Gender and attitude in suicide posts on Cairo Confessions: a corpus-based study

1 Introduction

Suicide is a public health concern worldwide, accounting for 720,000 deaths annually and ranking as the third leading cause of death among 15–29-year-olds (WHO 2024). In the Arab World, Egypt reports the highest number of adolescent suicides (Qassim 2021). Nonetheless, due to strong stigma in Egypt’s traditional, religion-conscious society, many suicide incidents and attempts go unreported, casting doubt on official statistics and hindering access to professional help. Consequently, individuals with suicidal thoughts often turn to anonymous virtual communities rather than therapy, where they feel freer to express their emotions and thoughts. Analyzing attitudes in these virtual posts can provide deeper insights into the authors’ emotional and evaluative states, potentially supporting suicide prevention and treatment efforts.

Most linguistic research on suicide-related discourse has focused on the Western Hemisphere, hence leaving a significant cultural and linguistic gap. In addition, social variables like age and gender have seldom been examined (see Section 4). The present study addresses these gaps by turning to the Arab World, specifically Egypt, and considering the gender variable. It analyzes suicide posts on Cairo Confessions (CC), an Egyptian Facebook chat group with 336k followers and 310k likes on Facebook alone. A corpus of 130 English suicide posts totaling 22,850 and spanning seven years (2018–2024), has been compiled. Based on the authors’ self-declared gender, the corpus has been divided into male and female sub-corpora. However, the age variable has been excluded due to sample homogeneity. Adopting a mixed-methods research design, blending both qualitative and quantitative corpus-based analyses, the study investigates potential gender differences within the expression of Attitude and suicidal ideation. The theoretical framework integrates Martin and White’s (2005) Appraisal Model with Daouk et al.’s (2023) psychological scaling of suicidality. This exploration of gendered suicide posts would contribute to our understanding of some darker aspects of human experience and would hopefully better inform much needed suicide prevention efforts. However significant, the present findings need to be interpreted with caution because of the limited size of corpus, focus on English-only posts, and reliance on unverifiable self-reported age and gender.

2 Appraisal model

Martin and White’s (2005) Appraisal Model offers a thorough and systematic view of how language is used to express attitudes and emotions through appraisal systems. The model rests on the framework of Systemic Functional Linguistics (SFL) of Halliday (2004/1994) and Halliday and Matthiessen (1999). It particularly underlines the interpersonal function of evaluative language which has the power of constructing identities and negotiating positions and social relationships. Within M and W’s (2005) model, evaluative language is analyzed from a cognitive semantic perspective across three systems: attitude, engagement, and graduation. Engagement addresses how the text authors position themselves regarding the different viewpoints and how they engage in a dialogue with their readers. Authors may choose to contract or expand the voices expressed in a certain text. The system of Graduation deals with the intensity and strength (Force) as well as the degree of precision or detail (Focus) while expressing attitudes and evaluations.

The system of attitude, which is the focus of the present study, comprises three semantic regions for expressing attitudinal meanings: affect, judgment, and appreciation. Affect is an expressive resource that deals with specifying whether our emotions and feelings are positive or negative. Judgement is concerned with attitudes towards behavior and/or ethics, which we approve or disapprove of, commend or criticize. Appreciation deals with evaluations of semiotic and natural phenomena, based on how they are appraised in a certain context.

Attitudinal appraisal can either be explicit/inscribed or implicit/invoked/implied. Explicit attitude can directly be detected through positive or negative lexical items (as illustrated in (1)a) while implicit attitude can only be inferred through understanding the sociocultural aspects of the context (as illustrated in (1)b).

In (1)a, the expression more rewarding indicates intensified positive appreciation of the valuation of the previous job. In (1)b, on the other hand, the factual expression seems to be superficially neutral. Yet, awareness of the disturbing sociocultural impact of mother (re)marrying in a traditional society would prompt negative interpretation. Additionally, (1)b may imply negative judgement of the mother’s propriety for forsaking her children.

3 Scaling of suicidality

Based on an extensive study of suicidality and its demographic characteristics among Arabic-speaking individuals visiting an online depression screener, Daouk et al. (2023) psychologically divide suicidality into four levels arranged in ascending order as (1) Thoughts of Death, (2) Wanting to Die, (3) Thoughts of Suicide, and (4) Suicide Attempt. The levels are not exclusive to one another; they may co-occur within the same post. Not all Arab countries have equally been represented in the online survey and nuances in the dialectal expressions of suicidal ideation may have been missed since the study has been conducted in Modern Standard Arabic. However, Daouk et al. (2023) conclude that, in line with global trends, men have been linked to death by suicide while women have reported higher suicidal ideation and attempt.

4 Review of literature

Linguistic investigations of suicide discourse have either focused on suicide notes or posts (Jasim and Jaafar 2022). The former have been left by confirmed suicide victims with trackable history, in the form of letters, notes, diaries, or audios, hence inviting contextually informed qualitative case studies (Fata et al. 2021). Suicide posts, in contrast, are not linked to actual suicide incidents. They are one of the emergent discourses on social media platforms within the confessional genre. Their abundance and accessibility attract corpus-based and computational linguistics studies (Jaafar and Jasim 2022; Lao et al. 2022; Pan et al. 2023; Roubidoux 2012).

Qualitative analyses of suicide notes originally emerged within forensic linguistics with a view to probe into their genuineness in settings such as police investigations or court cases (Coulthard and Johnson 2007). Hence, several studies aimed to contrast confirmed genuine notes with simulated ones. Shneidman and Farberow (1957), for example, contrasted 33 simulated suicide notes with 33 genuine notes from the Shneidman and Farberow (1957) collection which amounted to 721 notes left by confirmed suicides in Los Angeles. Three types of thought units were identified: discomfort, relief (e.g. that the person will get rid of his/her troubles), and neutral (e.g., giving directions to their loved ones about certain after-death arrangements). Shneidman and Farberow (1957) found that genuine suicide notes were more verbose and contained more discomfort and neutral thought units. Osgood and Walker (1959) contrasted 69 genuine suicide notes with 72 neutral everyday notes. They found genuine notes to use more stereotypical expressions, more grammatical errors, more depressing words, and shorter sentences. Prokofyeva (2013) identified several features like clear reasoning in identifying the trigger for suicidality, expression of negative emotions like insecurity and hopelessness, expression of relief, and use of the past simple tense. Fata et al. (2021) consistently found Prokofyeva’s (2013) features in 11 suicide notes linked to confirmed suicide victims, suggesting some stereotypical linguistic behavior. Fernández-Cabana et al. (2015) studied a sample of 23 suicidal notes written in Spanish, employing a text analysis software, Linguistic Inquiry and Word Count (LIWC), to compare linguistic features by gender, age and environment. Most relevantly, they found that Spanish women wrote longer notes, with more affective content and positive emotions. Interestingly, expression of positive emotions in suicide notes, particularly relief, were found to be used by depressed individuals just prior to their suicide (Lester 2010; Rudd 2008). Generally, studies have found that the language of depressed individuals, including those with suicidal ideation, features more negative and fewer positive emotional expressions (Rude et al. 2004; Tølbøll 2019).

Computational investigations of suicidal posts aimed to automatically detect and assess suicidality. Zhang et al. (2021) proposed a transformer-based learning model to detect and classify online suicide notes based on their semantic features, hence contributing to suicidal risk identification on social media posts. To train the program, Zhang et al. used three datasets: suicide notes from existing corpora (e.g., Schoene et al. 2019), last statements by criminals before death sentences obtained through the Department of Criminal Justice, and neutral posts from non-suicidal sub-Reddits. They claimed that their model outperformed conventional machine learning models and deep learning approaches (e.g., Schoene and Dethlefs 2016). Lao et al. (2022) demonstrated how machine learning (ML) models could be used to investigate markers of suicidality and depression on the lexical level in suicidal posts from mental health–related forums on Reddit to assess and better understand a person’s suicide risk. Pan et al. (2023) also applied machine-learning approaches to the analysis of Weibo posts. They found that depression-related linguistic characteristics such as negative emotions (e.g., anxiety and sadness) significantly predicted the subsequent development of suicidal ideation.

Most relevantly, corpus-based analysis of suicide notes by Jaafar and Jasim (2022) aimed to identify linguistic patterns characteristic of this psychologically complex and critical genre. Jaafar and Jasim conducted a comparative stylistic analysis of two corpora of suicide notes: Shneidman and Farberow’s (1957) collection versus more contemporary data from Reddit (2012–2020). Through the utilization of corpus tools, Wordsmith and CLAWS, Jaafar and Jasim showed shared features between both corpora, like informal writing. The Reddit notes showed a tendency to be longer and more detailed, with frequent use of first-person pronouns, verbs, and repetitive vocabulary. Notably, marked use of first-person pronouns, along with negatively valenced words, was observed in the written essays by depressed college students in the United States who were thought to have higher self-focused attention (Rude et al. 2004). A frequency-based, contextually informed pronoun analysis of text messages by confirmed depression patients in the United States was argued to be beneficial in predicting depression symptom severity (Ren et al. 2024). Alternatively, Roubidoux (2012) found that first-person pronouns, especially the active form, I, would set a tone that illustrated linguistic manifestations of power characteristic of genuine suicide notes. Roubidoux contrasted a corpus of 60 genuine suicide notes by males in the United States from the Shneidman and Farberow’s (1957) collection with 33 simulated notes from the Forensic Linguistics Institute.

Most linguistic research on suicide discourse has focused on the Western Hemisphere, leaving a notable cultural gap. In addition, social variables like gender and age have seldom been considered (Fernández-Cabana et al. 2015). Furthermore, previous emotion and sentiment analyses relied on broad positive, neutral, negative categorization. Hence, addressing non-Western cultures, attending to social variables (e.g. gender), and offering a fine-grained analysis of attitudinal language are essential for developing a more effective and culturally sensitive understanding of the emotions and evaluative thoughts expressed in suicide discourse. Ultimately, this would eventually contribute to better-informed and culturally-sensitive suicide prevention efforts.

5 Methodology

6 Qualitative analysis

Close examination of a sub-set of the suicide posts under study has enabled identification of nuances in the expression of mostly negative attitudinal meanings as well as of levels of suicidality. A sample of two posts, one from each data sub-set, has been selected for a detailed analysis based on the dimensions of attitude.

The sample posts are characterized with negative Affect where feelings of unhappiness, insecurity, dissatisfaction, and disinclination, reflect severe internal struggle. Expressions related to any level of suicidality highlight the seriousness of the authors’ emotional distress. Negative self-judgement and positive other judgement surface when authors compare themselves and their (lack of) achievements to (happier and better-achieved) friends. This indicates a sense of self-worthlessness. Implicit negative judgement of others is motivated by their (presumed) inability to support and care for the distressed author, hence reflecting a sense of disappointment and isolation.

In the analysis table under each example post, the Appraiser is the post author. Invoked/implied attitude is marked by the notation ‘t’ (Table 3).

Example Post 1: Participant Code F30-24-02

Example Post 1 opens with brief positive self-Appreciation as the author acknowledges her looks, personality, and job. However, this is swiftly overshadowed by the trigger for her suicidality: a lifelong struggle with obesity. This struggle elicits intense negative Affect (insecurity, unhappiness, dissatisfaction, and disinclination) to the point of entertaining thoughts of suicide (Level 3) and fear of hurting oneself. Throughout the post, she expresses deep frustration over repeated failed weight-loss attempts, blaming herself through implicit negative self-Judgement. She also expresses negative Judgement of others for not defending or supporting her (-tenacity) while growing up, revealing feelings of abandonment and insecurity. Hurtful comments from others have further damaged her self-esteem, invoking negative Affect and moral judgement toward them (Social sanction, -propriety).

The dominance of negative Affect highlights her deep self-involvement and the enormous emotional toll of her lifelong struggle with weight loss. The post depicts how young women find themselves under the pressure of measuring up to society’s standards for female body image. In fact, obesity is one of the recurrent triggers for suicidality in females. In another post, a 24-year-old writes, I died the moment my mother told me I look fat (F24-23-16).

Example Post 2: Participant Code M24-18-02

Example Post 2 is saturated with negative Affect, expressed through disinclination and lack of motivation to live, dissatisfaction with achievement, heartbreak, despair, and insecurity over financial instability. These negative emotions culminate with entertaining thoughts of death (Level 1: suicidality) marked by a rhetorical question that concludes the post. The author compares himself unfavorably to successful friends, invoking negative self-judgement and positive Judgement of others. Ironically, he implicitly mocks his failure at a suicide attempt (Level 4: suicidality). Overall, the text reflects a bleak view of life, deep feelings of hopelessness, and financial despair – which constitute key triggers of suicidality (Table 4).

7 Corpus-based analysis

7.1 Pronouns: self-focused discourse

Quantitative evidence indicates that this sub-type of confessional discourse is predominantly self-focused. As shown in Table 5 and Figure 1, first person singular pronoun forms (Nominative I, Accusative me, and reflexive anaphor, myself) appear with the highest frequencies in both data sub-sets when compared with second and third person pronouns. First person plural forms (we, us, ourselves) have been excluded due to their ambiguous interpretations, whether inclusive, exclusive, or generic.

Table 5:

Detailed counts of 1st, 2nd, and 3rd person singular pronouns.

	First person			Second person	Third person						Total
	I	me	myself	you	he	him	himself	she	her	herself
Male	825	160	45	42	43	19	4	66	61	2	1,267
%	65.1	12.6	3.6	3.3	3.4	1.5	0.3	5.2	4.8	0.2
Female	1,053	245	51	99	98	51	1	94	76	6	1,774
%	59.4	13.8	2.9	5.6	5.5	2.9	0.1	5.3	4.3	0.3

Figure 1:

Summary of 1st, 2nd, and 3rd person singular pronouns.

The most frequent 1st person pronoun, I, has surfaced within the linguistic clusters (N-grams) characteristic of each sub-corpus (see Section 7.3). The accusative form, me, has surfaced most frequently (4 times) in the cluster pray for me (to die) as in examples (2a, b) and less frequently (2 times) in each of the clusters [metaphorically] killing me and someone like me (in (2c, d), respectively) in the female sub-corpus. In the male sub-corpus, me appeared in everyone around me (4 times, example (2e)), in that made me (3 times, example (2f)), and in is eating me (2 times, example (2g)).

Across both sub-corpora, the 1st person reflexive pronoun, myself, has appeared more frequently after verbs with negative meanings (kill, examples (2a, h); hate, example (2i); and hurt, example (2j)) and less frequently after positive verbs like accept myself. The latter bi-gram, accept myself, has occurred only once in the female sub-corpus where it is portrayed as an unachieved goal (example (2l)) and twice in the male sub-corpus where it is negated (example (2k)). Hence, counting this bi-gram as positive ideation is doubtful as indicated by ‘??’ in Figure 2.

Figure 2:

Bi-grams containing myself in relative frequency.

From Figure 2, kill myself shows the highest frequency among the attested bi-grams for both genders, with a significantly higher frequency in the female sub-corpus, suggesting a higher occurrence of suicidal ideation or expression in females. Thoughts of self-harm (hurt myself) are exclusively expressed by females. Self-hatred (hate myself) is almost evenly articulated by both genders, with a marked absence of self-love, signaling lack of positive self-regard.

(2)

Extracted examples on pronouns

a.

By the time this gets posted i would have killed myself already. Pray for me , I’m going to hell anyway but pray for me . (F21-19-08)

b.

I jus wanna die I want it badly… pray for me to die . (F28-19-02)

c.

I wanna take off hijab but guilt is killing me , Idk. (F19-18-03)

d.

someone like me cant be loved (F21-19-08)

e.

I honestly believe everyone around me would be better off if I didn’t exist, I just want it to end. (M28-18-01)

f.

have depression that made me suicide once 3 years ago (M22-24-03)

g.

Depression is eating me alive. (M22-24-01)

h.

i want to kill myself i want to end all pain once and for all i’m just afraid to do it but i’m sure the fear will go away one day (M23-19-04)

i.

I hate myself and I don’t like the way I’m (M25-22-04)

j.

I thought I grew out of it but sometimes I subconsciously attempt to hurt myself in more discrete ways. (F23-18-05)

k.

i just dont accept myself … please dont advise me to accept myself because i dont want to! (M26-22-02)

l.

tried to accept myself but i always end up feeling shitty about myself (F21-18-11)

m.

How can you be truly happy?Well I used to be happy but that was 4 or 5 years ago (M22-24-06)

n.

Finally, thank you for your patience in getting through my scrambled thoughts, have a wonderful day!! (M32-21-04)

o.

I miss you as usual, the things that reminds me of you are infinite (F30-18-04)

Regarding the 2nd person pronoun, you, male authors have tended to use it either generically (2m) or to engage with the virtual reader(s) (2n). Female authors, on the other hand, have most often used you to address significant others in their lives (2o). Depressive thoughts and expressions of suicidal ideation can be observed across the examples on 1st person pronouns in (2a-j).

7.2 Adjectives: dimensions of attitude

To explore potential gender differences on the use of the three dimensions of attitude, this study contrasts the attitudinal meanings of the most-frequent 100 adjectives in each sub-corpus. While “a given attitude can be realized across a range of grammatical categories (Martin and White 2005: 10), yet adjectives are key to expressing disposition-oriented attitude. Appendices A and B detail the results on the positive versus negative attitudinal meanings across the three dimensions of Affect, Judgement, and Appreciation. Careful examination of the surrounding context through the concordance lines and/or file views has enabled accurate interpretation of the intended meaning.

To normalize the word count difference across the two sub-corpora, adjective relative frequencies have been calculated using the formula (Raw Frequency/Total Word Count of Sub-corpus*100,000) in Excel. Figure 3 displays the relative frequencies of positive and negative Affect, Judgement, and Appreciation among males and females based on the 100 most frequent adjectives. Both genders use more negative than positive attitude adjectives across all dimensions. Males use significantly more positive adjectives across all dimensions while negative Affect is similarly dominate in both sub-corpora. Females show a balance between negative Affect and negative Judgement, which appear at two pronounced peaks. Males, by contrast, show a significant dip with negative Judgement but surpass females in negative Appreciation.

Figure 3:

Relative Freq. of ± Attitude in the 100 most frequent adjectives.

Dominance of negative Affect across both genders suggests that these texts are uniformly charged with negative emotions. Females’ greater use of negative Judgement aligns with evidence presented in Section 7.3 on the characteristic tri-grams where females express negative Judgement of those who let them down. In contrast, male tri-grams in Section 7.4 reflect a sense of insecurity over material lack, which explains their higher expression of negative Appreciation of their life situation and financial troubles.

Figure 4 details the subcategories of negative Affect, Judgement, and Appreciation among the most frequent 100 adjectives by relative frequency. Males more often express insecurity, vulnerability, and dissatisfaction while females show more sadness and misery. Both genders reveal dissatisfaction through boredom and frustration. Males show no expression of anger or aggression whereas females show only one occurrence, (3).

Figure 4:

Relative Freq. of -Attitude sub-categories in the 100 most frequent adjectives.

(3)

I am a person who used to light up the room, now I became the person who suffers silently , I stopped talking, I started to accept the pain , the lies, the hurt, the words that tore me, I stopped caring, I stopped feeling , I became numb , I became angry and aggressive . I now feel heavy hearted when thinking about walking 2 steps to wash my teeth, I am dead inside . (F24-23-16)

Anger and aggression are outward emotions that may infringe on others. In (3), however, the author frames and softens these with internal feelings – withdrawal, silent suffering, acceptance, numbness.

Concerning negative Judgement, men would more likely criticize others’ abilities and personal qualities than their ethics or morality. Females, on the other hand, show a balance over the two sub-categories of Judgement. In negative Appreciation, males express more negative evaluations of the impact of objects and phenomena (e.g., boring and traditional, Appendix B), and their composition and value (e.g., weird and worthless, respectively in Appendix B). Females seem to focus their criticism on the negative quality of objects and phenomena, e.g. particularly reporting having suicidal thoughts (see Appendix A and Figure 8).

This section has highlighted areas where men and women display similar or different evaluative styles. Further gender effects are revealed in the following two sub-sections.

7.3 N-grams: gender-specific content differences

Distinctive linguistic patterns characteristic in the male and female sub-corpora have been identified by setting up each sub-corpus as a reference corpus for the other and then using the Clusters/N-Grams function in AntConc. Most characteristic tri-grams appear in the context of depressive and suicidal thoughts. Figure 5 illustrates the normalized relative frequencies of the most prominent tri-grams across both sub-corpora.

Figure 5:

Tri-grams across both sub-corpora.

Males have used the tri-gram I can’t twice as often as females, indicating greater openness in expressing vulnerability (e.g. examples (4a–c). Females have often balanced out expressions of powerlessness and despair with expressions of rationalization. In (4e), for example, the female author expresses a wish to die but then decides against suicide. In both sub-corpora, I can’t appears mainly in the context of depressive and suicidal thoughts as shown in (4).

(4)

Examples on I can’t

a.

I’m done with life I can’t do it anymore (M26-21-03)

b.

I can’t enjoy anything (M24-18-02)

c.

I can’t live like this (M17-21-01)

d.

I can’t communicate with others (F21-18-11)

e.

I can’t live anymore, i wish i could die but I’m not going to kill myself no I’m not stupid. (F19-19-06)

f.

I can’t take it any more I wanna kill myself. (F20-18-02)

The usage of I don’t and I want to is relatively balanced between the genders, with marginal differences between the two groups. The examples in (5) indicate that the desires expressed by I want to are strongly linked to dying and ending one’s life.

(5)

Examples on I don’t and I want to

a.

I don’t find a reason to be alive (M22-18-03)

b.

I don’t know why these thoughts pop into my head (F26-21-03)

c.

… i dont deserve her love and that i dont deserve anything (M18-24-07)

d.

i don’t deserve anything good (F24-23-11)

e.

I want to die i can’t stand myself (M24-20-07)

f.

I want to die as soon as possible I’m done with life (M26-21-03)

g.

I want to disappear off the face of the earth. (F42-23-02)

h.

I want to end my life (F24-19-05)

i.

i want to kill myself i want to end all pain once and for all (M23-19-04)

Characteristic of only the male sub-corpus are the expressions I tried to and I used to. The former is contextually linked to committing suicide as in (6a) and (6b). The latter, I used to, expresses (often happy) past states like being happy (6c), good (6d), and successful (6e).

(6)

Examples on I tried to and I used to

a.

killing myself…it’s s****d and boring yet I tried to do it twice (M21-24-02)

b.

i tried to end my life two times (M18-24-07)

c.

I used to be happy but that was 4 or 5 years ago. (M22-24-06)

d.

I used to be the “good boy” who doesn’t make sins (M22-24-06)

e.

I used to be successful I used to play sports (M21-24-08)

Characteristic of only the female sub-corpus are the expressions I’m not and I have no. The former expression has been used, for example, to express incapability (7a, b) and a resolution to not commit suicide (7c) for the sake of her family.

(7)

Examples on I’m not

a.

I’m not at my best mental state (F25-22-13)

b.

I’m not comfortable in MY OWN SKIN (F24-18-13)

c.

I’m not gonna commit suicide I wouldn’t do this to myself and my family (F19-18-03)

The latter expression, I have no, has been used differently in each sub-corpora as shown in Figures 6 and 7. Males seem to complain about having no dreams, family, ideas, or power.

Figure 6:

Usage of the tri-gram I have no in the male sub-corpus.

Figure 7:

Usage of the tri-gram I have no in the female sub-corpus.

Females, on the other hand, have focused their criticism on having no emotional support system which involves family and friends (see Figure 7). Females have expressed disappointment at having no one to confide in, someone they can pour their hearts out to.

Across the examples presented so far, some innovative expressions have been used to convey a desire to die or to kill oneself (i.e., commit suicide). To illustrate, consider “disappear off the face of the earth” (5g) and (5i). More innovative expressions of suicidal ideation are presented in the following section.

7.4 Suicidality

Expressions denoting all four levels of suicidality have been observed with varying frequencies in the data and overlapping patterns. In the following post, for example, Level 4 (Suicide attempt) appears at the beginning of the post, then Level 2 (Wanting to die) appears in the closing.

(8)

a.

Level 4: I tried to commit suicide multiple times. (F23-18-05)

b.

Level 2: Lately, all I can think of is how much I want to die. But I realize it’s a sin and I will nottt do it. But I pray to God all the time to take me. (F23-18-05)

In (7b), the author exhibits a fluctuating thought pattern, initially expressing a strong urge to die (or, commit suicide), followed by the realization that wanting to die/committing suicide would be a sin, and finally, a firm decision not to proceed. Hence, she says that she continually prays to God to do it for her, to “take” her life, as this would free her from committing a sin, leaving her hands clean.

Suicidal ideation has been targeted in the two sub-corpora through the following search terms: death, die*, dying, liv*, life, end*, stop*, suicid*, kill* myself. Close examination of the concordance lines enabled categorizing the search output across the four levels of suicidality. Normalized relative frequencies have been calculated in Excel to account for the size difference across the two sub-corpora. Figure 8 indicates that females consistently have a higher tendency to articulate suicidal ideation, with a marked surge in expressions denoting Level 3: Suicidal thoughts.

Figure 8:

Relative frequency of expressions of suicidality.

For both genders, Levels 2 and 3 are higher than Levels 1 and 4. Level 1 expressions usually serve, among other alternatives, as an introduction to Level 2 or Level 3 expressions. Consider examples (4e, f) above repeated below as (9a, b).

(9)

a.

I can’t live anymore, i wish i could die but I’m not going to kill myself no I’m not stupid. (F19-19-06)

b.

I can’t take it any more I wanna kill myself . (F20-18-02)

c.

I am tired and I want to die (F23-23-09)

In (9a), expressing inability to live anymore (Level 1) introduces a wish to die (Level 2) while in (9b) Level 1 introduces suicidal thoughts about wanting to kill oneself (Level 3). Alternatively, as illustrated in (9c), expression of exhaustion or boredom may pave the way for wanting to die (Level 2).

The sharp decline from Level 3 to level 4 for both genders (Figure 8) indicates that while many express suicidal thoughts, fewer admit to previous attempts. This may reflect fear of stigma or a choice not to act on such thoughts. Harmer et al. (2024) discuss the Ideation-to-Action framework, where suicidal thoughts and attempts are connected yet distinct behaviors. The transition from contemplation to action remains unpredictable, being subject to several personal factors.

A great deal of similarity has been observed across both genders in terms of the reported reasons for not proceeding from ideation to action. The reasons included negative evaluation of the action as stupid or irrational ((4e), (11a)), considering its impact on their families ((7c), (11b, c)), religious apprehensions ((8b), (10a, b), (11d)), or just being too scared to do it ((10d), (11e, f)).

(10)

Examples of females deciding against a suicide attempt

a.

i tried many time to commit suicide before but because its haram [i.e. forbidden in Islam] i stopped (F21-18-14)

b.

I’m starting to have suicidal thoughts maybe the only thing stopping me is religion (F19-22-10)

c.

I have suicidal thoughts, and I know this is the answer, but again I am too coward to commit suicide (F28-19-13).

d.

I want to die and yet I’m very scared of death (F33-19-12)

(11)

Examples of males deciding against a suicide attempt

a.

I really am hoping I would die, I am just too sane to commit suicide (M30-23-01)

b.

i need someone to motivate me not to kill myself because it will breaks my mother’s heart (M24-19-01)

c.

the only thing stopping me from taking out my useless life is my mom bc i’d break her heart (M28-20-08)

d.

if committing a suicide was a halal thing [i.e. allowed in Islam] I would do it without hesitation (M25-22-04)

e.

i wanna suicide but i know i wont because im way too coward to do it! (M26-22-02)

f.

I considered k*lling myself but I got too scared (M36-23-02)

Although both genders share similar reasons for not acting on suicidal thoughts, females have often exhibited a more assertive language style in conveying suicidal ideation. In (10c, d), for example, the female authors use concise factual statements in the present simple tense. In (10d), the female author further includes an assertive convening statement, and I know this is the answer. Male authors, in contrast, hedge their expressions through using the present progressive (instead of the factual present simple), hedging modals (really, would), and less forceful lexis (hope) in (11a). In (11b), a lonely male author turns to the virtual audience for intervention. In (11d), the conditional and the hedging modal (would) weaken the progression to action. Further support to female assertion can be observed by comparing the concordance of decided across the two sub-corpora as illustrated in Figures 9 and 10.

Figure 9:

Concordance of decided in the female sub-corpus.

Figure 10:

Concordance of decided in the male sub-corpus.

One of the assertive frames in English is the expression I decided to …. In the female sub-corpus (Figure 9), this frame occurs five times, three of which (i.e., 60 %) with the complement VP end my life/end it all. In the male sub-corpus (Figure 10), this frame occurs six times, none of which with the complement VP end my life.

Concordance of end my life also reveals more assertion in the female sub-corpus to kill oneself (Figure 11), where it is preceded by want to (i.e., desire to), tried a lot to (i.e., several repeated attempts), and twice after decided to (i.e., resolution to). In the male sub-corpus (Figure 12), authors less assertively articulated battling with the urge to (i.e., resisting the idea) of ending one’s life, or reporting two past attempts.

Figure 11:

Concordance of end my life in the female sub-corpus.

Figure 12:

Concordance of end my life in the male sub-corpus.

8 Discussion and conclusion

This study examined gender differences in expressing Martin and White’s (2005) Attitude dimensions (Affect, Judgement, Appreciation) and Daouk et al.’s (2023) four levels of suicidality in 130 English digital suicide posts by mostly young Egyptian users. Gender-neutral traits are summarized in (12) and gender-specific differences in (14).

Notably, high use of 1st person pronouns in the suicidal genre, (12a), was reported in the literature (Jasim and Jaafar 2022), and was found to signal depression (Rude et al. 2004; Tølbøll 2019) or predict its severity (Ren et al. 2024); or to manifest power (Roubidoux 2012).

As expected in this genre and in conformity with previous studies (Rude et al. 2004; Tølbøll 2019), negative emotions and evaluative thoughts surpassed positive ones across both sub-corpora. However, closer examination of the sub-categories within Affect, Judgement, and Appreciation revealed that males and females had different evaluative styles (Section 7.2). With regard to the sub-categories of negative Affect, for example, males displayed higher expression of insecurity as well as dissatisfaction while females tended to express more sadness and misery, (14b). Obesity was observed as a recurrent trigger for suicidality in the female sub-corpus. Material and psychological losses were the recurrent triggers in the male sub-corpus. Observing these gender differences would result in more targeted psychological support.

Religion has been reported by the males and females in this study as one of the reasons for them to not act upon their suicidal thoughts, (12f), (see examples (8b), (10a, b), (11d) in Section 7.4). Egypt’s dominant religion is Islam where the sanctity of life is stressed and suicide is haram (i.e., prohibited) as a major sin. In Surah An-Nisa (4:29), for example, there is a direct order against the killing of oneself in “وَلَا تَقْتُلُوا أَنْفُسَكُمْ إِنَّ اللَّهَ كَانَ بِكُمْ رَحِيمًا …” ‘And do not kill yourselves (nor kill one another). Surely, Allâh is Most Merciful to you.’ (Trans. Al-Hilali and Khan 1998, p. 125). In prophetic hadith, suicide is punishable by eternity in Hell. Nonetheless, religious awareness does not seem to prevent expression of a strong drive to proceed. In example (2a) from section 7.1, repeated below as (13), the author, who seems intent on killing herself, is aware of the punishment for such sin, yet desperately and repeatedly asks the audience to pray for her anyway.

Gender effects have been observed, (14). Contrary to the Spanish women in Fernández-Cabana et al.’s (2015) study, Egyptian females in the present study wrote briefer suicide notes and expressed less positive Affect, Judgement, and Appreciation. Other findings listed under (14a) suggest that Egyptian females either have higher suicidal ideation due to life stresses or are more open (than males) to express such emotions and thoughts. In psychological research, Deng et al. (2016) distinguish between emotional experience and emotional expressivity as two types of emotional response. They have found that, when watching videos that induce an emotional response, Chinese men have had intense emotional experiences, whereas Chinese women have displayed higher emotional expressivity, specifically for negative emotions. American women have been found to show greater emotional expressivity on the overall (Kring and Gordon 1998), and in response to negative stimuli in a picture-viewing task (Gard and Kring 2007). American men have been found to ‘take in’ their emotions despite being more physiologically affected by stressors (Levenson et al. 1994). In the Arab World, Daouk et al. (2023) conclude that men have higher rates of death by suicide while women have higher suicide ideation, planning, and attempts. El Halabi et al. (2021) argue that masculinity is traditionally linked, in the minds of Arab men, with physical strength, prestige, and resilience. They have found that middle-aged Lebanese men often find it difficult to admit to defeat and depression, which results in a higher risk of developing depressive pathophysiology and a reluctance to seek professional help.

Further to the gender-specific differences, Egyptian males have been more open than their female counterparts to express their own inability and vulnerability, which runs against the traditional social construct of Arab masculinity (El Halabi et al. 2021). This may suggest generational development, especially considering that the males in the present sample are relatively young, with an average age of 24.18 years.

The findings of this study should be approached with the understanding that since all the examined posts are written in English (L2), emotional expressivity in L2 may have had an effect, especially that the posts often show poor command of the language with frequent grammatical and language mechanics issues. Focusing on posts written in English is because the dominant language on CC is English, not Arabic (L1). English dominance could be a characteristic of writing in the digital medium regardless of the genre. Alternatively, authors under emotional stress may avoid L1 (Arabic) and opt for L2 (English) to distance themselves emotionally from their (especially sensitive) troubles and reduce empathy, i.e., achieve emotional blunting (Larson 2018; Romero-Rivas et al. 2020; Youssef forthcoming).

Challenges to corpus-analysis in this study stem from the nature of the genre under examination. One challenge is the use of informal language and edited forms like I’ve been thinking about su**ide for some time… (F28-24-03). Another challenge is the metaphorical use of dying as in I have been dead inside for 4 years now. Depression is eating me alive. I died and I watched everyone around me progress with their lives (M22-24-01). A third challenge is the nuances in expressing suicidal ideation. Level 2-Wanting to die, for example, has been expressed through phrases that do not include the performative verb die such as I pray to God all the time to take me (F23-18-05), i want to kill myself i want to end all pain once and for all (M23-19-04), and I just want it to end. (M28-18-01).

Future research could explore suicide-related discourse in other cultures and languages, the triggers of suicidality, the nuances in the expression of suicidal ideation, and the extent to which these may be influenced by variables like self-declared age and gender, which remain unverifiable online.