Sociological autopsy: An integrated approach to the study of suicide in men

This paper ’ s main aim is to argue the methodological case for a particular approach to researching the sociology of suicide. By way of illustrating the use of this approach it also offers some brief examples of substantive ﬁ ndings about the gendered character of men ’ s suicides. The ﬁ rst half of the article explains and justi ﬁ es the research approach. This is a qualitatively-driven mixed method and dual paradigm study of individual suicides. It is a sociological study which draws on the tradition of psychological autopsies of suicide; hence the term ‘ sociological autopsy ’ . The second half of the article offers brief illustrative ﬁ ndings from a speci ﬁ c research project which employed the sociological autopsy approach. This was a study of 100 suicide case ﬁ les from a coroner ’ s of ﬁ ce in the UK. There is discussion of common sense assumptions about suicide in men; the construction of evidence in case ﬁ les; a typology of gendered suicides where relationship breakdown seems to be the principal trigger; and the value of case-based analysis, with a single case discussed in some detail.

an explanation of the method.The second half offers brief illustrations of applying the method, in relation to suicidal behaviour in men.The paper is distinct from the rest of this special issue in having a primarily methodological rather than an empirical focus.

The development of sociological research on suicide
To introduce this method, we begin with a brief summary of the development of sociological research on suicide.Sociology is in fact only a bit-part player in terms of worldwide suicide research.Agerbo, Stack, and Petersen (2009) found that out of over 30,000 academic papers on suicide published since 1980 (and in the Institute of Science Information database) there were only 400 which could be categorised as sociological.Despite the marginal position of sociology with the wider field of suicide research, the best-known sociological approach to the topic, that of Durkheim ([1897] 2002), has been very influential both within and beyond sociology and is still very frequently used in contemporary studies.Durkheim's idea was to study the social context of an ostensibly individual act by examining associations between suicide rates and various social factors (such as integration and regulation).This approach has generated considerable debate.Douglas (1967),inan important contribution, argued in opposition to the Durkheimian tradition for a Weberian emphasis on the subjective meanings of suicidal behaviour to social actors.Although his book is often cited now as an important contribution, Douglas's research agenda was not taken up to any noticeable extent.As Stack's (2000aStack's ( , 2000b) ) review papers show, most contemporary sociological research on suicide is exclusively quantitative and much of it is in the Durkheimian tradition of research on suicide rates in relation to a wide spectrum of social variables and does not explore individual-level data.As for the wider field of suicidology beyond the discipline of sociology, this is also dominated by quantitative methods.Hjelmeland (2008) found that 2.8% of all articles published in the three international suicide research journals in 2005e2007 were based on qualitative research.Atkinson (1978) made another important contribution to the sociology of suicide, in deconstructing through careful ethnomethodological research the reliance of published suicide rates on common sense reasoning by coroners in making their verdicts.Atkinson's conclusion was that we can only address suicide prevention through sociological research to a very limited extent, given the problems that there are with knowledge about suicide cases.

A sociological autopsy approach to suicide research
There is an important and well-established tradition of psychological autopsy studies within the field of 'suicidology'.These are studies of individual cases of suicide that are conducted post-mortem, usually conducted by psychiatrists or psychologists.They typically involve the measurement of risk factors, using a relatively small sample of suicide cases.The study of individual suicides is generally seen by sociologists as irredeemably psychological.Durkheim rejected both the study of individual suicide cases and the relevance of a psychiatric dimension.He has arguably been proved wrong, insofar as the study of individual cases by psychiatrists and psychologists has revealed the high proportion of suicides which feature mental illness, often undiagnosed (see Cavanagh, Carson, Sharpe, & Lawrie, 2003).The term 'social autopsy' has been used by sociologist Eric Klinenberg (2002),t o convey the idea of the social and political context of deaths during the Chicago Heat Wave and Klinenburg does work with individuallevel data, although he has been criticised by Duneier (2006) for a lack of depth of engagement with individual stories, resulting in the ecological fallacy (Robinson, 1950) of assuming individual circumstances from area-level data.Owens, Lambert, Lloyd, and Donovan (2008) have applied qualitative approaches to a psychological autopsy study of suicide and have generated important sociological insights, but they are cautious in maintaining the term 'psychological autopsy'.
We differ from both Klinenburg and Owens et al. in preferring the term 'sociological autopsy', with its conscious mimicry of the disciplinary claim of the psychological autopsy.We are not the first to use the term 'sociological autopsy'.It can be found in the work of Chatterjee and Bailey (1993) and Slater (2005), though only the latter in relation to suicide.We have set out to explore the possibilities for sociological research on individual suicides.In the light of the pioneering work of Douglas and Atkinson, a major challenge is to construct a successful study which both examines what we know about suicidal lives and takes a critical stance on the knowledge itself.Such a study needs to have a very broad interpretation of what is meant by the social context of suicide, to include the social construction of knowledge.Within the mainstream psychological autopsy literature, it does seem as though there is room for a social focus.Cavanagh et al.'s (2003) systematic review of psychological autopsy studies noted that evidence from these studies on psycho-social factors is limited.Our study can help to fill that gap, by taking a consciously sociological approach.Our study is also unusual in using a qualitatively-driven approach to this kind of study (though see Owens et al., 2008;Stack & Wasserman, 2007), allowing for case-based as well as variablebased analysis.

Study methods and data analysis
The study's data set is taken from the cases files of a district coroner in the UK.The study was given ethical approval by an NHS Multi-centre Research Ethics Committee.The committee's regional location is intentionally removed here to preserve the anonymity of the coroner's district, given the distinctive features of some suicide cases.A team of three researchers read a sample of 100 suicide case files in a coroner'so f fice which covers a medium-sized city, an adjacent rural area and an industrial town.The sample was of the first one hundred suicide verdicts encountered from 2002 to 2005.The 100 cases are broadly similar in terms of age and sex to the picture for the whole of England and Wales; the ratio of male suicides to female was just under 4:1 and the average age was 46 (44 for males, 53 for females).The city which makes up the biggest part of the population of this district has a minority ethnic population a little above the UK average of 7.9%, but it should be noted at this point that there were insufficient data on ethnicity in the coroners' files to allow for inclusion of ethnicity in the data analysis.Data providing indicators of social class, such as occupation, were also partial or missing in many cases and data on sexual orientation were not collected by coroners.Because of the amount of data in the files, decisions had to be made about what to record.What was taken away from the coroner'soffice in the form of an anonymised data set is therefore a combination of the researchers' own notes and verbatim data excerpts, some of which are quite lengthy.The process of analysis had therefore inevitably begun as we were reading the files.
The data in the case files were so diverse that to consider documentary research as mono-method in this context would be inaccurate.The various kinds of sources included forms filled out by coroner; scribbles by the coroner on file wallets; police statements from witnesses and significant others; forensic pathology reports; medical letters and reports, especially psychiatric ones; suicide notes; mobile phone records; photographs of corpses; letters to the coroner and newspaper clippings.These data sources are so diverse in terms of the conditions under which they were designed and produced that this could arguably be seen as a 'multi-modal' data set (Fincham, Scourfield, & Langer, 2007).Coroners' officers have a key role in putting together the file of evidence for the inquest.In some districts, these officers will also take statements from family members.In the district where our research was based, however, it was police officers in local stations who took evidence from witnesses, including family and friends.As research by Davis, Lindsey, Seabourne, and Griffiths-Baker (2002) shows, the role of coroners' officers is very variable between coroners' districts.Arguably, therefore, local police officers emerge in our study as the key professionals in relation to generating evidence about suicide, although it is coroners who make the crucial judgement about how a death should be categorised.
In contrast to mainstream psychological autopsy studies, we believe it is important to recognise the inevitably interpretive dimension of the analysis of suicide cases.Rather than employing solely qualitative approaches, however, we have attempted to go beyond a small case study sample by generating a robust sample of 100 suicides over a three-year period.This has allowed us to conduct a quantitative analysis of themes that were originally generated via qualitative code-and-retrieve analysis.Ours is a 'qualitatively-driven' mixed method study (Mason, 2006) that affords the opportunity for both case-based and variable-based analysis (the distinction made by Ragin (1987)).The three researchers agreed a coding frame and thematically coded whole cases using N-vivo software.Coding whole cases, rather than breaking data up into excerpts, meant that the complexity of cases could be preserved within the N-vivo project and also that a tabular coding profile could be quickly produced by the software.This tabular coding profile was exported into SPSS for statistical analysis.Not everyone would agree that qualitative coding is suitable for quantification.We were careful to agree a coding frame, based on initial analysis, which was then systematically applied to the 100 cases on the basis that if a social factor was indicated in the evidence the relevant code would be attached to the case.Further analysis on the relative importance of this factor was conducted for specific themes, such as relationship breakdown.
There are potential difficulties with using coroners' data of course.There can be variation between coroners as to which cases deserve a verdict of 'suicide' and which an 'open' verdict.Timmermans (2005) has conducted ethnographic research with medical examiners, the nearest equivalent to coroners in the USA.He argues that an under-reporting of suicides can be explained by the very professional characteristics that protect the medical examiners' authority, namely the insistence on legal thresholds, the privileging of pathological evidence and maintenance of close relationships with law enforcement and medical personnel.We visited two neighbouring coroners' offices, to interview the coroner and read ten files from each place.This satisfied us that the kinds of evidence used were broadly similar to those in our main research site.This does not address differences in interpretation by coroners, but some published research on this issue would justify the use of data from only one coroner.Sainsbury and Jenkins (1982) argued, on the basis of data from England and Wales in the 1950s, 60s and 70s, that errors in reporting deaths were randomised and therefore made no difference to comparisons within and between countries.Pescosolido and Mendelsohn's (1986) research in the USA found there was systematic misreporting of suicide, but that this had little impact on the variables commonly used to test sociological theories of suicide.It is important to note the silences there can sometimes be in coroners' files.There are some very thin files in which the contextual social information that we might expect to find is absent and few data are available beyond reports on pathology and location of the corpse.

The theoretical implications of working with diverse documentary data
The first important theoretical issue to consider is what kinds of social phenomena are evidenced in these data.Our argument in relation to this project is that we need to seek insights into two different dimensions of reality revealed by the case file data.The first of these dimensions is the way that evidence is constructed by all parties (both living and now dead) and the second is the evidence we have about the beliefs and actions of suicidal individuals.This is a sociological necessity, as understanding the social context of suicide inevitably involves attention to how knowledge about suicide is constructed by professional and lay actors as well as attention to the circumstances of the suicidal individual, insofar as it is possible to know about these circumstances.We could therefore make a rather tentative claim to be operating dual paradigms in this project as well as combining qualitative and quantitative methods.We are relaxed about the relatively unusual juxtaposition of objectivism and constructionism.Whilst recognising it is a concept developed in a rather different context, the 'cautious naturalism' outlined by Fine (1997Fine ( , 2007;;after Gubrium, 1993) in relation to historical interpretation and the construction of social problems is helpful here.Fine argues that for strict constructionists to reject the possibility of studying objective structural conditions is 'ontological hopelessness ' (1997: 298) and that for sociologists to 'ignore the effects of social structure is to deny their birthright ' (1997: 317).A cautious naturalism 'recognizes that confident knowledge and interpretive schemata both contribute to our learning from the past' (Fine, 2007: 33).From coroners' files on suicide it is possible, we would assert, to make some moderatum generalisations (Payne & Williams, 2005) about the social structural context of suicide.So if, for example, a relationship breakdown is being cited as significant in a suicide note and by friends and family of the deceased then we can reasonably conclude that it was an important part of the suicide's social context.If there are gendered features of the reaction to the relationship breakdown, as described in the inquest evidence, then we can reasonably make some connection to social structure.We must also accept that evidence about relationship breakdown is constructed on the basis of various common sense theories about relationships and about suicides.These assumptions are of course amongst the repertoire of knowledge available to suicidal individuals (Canetto, 1992e1993) and should therefore be taken seriously.
In contrast to the Durkheimian tradition of research on suicide rates (Durkheim, [1897] 2002), we would argue the sociological relevance of qualitatively-driven mixed method research on individual suicides for generating insights into the social structural context of suicide.What this leads us to is an acceptance of the complexity of individual cases and a reluctance to reduce any sociological analysis to a mono-causal interpretation.Sibeon (1999) has provided the theoretical basis for an anti-reductionist sociology.This is not a post-modernist approach, as post-modernism is concerned with fragmentation but not necessarily consolidation.The aim of an anti-reductionist sociology is to consolidate a complex mix of social explanations.Encompassing tensions and even contradictions in data rather than eliminating them might mean having messy, not smooth accounts, however (Law, 2004).It also ultimately means employing a psycho-social approach, as although we are arguing the relevance of sociology to the study of individual suicides, it would be disciplinary arrogance to deny the importance of psychological insights.
At this stage we move on to present a brief example of the insights that can be developed from a qualitatively-driven sociological autopsy of individual suicides.The example of an empirical theme we have chosen is the gendered character of men's suicides.Gender is important because of the disproportionate number of suicides in men (at least three times the rate of female suicides in most countries, China being the most notable exception) and higher levels of suicidal thinking and behaviour in women.Canetto and Sakinofsky (1998) use the term the 'gender paradox' of suicidal behaviour in relation to these trends.We attempt to focus in the following material on gendered identities and gendered practices rather than a simple comparison of men and women as sex groups, as this is of course necessary for a proper social scientific understanding of the issue (Canetto, 1997, Scourfield, 2005).

Brief illustrations of the research approach: understanding suicide in men
We present in what follows some relatively brief examples of the insights that can be gained from the different dimensions of a sociological autopsy.This is not to present any major empirical claims, but simply to illustrate briefly the potential of the method.We begin with a constructionist approach to critically evaluating the evidence and then move on to more objectivist conclusions about the social and cultural context of suicidal behaviour.We conclude by discussing the importance of preserving the complexity of individual cases.
To explain how substantive themes emerged from the data it is worth returning to the process of data coding.Case files were read by the research team, a coding scheme agreed and whole cases (not data excerpts) were thematically coded using N-vivo version 2.0.As is often the case with qualitative research, the coding system was developed through contact with the data rather than being designed to test prior hypotheses.Initial codes referred to evidence in a case file that a social or behavioural factor was present.Any indication from the file of, for example, isolation or alcohol problems would be coded accordingly.Although the analysis was inductive, the research team's attention was inevitably drawn to some factors which are well-known from the epidemiological literature on suicide.This initial coding did not reflect the relative importance of any factor to the suicidal act, but simply evidence from the file that the factor was present.More detailed qualitative analysis then followed for key themes.
In the brief empirical illustrations which follow (in which all names are pseudonyms), we concentrate on money/work problems and relationship breakdown.We have focused on these in part because there were interesting gendered patterns within the data set.It can be difficult to quantify gendered patterns because of the small numbers of women in the sample of 100 suicides.However, as is explained in full in Shiner, Scourfield, Fincham, and Langer (2009), an analysis based on Cramer's V found the following factors to be associated with gender within the data set: employment problems and criminal activity (more frequent in suicides by men); isolation, diagnosis of mental illness and problems related to children (more frequent in suicides by women).Also, although this was not a statistically significant pattern, debt was mentioned as relevant to the suicide by inquest witnesses for 11% of men (n ¼ 9) and 5% of women (n ¼ 1).Similarly, relationship problems were identified by the research team as ostensibly the main trigger for the suicide at twice the rate in men as women (38% [n ¼ 30]: 19% [n ¼ 4]).To introduce the cultural context of the suicide cases, as noted earlier, ethnic monitoring data were lacking in the coroners' files.Since more than 90% of the local population are white British, it is fairly safe to assume this is a very largely white sample.Given diversity of gendered identities and practices it is difficult to concisely summarise the cultural context of white British masculinity.Arguably the most important context to note is the widespread discourse of the crisis of masculinity.As in other Western countries, there is a preoccupation in media commentary with the problems men have adjusting to a changing gender order (Coyle & Morgan-Sykes, 1998).Specifically, connections are made within this discourse between mental health problems in men and rising rates of divorce and changing work patterns, with the rise in feminised jobs in service industries and a decline in traditional manly manual labour for working class men (see Shiner et al., 2009).There is a certain continuity here with the gendered concerns about suicide and social change noted by Kushner (1993) in the 19th century.

Common sense theorising about gender and suicide
There is a traditional association between financial problems and men's suicides (Canetto, 1997;Kushner, 1993).In our sociological autopsy study debt was mentioned by at least one of the inquest witnesses as a relevant factor in 11% (n ¼ 9) of the male suicide cases.This proportion is perhaps rather lower than might have been expected in the light of the traditional importance of money in lay theorising about men's suicides.It should be noted, however, that when we consider the related issue of work problems e related because both money and work are important aspects of the traditionally hegemonic breadwinner model of masculinity e then the proportion was rather higher at 23% (n ¼ 18) of male suicide cases.
To illustrate its currency as a common sense explanation, we see examples of cases where financial problems are cited as relevant to a male suicide by one of the witnesses, even where there is no other evidence to support a connection and the reported views of the deceased on their distress make no reference to money matters.So, for example, in case 18, the father of the deceased (male, age 26, 'Jack') begins his statement thus: 'Jack up to a week before the incident had had financial worries regarding a car he had purchased approximately 8 months earlier.'This suicide was apparently prompted by Jack having been due to meet his ex-girlfriend for a drink and she did not turn up.We do not mean to suggest that the father was wrong to cite financial worries as relevant.In fact we would stress the importance of preserving the complexity of social circumstances.We can, however, conclude from this case that financial troubles are most likely being cited as relevant because of a common sense assumption that this is an understandable reason for a suicide, and perhaps especially a male suicide, given the continuing cultural association between men and work/money.This common sense association is in fact borne out by research evidence which shows an association between unemployment and male suicide (see Chung, 2009).In Stack and Wasserman's (2007) qualitative study of 62 suicide cases in Detroit, most of whom were men, common contextual factors included unemployment and income loss, as well as other contributory economic strain factors including loss of a car and pension plan cutbacks.
Relationship breakdown also emerges from the data set as a common sense justification for suicide in the eyes of inquest witnesses and interestingly this appears to be especially the case for men.This might be seen to challenge historical stereotypes which connected suicide in women with a reaction to failed relationships (Canetto, 1992e1993;Kushner, 1993).We can see from the files that in situ theorising by police officers who present at suicide scenes often includes mention of failed relationships, whether or not this proves to be relevant in the evidence provided by other witnesses.Also, when visiting another local coroner's office (not our main site) we were told by the coroner's secretary that the suicides in her district were 'mostly young boys who've split up with their girlfriends'.In fact the picture for England and Wales shows that suicide rates are higher in mid-life and old age than in youth (Shiner et al., 2009) and this district did not depart from this trend.The secretary's summary of suicide trends revealed a popular preoccupation with the problem of suicides in young men, as shown in media reporting, but there is perhaps a newer focus in the popular imagination on men's responses to relationship breakdowns.It is an association which fits with portrayals of suicide in popular culture.Failed relationships predominate in films featuring men's suicides, for example (Agerbo et al., 2009).
Relationship breakdown coupled with conflicts over children seems to be taken to be an especially understandable reason for suicides in men e understandable perhaps in the context of the popular assumption that the family courts 'favour' women as carers of children and the angry politics of fatherhood, with fathers' rights groups citing suicides in separated fathers as evidence of the supposed cruelty of the judicial system.Case 30 (male aged 32) was a very thin file with little evidence at all.Pretty much all we know about the man is that he had a custody battle with his ex-partner over his children aged 5, 8 and 10 and was concerned about his expartner's care for the children.We were told that he was due in court when he killed himself.This seemed to be an example of a case where the coroner and his staff assumed it should a straightforward 'suicide' verdict because of the custody battle e straightforward enough that little evidence was needed.This single explanation was fairly stark in the evidence from witnesses, including those who knew him well.Again this seems to be a case where an apparent cause 'makes sense' to those lay and professional people who were assessing the death.This popular association of men's suicides and conflicts over children might possibly suggest a shift in the hegemonic model of masculinity in the UK from a narrow focus on breadwinning to a broader conception which encompasses involved fathering, with this role including an actual physical presence of some kind in children's lives and contribution to care as well as finance (Shiner et al., 2009).
In noting the currency of relationship breakdown as a common sense gendered explanation we do not intend to distance ourselves from this theorising.In fact as can be seen below we have built on this connection to devise a typology of gendered responses to relationship breakdown.As Stack's (2000b) review notes, numerous studies have pointed to a connection between divorce and suicide, with marriage offering stronger protection against suicide for men than for women.Our general argument in relation to common sense theorising about suicide is that rather than dismissing it as 'only' common sense we should generally respect the interpretations of social actors who are close to the deceased and we should acknowledge that the stock of knowledge about distress and what might reasonably make someone feel suicidal that inquest witnesses draw on is broadly the same as that which suicidal individuals draw on when experiencing distress.

The construction of evidence about suicides
As well as acknowledging that evidence in case files is based on taken-for-granted assumptions about reasonable cause, we have to consider that the concerns of the living feature in the case files as much as those of the dead (Langer, Scourfield, & Fincham, 2008).Acknowledging this reality emphasises the complexity of evidence about individual cases and the inevitably interpretive dimension of this kind of research.As with psychological autopsy studies, decisions have to be taken by researchers.The data cannot simply be assumed to speak for themselves.For example, although there is plenty of evidence in the files of relationship breakdown being a significant factor in suicides (it is present in 55 of the 100 cases and ostensibly the main trigger to the suicide in 34), there are also some silences.There are cases, for example, where partners deny any tensions but other witnesses such as friends of the deceased make it clear that relationship difficulties were cited by the deceased as amongst the principal triggers for the suicide.In Case 41, there was some indication in the case file of work-related stress e this was a self-employed man in the property business.The police seemed perhaps to be fishing for debt as a possible understandable reason for suicide.In fact the man's secretary told the police her boss was reasonably happy, although he did experience stress and had a drink problem.The best clue to what may have been an important trigger was the word of a friend, who would not give a formal statement, that the girlfriend of the deceased was having an affair and he had only learned this information on the night he died.The girlfriend herself made no mention of any affair.In Case 48, we were clearly told by several witnesses that the context of the suicide was divorce, following domestic violence.The ex-wife provided detailed evidence, including the no doubt difficult admission that on the day before he died, when he drunkenly talked of killing himself, she and her daughter had left him on his own 'as he had used this type of behaviour in the past to seek attention from family members'.What she did not mention was any difficulty he had with her new partner, whereas the fact that he took an overdose on the day she had arranged for him to socialise with her new partner might indicate this had a certain importance.
These two cases once again hint at the complexity of individual cases and the impossibility of reducing suicides to a single cause.They also suggest that witnesses may control the flow of information in the inquest processes.This is hardly surprising in the context of the moral judgements which follow suicides and the connection between morality and responsibility (see, for example, Coyle and MacWhannell (2002) on media discourse).Those who were close to someone who has died by suicide have to negotiate the hovering responsibility (Owens et al., 2008), perhaps minimising their own connection to any distress experienced by the deceased or at least drawing certain lines around their own role.Witnesses' moral positioning and disclosure of evidence are also likely to be affected by the legal context of the inquest.For example they may be less likely to discuss potentially relevant participation in illegal activities such as crime or drug use, or behaviours which carry particular moral judgements (e.g.some sexual practices).Interpreting inquest evidence requires a critical stance on the available sources and a consideration of witnesses' positionality and expectations of audience.

Gender and relationship breakdown e a typology of suicides
We change our theoretical lens at this point, moving constructionism into the background and instead applying a reasonably objectivist understanding to the suicide cases.Whilst it is vital to capture the complexity of cases, there is also within a sociological autopsy such as we have outlined room for variablebased analysis.Given the nature of the evidence in coroners' records, we see any analysis of them as unavoidably interpretivist and therefore we part company with the psychological autopsy tradition which claims to isolate variables (e.g.mental illness) for suicidal individuals post-mortem in an unproblematic way.It is important to acknowledge that judgements have to be made about any social context variables in suicide cases at several points.There is the process of data recording in the coroner's office, which has to filter data out because to note every word in a file would not be feasible.There is the process of coding whole cases according to whether a social factor (e.g.isolation, alcohol dependency) is mentioned by any inquest witness as relevant.There is then further interpretation of how significant a given social factor is for each case.
As explained earlier in the article, any coding of whole cases can be exported into statistical software, allowing for mixed method analysis.We present some of this variable-based analysis in what follows.Another paper we have written (Shiner et al., 2009) presents a statistical overview of the data set and notes some genderand age-related trends.We briefly summarise some of the findings here as a further illustration of how the sociological autopsy method can be applied.Considerably more detail can be found in the published paper.
Problems related to children, as reported by any inquest witness, feature most strongly for women, whereas problems related to work and debt feature most strongly for men, confirming traditional gender narratives.These problems were concentrated in mid-life for both sexes.Relationship breakdown was most common as a main trigger for suicide in those aged 25e54.On the basis of an inductive hunch about gendered reactions to relationship breakdown, one of the researchers returned to the cases coded under this heading (i.e.there was evidence of breakdown or difficulties in an intimate/sexual relationship) to do some further categorisation.This led to two kinds of typologies.The first typology was of the relative importance of relationship breakdown to the suicidal act e as far as it was possible to interpret this from the evidence.The second typology involved only those cases which were identified as having relationship breakdown as the main trigger.These were categorised in terms of the apparent response of the suicidal individual to the relationship problems (or relationship termination in many cases).The categories of suicidal response to relationship breakdown were these: murder/attempted murder punishment dependence sexual jealousy separation from children (see Shiner et al., 2009).It should be noted that these are of course not discrete categories.'Punishment' cases tended to also involve sexual jealousy and/or separation from children, for example.The typology is simply designed to highlight the dominant circumstances, as revealed in the evidence presented to the coroner.It illustrates the difference between the emphasis of a variable-based analysis e albeit a qualitatively-driven one e and a case-based analysis which maintains an emphasis on multi-factorial social context.It illustrates the benefits of qualitative analysis, as an inductive approach to inquest evidence can yield distinctive insights into the social contexts of suicidal acts.
It should be noted that, with the exception of one woman, all these cases ostensibly involved the breakdown of heterosexual relationships.Although relationship breakdown was cited as a relevant social factor at a similar rate for males and females (56 per cent [n ¼ 44] and 52 per cent [n ¼ 11] respectively), it was more likely to be identified as the main trigger for male than female suicides (38 per cent [n ¼ 30] and 19 per cent [n ¼ 4] respectively), which is in keeping with most research to date (see Stack, 2000b).When we quantify the different categories of relationship breakdown trigger, it is more difficult to maintain robust statistical analysis because of the small numbers involved.We can note, however, that if we regard homicidal violence, punishment, sexual jealousy and conflict over children as all indicating different aspects of domestic abuse e and the qualitative analysis would broadly support this interpretation e then 23 suicides out of the sample of 100 could be characterised as indicating some features of abusive behaviour on the part of the deceased.All but one of these 23 individuals was male.

Case-based analysis of suicidal masculinities
We return to case-based analysis at this point to round off our analytic examples with further illustration of the complexity of evidence and understanding that looking at individual suicide cases has to involve.We have chosen to focus on Mark (case 7), a 30-year-old man who killed himself by carbon monoxide poisoning in his car soon after a very serious suicidal act by the same means.Mark's suicide is interesting because the social context seemed to include a toxic combination of spite to his expartner (the mother of his child) and anxiety about his public image.
A psychiatric report noted that Mark 'felt very humiliated, embarrassed and quite upset' about his suicidal act.Mark had no history of psychiatric involvement until this serious suicidal act.His embarrassment may well be understandable in the light of Canetto and Sakinofsky's (1998) insights about gendered cultural scripts of suicidal behaviour.These authors note that 'successful' suicide tends to be masculinised (as decisive and requiring courage) whilst a 'failed' suicidal act is often feminised (as weak and ambivalent).Mark's reaction of embarrassment and humiliation could easily be understood in these terms as shame related to behaviour which is feminised.Mark was generally concerned about how he was seen by others.He had been imprisoned some years before his death for theft from an employer and told his best friend shortly before his death that he felt he had hurt his family.He further said that 'he felt he had built up an image to everyone that he could no longer live up to' (his best friend's statement as drafted by a police officer) (see Canetto, 1995).
This explicit mention of 'image' is unusual, although similar issues are likely to be implicitly present in many male suicides.The idea of a man's 'image', viewed from the perspective of critical men's studies, connects with Connell's (1995) concept of 'hegemonic masculinity'.Although this concept has been criticised as suggesting 'the notion of a fixed (male) structure' (Whitehead, 2002: 94), Connell insists it refers to 'the circulation of models of admired masculine conduct' that do not 'correspond closely to the lives of the majority of men' but express 'ideals, fantasies and desires' (Connell, 2002: 90).Bourdieu (2001) emphasises the important role of honour in masculine domination.He describes honour as 'a system of demands which remains, in many cases, inaccessible' (p.50).Honour, shame and manliness all require public affirmation; they must be 'validated before other men' (p.52).Applying these ideas to a sociological understanding of gendered suicide, Scourfield (2005) has highlighted the tension that can be produced when a gap between aspiration and reality in suicidal men becomes absolutely overt and recognised, rather than submerged or socially embedded, and a life marked by a loss of masculine honour is not seen as worth living.Mark wrote in one of his several suicide notes 'I Just wished that I was normal and Respectable instead of who I am'.
The public account Mark gave for his suicidal feelings rested on his failure to conform to a respectable image and also on the debts he had accrued.Debt is perhaps an expected and even acceptable reason for suicidality in men, given the traditional association between masculinity and work, as noted earlier in this paper.Mark did in fact hint to his best friend that there were 'more things in the pipeline that would become evident' (best friend's statement to police).In fact there was an apparently very significant factor which only became evident from the suicide notes he left.He wrote several notes, one of which was to his ex-partner who was the mother of his child.He specified that he did not want his family to know about this note in case it further upset them.He started the note to his ex-partner by writing 'if you are reading this letter I am no longer here thank God, no more thoughts of what a Spiteful bitch you are' and proceeded to write a note that was designed to cause her intense distress.He wrote 'make no mistakes this is because of you and nothing else' and went on explain how difficult it should be for her to live with herself knowing she was responsible for his death.He carefully listed everyday events and objects that ought to remind her of him and cause her pain.He twice mentioned the potential impact on their son, again making it clear she was responsible for this.In stark contrast, he left a note for the police and any person who found his body, apologising for any 'hassle' and leaving his name, address and details of next of kin.
In the light of his apparent preoccupation with how others saw him and a sense of respectability, we might speculate that the motive of punishing his ex-partner was apparently not mentioned to the best friend (or at least not mentioned in the friend's witness statement) because he was aware of possible moral censure, especially in relation to the impact on his young son.It was possible for Mark to be concerned about avoiding negative public reaction, to the extent of apologising to a stranger who might find his body, and simultaneously to carefully construct a suicide note to cause maximum anguish to his ex-partner.We do not have any data on the history of his relationship with any ex-partners, but we can arguably see in this case a similar dynamic that can be seen in some domestic abuse of a concern with public respectability alongside simultaneous extremes of private abuse.Such extreme denigration of the ex-partner does suggest a personalised misogyny; not necessarily a generalised hatred of women but a hatred of an exwoman partner which is somehow allowable.He did not want to make the hatred public, but putting these thoughts down on paper does perhaps suggest that there is a social or cultural space allowed for such extreme views on an ex-partner.
Case-based analysis such as this example can bring out the individual complexities referred to in the quotation from Redfield Jamison that opened this paper.It is ideally suited to making sense of gendered identities and practices in suicide cases.The sociological autopsy approach allows for case-based insights to be combined with variable-based analysis, as well as for the negotiation of constructionist and objectivist ontologies.It is therefore an integrated approach to the study of suicide.

Conclusion
There are clearly limitations to the sociological autopsy.We have not succeeded as Klinenberg (2002) did in his 'social autopsy' of the Chicago heat wave in locating deaths within the ecology of local areas.This was in part due to considerations of anonymity e choosing from the outset not to identify the location of the study e and also because of limited resources.We would certainly recommend enhancing our method with area-based data in future studies, as long as anonymity could be carefully negotiated.The data in coroners' files on social class and ethnicity were sparse, placing further limitations on our sociological reflections.We did not check the witness statements that were in the files by interviewing significant others as Duneier (2006) recommends and psychological autopsy studies routinely attempt to do.To do this with a large enough sample for robust statistical analysis is a considerable challenge, of course, but we agree it is desirable.All this is to acknowledge what could be seen as the limitations of our approach.In its defence, we might claim that our approach falls somewhere between Klinenburg's macro-level analysis and the ethnographic approach recommended by Duneier.We do have individual-level data which include, in almost most cases, the testimonies of several witnesses who knew the deceased well.We have a reasonably large and robust data set compared to most qualitative research.Our approach can potentially help researchers avoid the pitfalls of the ecological fallacy and over-generalisation from very small numbers of cases.Where we can most justifiably claim to be extending the methodological options for social autopsy is in negotiating dual paradigms.Social autopsy data can be used for study of the social construction of death and also as the basis for reasonably objective conclusions about the social circumstances of individuals in life and death.
In the short term, sociological factors in suicide can perhaps best be documented via research which takes advantage of the wealth of information in coroners' files.There are few large datasets available with individual-level data on suicides and where they do exist (e.g. the US National Mortality Followback Surveys) they contain relatively few sociological variables.Certainly for researchers with limited or no funding, coroners' files, despite their limitations, offer an opportunity for suicide research.Longer term there needs to be wider recognition that qualitative and quantitative approaches to suicide research are complimentary, with each informing the other.Ideally, major research funders will in future invest in studies where contributing factors uncovered in qualitative research could be tested in 'psycho-social' autopsy studies, in which investigators use scales to measure both psychological and sociological factors for individual suicide cases.
We might claim that suicide is in some respects a uniquely fascinating topic.This is in part because the non-suicidal majority who put considerable effort into living find it hard to understand the desire to end life.It is also perhaps because it reminds us of our own psychological and physical frailty.Encountering death, according to Berger (1990: 43), causes us to radically question 'the taken-for-granted "business-as-usual" attitude in which one exists in everyday life'.Suicide is certainly a topic of enduring sociological interest.The epistemological and theoretical debates that Jamison noted in the quotation that opened this article are still thriving.In the article we have presented an approach to researching suicide that we believe draws on both the traditions within sociology which are more obviously allied to science and those which have more obvious affinities with the humanities.We have stated the case for a qualitatively-driven mixed methods sociological autopsy study, which mimics the approach of psychological autopsy studies but has an avowedly sociological purpose, both encompassing the construction of knowledge about suicide cases and also aiming towards reasonably objective judgements about the circumstances of suicidal individuals.Our paper is primarily methodological and does not make any major empirical claims.We have, however, briefly illustrated the application of the sociological autopsy approach with reference to men's gendered identities and gendered practices.The illustrations dealt in particular with the important issues of men's economic life and intimate relationships; issues which are taken up with a more dedicated empirical focus in the other papers in this special issue.