, 36]: anxiety disorders, feeding and eating disorders, mood disorders (bipolar and related disorders and depressive disorders), schizophrenia spectrum and other psychotic disorders, personality disorders, and substance-related disorders. Searches were limited to Irish national news articles between 12^th March 2018 and 12^th March 2022.All national news sources available through Nexis Advance were included, encompassing 11 online and print news sources (S2 Table). Local newspapers were excluded. Articles identified from the search strings (n = 1,892) were downloaded to NVivo for content analysis [37].

Data analysis

Articles were categorised based on mental disorder category and whether they were published before or after the beginning of the COVID-19 pandemic. The 12^th March 2020 was chosen to denote the beginning of the pandemic, as on this date the Irish Taoiseach (Prime Minister) issued an official announcement formally acknowledging the COVID-19 crisis as a pandemic and introducing measures to tackle the virus.

All articles were read through to identify any statement indicating a causal explanation of a mental disorder, defined as any explicit or implied reasoning for the onset/progression of the mental disorders in question. Inductive content analysis was used to identify and classify different types of causal explanations. A preliminary coding frame was developed based on the analysis of 400 randomly-chosen articles covering all of the six mental disorder categories. Similar attributions were grouped into basic codes (e.g., social media/the internet), which were then organised into overarching superordinate code categories (e.g., cultural/societal environment; S1 Table). For context, each attribution was also coded to denote its source (i.e., the type of person/organisation who offered it).

Relevant codes were attached to each article using NVivo. Codes were not treated as mutually exclusive, and where relevant, explanations presented in the articles were assigned one or more codes. All articles were coded by reviewer one (L.H.), and approximately one-third of the coded articles (n = 178) were additionally independently coded by reviewer two (C.O.C.). Inter-coder agreement [38] was calculated using the coding comparison function in NVivo. Overall agreement was 92.8% (k = .874; S1 Table). The coding frame was then applied to the remaining articles and slightly modified as appropriate. A frequency analysis was conducted using the Matrix function in NVivo.

Ethics statement

Since all data were publicly available and written with the intent of public consumption, ethical approval was not required for this analysis.

Overall prevalence of causal explanations

Overall, 24.63% (n = 466) of the identified articles contained causal explanations for a mental disorder (Fig 1). Due to many of the articles containing multiple attributions, the total number of causal explanations came to a total of 628. Frequencies of articles containing causal explanations varied across type of mental disorder, from 8.70% for substance related disorders, to 29.50% for eating disorders.

In terms of the source of attributions, most explanations (51%) originated from an individual’s perception of their own experience with mental illness, whether directly relayed by themselves or someone else. Other sources included relevant experts, such as doctors, psychologists, and scientists (19%), scientific research (17%), and other origins (7%) such as political figures or commentators presented as knowledgeable on the topic.

Categories of causal attributions

Using content analysis [37], eight superordinate attributions for mental illness were identified, and their frequencies were calculated as a percentage of the total number of attributions overall (n = 628; Fig 2). These were: life events/experiences (20.86%, n = 131), the cultural/societal environment (16.40%, n = 103), interpersonal relations (16.24%, n = 102), health and lifestyle factors (13.22%, n = 83), biological factors (11.30%, n = 71), psychological factors (9.24%, n = 58), socio-economic conditions (8.12%, n = 51), and the family environment (4.62%, n = 29).

Comparison across mental disorders

An analysis of frequency was conducted which compared the prevalence of the eight identified causal explanations for each of the included mental disorder categories; psychotic disorders, personality disorders, mood disorders, eating disorders, and anxiety disorders (Fig 3). Results pertaining to the substance-related disorders category were omitted due to the low prevalence of articles containing causal explanations (n = 2), both of which described health/lifestyle attributions. While the value of this data is primarily descriptive due to challenges in meaningfully interpreting a 5×8 statistical comparison, chi square analysis did confirm a significant association between causal explanations and mental disorders category, χ²(28, N = 637) = 286.55, p < .001. Results show that life events/experiences were the most common explanation for anxiety and personality disorders, cultural/societal environment for eating disorders, and health/lifestyle factors for mood and psychotic disorders. Biological factors were most invoked for psychotic or mood disorders, and appeared infrequently in relation to anxiety, eating or personality disorders.

Frequency comparison before and during the COVID-19 pandemic

To assess whether there were observable differences in representations of causal explanations before and during the COVID-19 pandemic, a frequency analysis was conducted (Fig 4). Overall, there were slightly fewer causal explanations in news media during the pandemic (n = 289) than before (n = 339). A chi square analysis revealed that the distribution of causal explanations differed significantly before and during the pandemic, χ²(7, N = 628) = 37.77, p < .001. To pinpoint the specific categories responsible for this difference, eight 2×2 post-hoc analyses using a Bonferroni correction were conducted. This revealed a significant effect (p < .006) only within the interpersonal relations category, which rose from 8.62% of attributions before the pandemic, to 26% of attributions during the pandemic.

How are causal explanations for mental disorders represented in Irish news media?

Content analysis of over 400 news articles identified eight superordinate types of explanation for mental disorders in media discourse: life events/experiences, the cultural/societal environment, interpersonal relations, health/lifestyle factors, biological factors, psychological factors, socio-economic conditions, and the family environment. Overall, the most commonly occurring explanations involved attributions to social phenomena, such as life events/experiences, the cultural/societal environment, and interpersonal relations. Conversely, biological factors accounted for just over one-tenth of observed causal attributions. This arguably indicates the relative volume of interest in biological attributions in previous research may be disproportionate to their real-world frequency within naturally occurring public discourse.

As social attributions are much more frequent in real-world discourse about mental disorders than currently represented in attribution research, they merit more direct theoretical and empirical elaboration on their own terms, rather than as merely a foil or control for biological attributions. While most previous research has collapsed social explanations into a single unidimensional category, the current data illuminates the multi-layered nature of attributions to social causes. Articles invoked determinants that traversed multiple ‘levels’ of the social world: from individual biographies to interpersonal relations, family dynamics, socio-economic factors and cultural context. Research from other domains suggests that different social determinants can trigger variable degrees of social essentialism, and hence more negative intergroup attitudes [22]. While investigating the differential implications of divergent social attributions was beyond the scope of this study, the analysis provides proof-of-concept of substantial variability within the content of social explanations and emphasises the need for future research to take account of this nuance.

The high prevalence of social explanations overall in comparison to biological factors is consistent with some previously documented survey data, which imply that people most commonly attribute mental disorders to psychosocial factors, rather than biological [5]. However, previous studies have found personal factors comprise a large proportion of commonly-attributed ‘psychosocial’ factors [5]; in the current study, psychological factors accounted for only 8% of the present explanations. This difference may be a methodological artefact. Previous research regarding public aetiological beliefs has typically administered surveys containing a restrictive range of options from which participants may respond [5]. This can obscure attributions that the researchers have not anticipated or spotlight attributions that participants would not spontaneously volunteer–hence the importance of observing naturally occurring discourse. Of course, even naturalistic datasets are subject to the unique contingencies of that form of communication. For example, the infrequent reference to psychological causes in media data may reflect writers’ tendencies to avoid introducing causal explanations that imply a high degree of personal blame towards the affected person, in an attempt to ameliorate sigma narratives. Additionally, the position of life events/experiences as the most frequent attribution overall (20.86%) may be a result of news sources placing greater focus on reporting recent events of significance, such as crimes or serious accidents. Analysis of more informal naturalistic data, such as social media or verbal conversations, may yield different results.

The relatively low prevalence of biological explanations is notable. It suggests that concerns about the rising societal prominence of biomedical understandings of mental ill-health may be premature, at least within this cultural context. However it is worth noting that in the present findings, health/lifestyle factors presented similar patterns to biological factors in terms of their distributions across mental disorder categories. Features of this attribution category included factors such as substances/medication, lack of sleep, health issues, and diet/malnourishment. As these are primarily physical in nature, it is possible they are conceptualised in a similar manner to biological factors, though further research is needed to investigate this.

Are causal explanations represented differently across different categories of mental disorders?

While the overall findings presented higher rates of social attributions, this was not universal across mental disorder categories. Indeed, the overall figures disproportionately reflect the mental disorder categories which contained more causal explanations, such as eating disorders. This reiterates the importance of focusing analysis at the level of specific mental disorder categories, rather than a generic construct of ‘mental illness’. While eating disorders and anxiety disorders were most commonly attributed to social factors such as the cultural/societal environment, interpersonal factors, and life events/experiences, psychotic and mood disorders were most commonly attributed to health/lifestyle or biological factors. Additionally, while personality disorders were most commonly attributed to life events/experiences, they generated comparatively frequent rates of health/lifestyle and biological attributions.

These trends present both similarities and differences to previously documented patterns in public perceptions of aetiology. In terms of similarities, the findings reflect public tendencies to attribute personality disorders to trauma and stress [39, 40], and to more frequently endorse biological determinants for schizophrenia than other mental disorders [23]. However, the present findings differ from research which suggests that eating disorders are more likely to be attributed to personal factors [41], and that depression and schizophrenia are most likely to be attributed to acute stress [42]. Again, these differences may reflect divergence between media and survey methodologies; further research is required to clarify.

Because of the stigma implications of public perceptions of aetiology, it is also beneficial to consider how the findings map onto previously observed patterns in societal stigma. In the present study, more traditionally stigmatised categories of mental disorders were more likely to be attributed to biological or health/lifestyle factors than less stigmatised categories. Previous research has found that personality and psychotic disorders are highly stigmatised in comparison to other mental disorders types such as eating disorders or anxiety disorders [43]. These observed patterns align with theories of psychological essentialism which imply that biological attributions are associated with greater stigma attitudes, such as perceptions of dangerousness and desire for social distance [1].

What proportion of articles that mention mental disorders contain causal explanations?

Results indicated that approximately one-quarter of all articles that contained one of the specified diagnostic keywords included an explanation of the causes of that disorder. While it is difficult to interpret this figure without appropriate benchmarks, results confirm that causal attributions spontaneously occur on a regular basis in mental disorders discourse [5]. Given their documented significance for stigma dynamics [1], the study thus validates the maintenance of causal attributions as a research priority.

Did patterns of causal explanations change during the COVID-19 pandemic?

As a major societal event, the COVID-19 pandemic offers a good test-case to explore the malleability of causal attributions across socio-cultural context. The analysis produced some tentative evidence of temporal changes in attribution patterns post-pandemic, most notably a sharp rise in attributions to interpersonal relations, which included factors such as loneliness and social isolation. Along with prior research [27], this finding suggests that the pandemic may have increased public sensitivity to the social causes of mental disorders, though the durability of such effects remains to be seen. This emphasises the importance of situating research on causal attributions within their cultural/societal environment, and diversifying the geographic and historical evidence-base from which conclusions about attributions’ trends and effects are drawn.

Strengths and limitations

The study’s original contributions emanate from its observation of naturally occurring discourse, which garnered ecological validity that cannot be generated by artificially-constructed survey data alone. However, there are several limitations to the present study that should be considered. Firstly, it should be noted that public consumption of media is not exclusive to print and online news media sources alone, with notable alternative sources including social media, television, radio, and film. However, a strength to assessing news media specifically is the high level of trust and perceived reliability associated with it in comparison to other sources of information, such as social media [44, 45]. Moreover, much social media content involves dissemination of information originally issued via traditional media channels [46].

Secondly, the search terms in the current study present some limitations. While the use of DSM-5 diagnostic terms allowed for meaningful comparison across mental disorder types, and ensured the dataset was not overwhelmed by non-clinical references to terms such as ‘anxiety’ and ‘addiction’, results were affected by some diagnostic terms being more widely used than others. For example, the terms “anorexia nervosa” and “phobia” were used relatively faithfully, but the number of articles recovered about substance addiction was low due to infrequent lay use of formal diagnostic terms like “alcohol use disorder”. Similarly, the mood disorder category was more representative of bipolar disorder than major depressive disorder due to greater commonality of the term “bipolar disorder” than “major depressive disorder” in everyday language. The search terms also did not capture generic discussions of “mental health” or “mental illness”. While the analysis did not aim or claim to cover every single article published on the topic, it is unclear whether these keyword contingencies may have systematically biased the dataset in some way. The analysis itself sought to guard against bias through using an inductive code-development strategy and inter-coder reliability assessment.

Thirdly, the findings are necessarily limited to one particular cultural context, in a unique historical moment. This in itself represents an original contribution, as minimal published research has investigated Irish patterns of explaining mental disorders, or pandemic-related effects on attribution internationally. It is unclear whether or how unique features of the Irish Media landscape affect results. For instance, Irish newspapers typically do not have clear affiliations with certain political parties or ideologies. As political affiliations are associated with attributional tendencies, with right-wing attitudes more liked with biological and left-wing with social attributions [47], a more politicised media landscape may yield different results. Further work is required to establish how the patterns detected in this Irish study compare with other cultural contexts.

Finally, results are descriptive and do not facilitate inference about the cause or effects of the attributional patterns demonstrated. However this is appropriate for an initial exploratory study. This descriptive data is vital to inform future ecologically valid, hypothesis-driven research, for example regarding the influence of COVID-19 on attributions or the differential effects of various social attributions.