This post aims to bring some application value to the accent strand newly introduced to A’Level English Language. Accent and dialect differences are of course interesting in their own right. However, they can also be useful to real-life applications. Here, I’m going to shed light on just one of these: forensic speech science.
Forensic speech practitioners analyse recordings which might feature as evidence in legal casework. Often, it’ll be telephone calls and we want to answer various questions about the speaker or what was said. One task analysts might be asked to do is called ‘speaker profiling’. Speaker profiling is the task of extracting various identifying information about the speaker in the recording. We could think about this in the context of a ransom telephone call, for example, where we don’t have any information about the speaker, but we want to narrow down the pool of possibilities to assist investigative teams. Information like where the speaker is from, or what speech community he/she belongs to, could be really useful to a cause. A thorough analysis of the speaker’s accent can help us to do this. Outlined below are a couple of real-life cases where speaker profiling/accent analysis played a part.
Case 1: The Yorkshire Ripper
The most famous case involving forensic speaker profiling dates back to the late 1970s - The Yorkshire Ripper case. Around this time, and over a the course of a few years, a serial killer was at large, brutally murdering women across Yorkshire. The lead investigator for the case, George Oldfield, received a recorded message from a male claiming to be the Yorkshire Ripper. Below is what the speaker in the recording said:
I’m Jack. I see you have no luck catching me. I have the greatest respect for you, George, but, Lord, you are no nearer catching me now than four years ago when I started. I reckon your boys are letting you down, George. They can’t be much good, can they? The only time they came near catching me was a few months back in Chapeltown when I was disturbed. Even then it was a uniformed copper, not a detective. I warned you in March that I’d strike again. Sorry it wasn’t Bradford, I did promise you that I couldn’t get there. I’m not quite sure when I’ll strike again but it will be definitely some time this year, maybe September or October, even sooner if I get the chance. I’m not sure where. Maybe Manchester, I like it there, there’s plenty of them knocking about. They never learn, do they, George? I bet you’ve warned them, but they never listen. At the rate I’m going, I should be in the book of records. I think it’s eleven up to now, isn’t it? Well, I’ll keep on going for quite a while yet. I can’t see myself being nicked just yet. Even if you do get near, I’ll probably top myself first. Well, it’s been nice chatting to you, George. Yours, Jack the Ripper. No good looking for fingerprints, you should know by now it’s clean as a whistle. See you soon. Bye. Hope you like the catchy tune at the end. Ha Ha.
This is when Stanley Ellis, a leading dialectologist, was brought in to lend a hand with some expert analysis. With the belief that they had a recording of the perpetrator, they thought that they could identify him. Stanley Ellis’s spanning experience in dialectology and fieldwork equipped him to be able to make an initial broad diagnosis of the speaker’s accent, saying that the speaker sounded like he was from the general Sunderland area in the North-East of England. Using various cues from the recording, Ellis was able to pinpoint, to a finer degree, where he believed the speaker in the recording was from. In his account of the case, here are some cues Ellis used to do this (remember that these were of relevance to these particular varieties back in the 1970s - accent features may have changed since then):
- The vowel quality of the pronoun ‘I’ suggested that Ellis could perhaps eliminate Tyneside or North Yorkshire as possible areas. If the speaker were from Tyneside or Yorkshire, we could perhaps expect an elongated version of the vowel we find in ‘cat’.
- The word ‘strike’ was another useful clue. The speaker’s pronunciation of the vowel in this word was closer to what we would expect in, say, Received Pronunciation, than what we might expect from the spoken variety in the north of County Durham. Typically, speakers in this area would produce a similar vowel to what we might hear in RP ‘steak’ (roughly speaking). This observation therefore meant that the speaker was unlikely to be from north County Durham.
- Also of note was the fact that the speaker in our mysterious recording h-dropped (so, not pronouncing the /h/ in the words ‘have’ and ‘hope’). Having researched this area, Ellis was able to suggest that this means that the speaker is unlikely to be from areas north of the River Wear, where h-dropping is less common.
All kinds of these sorts of cues came together, along with further data collection from this part of the country, to home in on two possible areas: Southwick or Castletown. It is important to note that these kinds of analyses only offer an indication, rather than ground truth results. Forensic analysts present their conclusions in terms of likelihoods.
Based on the outcomes of Ellis’s analysis, police investigation efforts targeted Southwick and Castletown, but there was no luck in identifying a specific individual. However, in 1981, police arrested Peter Sutcliffe, a lorry driver from Bradford (with a Bradford accent), who, it turned out, was responsible for the murders. The tape recording was a hoax, and it wasn’t revealed who the hoaxer was until 2006 - John Humble. Because of advances in techniques, forensic scientists were able to find a DNA match that police had stored from a minor incident Humble was involved in in 1991. It turned out that Humble was indeed from near the areas Ellis had identified. Unfortunately, however, the distraction this hoax created at the time of the Yorkshire Ripper investigation meant that the Yorkshire Ripper was able to go on to murder three more women.
Case 2: Mysterious Bomb Threats to Mr HOW
This case is about 40 year-old Richard Carl who lived 12 miles away from Philadelphia in the US. This case brings together elements of speaker profiling, as well as more specific speaker comparison elements. His wife had been laid off from her job at a company called Mr HOW. Richard Carl called the company and spoke to the supervisor to express that he thought his wife had been unfairly treated. Within a short period of time after this phone call, four phone calls were made to the local police and fire departments claiming that there were bombs and fires at Mr HOW. Richard Carl was accused of making these obscene phone calls, and at this point, Sharon Ash was brought in to make an analysis of the recorded threatening calls with Richard Carl’s speech.
Having closely analysed Carl’s vowels, Ash could confidently express that Carl showed the details of a typical speaker of Philadelphia English. Ash could then analyse the speech of the bomb threat caller and compare the two speakers’ pronunciations. One example of the sort of features Ash looked at was the vowel in ‘gonna’. Ash was able to compare the pronunciation of the first vowel in this word for both the bomb threat caller and Carl. She spotted that while Carl’s vowel matched with the vowels you would find in ‘on’ or ‘off’, the bomb threat caller produced something more like ‘gunna’.
This analysis contributed to Carl’s overall case, and he was acquitted (as a result of a combination of factors).
Practical Challenges
For both Case 1 and Case 2, the accents involved had been previously studied and documented in academic research. We don’t always have access to the sociolinguistic expertise to indicate the specific featural diagnostics which point us towards an overall accent label. For example, in a case, we might have accent varieties which have never been visited in academic research before. Forensic casework, in its nature, is very unpredictable, and caseworkers could be asked to work on anything. As well as this, we know that accents and dialects change, and they change rapidly. This means that our documentation of varieties becomes outdated, and therefore invalid, quite quickly.
My own research aims to make improvements to the way forensic practitioners conduct the speaker profiling task. I have developed the Y-ACCDIST system. Y-ACCDIST is a software tool which can be trained on a database of accents, to then identify what the systematic differences are between these varieties. Given a speech sample of interest, Y-ACCDIST will then process and classify the unknown speaker into one of the previously trained accents. It doesn’t work 100% of the time (I’m not completely deluded), but technological developments are (hopefully) being made in the right direction. The intention is not to replace the forensic analyst in these speaker profiling tasks, but to have the analyst and Y-ACCDIST work in conjunction with one another.
When attitudes to accent might matter
The cases above illustrate how, by being analytical about speakers’ accents, we can perhaps provide an evidential contribution. However, our attitudes to accents can also be significant in a legal context. James Tompkinson and Katherine Weinberg at the University of York are currently researching into this nook of the field. James is looking into listeners perceiving threats. More specifically, he’s looking into how the speaker’s accent affects how threatening the listener finds certain utterances. For example, does a speaker with a Cockney accent come across as more threatening than a speaker of RP? Developing our understanding of this could be valuable to court cases where threats are involved.
Katherine has been looking at threatening speech/language in the context of Anglo Americans’ perceptions of African Americans. Recent relevant cases might include the shooting of Walter Scott, an African American who was shot by law enforcement. It was claimed the the officer responsible felt threatened. However, video footage did not indicate that the victim was threatening the officer in any way. Removing physical appearances from the equation, are there elements of African American speech that Anglo Americans perceive as threatening? Katherine has been analysing phonetic, lexical and grammatical aspects of African American speech and what listeners may associate with these.
With any luck, this post has provided some insight into the murky world of forensic speech science, with particular attention paid to accent analysis. There are many more types of task which forensic analysts have to contend with, but I think this post offers some justification to studying accents. They’re not just interesting, but studying them can also be useful.
