It was after attending an International Women’s Day event in 2019, where I heard journalist and feminist Jane Gilmore speak about her project ‘FixedIt’ – an initiative that corrects victim-blaming headlines in news articles on violence against women (VAW) –  that I knew I wanted to focus my dissertation on how the Australian media reports on VAW. After noticing differences in media coverage (and consequent public outrage) for some cases compared to others (thanks to Destroy The Joint’s ‘Counting Dead Women Australia’), I decided I also wanted to examine whether there are differences in how domestic violence (DV) is being portrayed compared to non-domestic violence (NDV). Considering the influence the media has in shaping people’s attitudes, behaviours and understandings of the issue, I was particularly interested in examining how blame and responsibility is encoded in articles. Since existing research on the coverage of VAW in Australian media has largely been undertaken in the fields of media communications and feminist studies, I was keen to explore how linguistics might be able to contribute to the conversation.

With this being my first empirical research study in linguistics, my supervisor Monika Bednarek provided great insight into how different linguistic tools might help me to explore the questions I had posed. I came to see how combining corpus methodologies with critical discourse analysis (CDA) was the perfect method for my study. It reduced the risk of being accused of ‘cherry picking’ articles in order to confirm preconceived ideas (a common criticism of CDA), and also allowed me to examine initial findings against a larger set of data. After undertaking extensive research into recent cases of VAW in Australia, I decided on six cases to include in the study – three domestic violence cases and three non-domestic violence cases. I collected data from two Sydney newspapers – The Sydney Morning Herald and The Daily Telegraph – and created a small, highly specialised corpus of 54 texts for corpus linguistic analysis and then used a subset of these articles for computer-assisted ‘manual’ text analysis (with the help of UAM corpus tool).

I undertook the ‘manual’ text analysis on eight articles first, focussing on social actor and appraisal analysis to examine how the victim, perpetrator, and the act of violence (AOV) are represented in articles of DV vs. NDV. The text analysis uncovered some interesting insights – highlighting first how different lexis was being used to describe the act of violence in DV and NDV cases – e.g. DV articles were more likely to include graphic representation of the AOV compared to NDV articles, using words such as stabbed, stuffed, strangled and suffocated. In addition, the text analysis found that DV articles were more likely to describe the AOV using neutral words like death or dead, compared to NDV articles which were more likely to describe it with more emotive language, either as a murder, attack or assault.

Amongst other findings, which you can read about in more detail in my dissertation, the text analysis also uncovered differences in the naming practices used for perpetrators of DV compared to NDV, with DV perpetrators more likely to be referred to in personalised terms that reflect their social status, such as boyfriend, dentist, doctor, lover, compared to NDV perpetrators who were more likely to be labelled as a killer.

The importance of utilising corpus analysis as part of this study became apparent when I tested selected findings from the text analysis in the larger corpus. Primarily using keyword analysis, I found that while findings related to the graphic and sensationalist reporting of DV were supported in the larger corpus, the finding that NDV cases were more likely to describe the AOV as a murder, attack or assault, was not. In other words, this particular result needs further research. However, keyword analysis also confirmed initial findings from the text analysis that DV perpetrators were more likely to be referred to in personalised terms.

Overall, using the emergent technique of corpus-based CDA for this study was fantastic, as it allowed me to approach the topic using a different methodology in order to contribute something else to the research that currently exists. One of the benefits of this combined approach was that each method uncovered findings that the other one might not have on its own. The corpus analysis also allowed me to cross-check initial findings from the text analysis in a larger dataset, which strengthened the validity of a number of findings and indeed, the overall study.

You can read more about corpus-based CDA and the research I undertook in my dissertation. You can contact me via LinkedIn.