Murderer, mother, slave, or skivvy: XML annotation to enable social actor analysis in a small corpus of English sentencing remarks for women who kill

Although women make up a minority of homicide convictions in England and Wales, this is a statistically and socially significant minority. Their comparatively minor involvement in such criminal behaviour has resulted in a comparatively small (but growing) body of research engaging with such cases. Much of emerging scholarly work argues that these women are constructed in relation to their deviance from appropriate femininity and are ultimately constructed as ‘mad’, ‘bad’ or ‘victims’, depending on the circumstances surrounding their cases and whether they successfully use any legal defences (see for example: Seal, 2010; Weare, 2013). This paper utilises corpus linguistic tools on a specialised corpus of courtroom language to explore constructions of women who kill. This group might be considered triply deviant, for: 1) being a small minority group, in 2) committing a very serious crime, which 3) goes against traditional conceptions of appropriate feminine behaviour. Though socio-cultural conceptions of gender should not contribute to legal proceedings, we will demonstrate how judge’s sentencing remarks routinely stereotype women who kill, further ‘othering’ them, removing their agency, and seemingly impacting the sentencing outcome. This paper illustrates some of the latest work being undertaken at the legal- linguistic interface. Using methods from corpus linguistics, we analysed recurrent (statistically significant) patterns in a collection of sentencing remarks to explore how female defendants are routinely constructed. Analysis is based on a custom- collected corpus of English Crown Courts sentencing remarks, 2012-2015, where a jury has found a woman guilty of unlawful killing. At present, the corpus contains 17 sentencing remarks comprising over 50,000 tokens total. The corpus is analysed in SketchEngine (https://www.sketchengine.co.uk), making use of XML compatibility and WordSketch features. The small size of the corpus reflects the comparatively small number of women convicted of a homicide offence, but has allowed for full reading of the texts and manual XML annotation (Hardie, 2014) of salient features therein. This approach allowed us to search for all references to women who kill using single query, and to view all instances in context. Based on this, we could indicate frequency of naming strategies and calculate collocates of women who kill, which informed our discussion about the social construction of these participants by judges in sentencing remarks. Women who kill are referred to 1,809 times in the corpus. Several findings emerged from our initial analysis. Frequency analysis of the textual contents of the XML tags provides insight into common referencing strategies. These women were most frequently referred to directly by the judge using the second person pronoun, but additional functionalization (van Leeuwen, 2008) and nomination strategies construed these homicidal women in terms of familial position, e.g. “you, Keanu’s mother”. The use of such nomination strategies construing these women in terms of familial position also reflects the perceived gendered deviance of women who kill, who have both departed from, and undermined, their roles as the care-givers and nurturers within society. Annotating all references to a class of social actor also allows for collocation on this group (including even low-frequency instances). Collocations—with concordance analysis—have initially showed a lack of agency associated with women who kill. Verbal processes and mental processes of emotion associate most strongly with women who kill, while active, material processes are much less frequent and are of lower statistical significance. As a result, rather than acknowledging the choice of these women in relation to their homicidal behaviours (which would, in turn, actively challenge existing gender discourse), excuses and explanations are typically invoked when exploring their offending, and the women offenders are routinely constructed as being controlled by mental illness or deviant sexual impulses, or victimised by others. We sum up by reflecting on the strengths and limitations of working with a very small, extremely specialised corpus of legal language. This work contributes to both legal and linguistic studies. From a corpus linguistic perspective, close reading and manual annotation of features is increasingly rare. From a legal perspective, applying corpus linguistic approaches to legal texts allows for a previously unexplored depth of analysis of these texts to be undertaken. In turn, this allows us to further develop understandings of the discourses and narratives that emerge and thus take steps to combat those that are problematic and damaging