Attitudes at work
What are the attitudes to accents among employers in elite professions?
We asked 61 legal professionals in the North and South of England to listen to ten mock interview answers, and to rate the candidate’s performance and professional aptitude.
The goal of this survey was to explore in detail whether accent bias interferes with the judging of professional skill. It focuses in on a sector previously described as lacking diversity (law), and uses expert raters in two different UK regions.
We prepared 10 short mock interview answers, varying between ‘good’ and ‘poor’ quality. These were pre-tested in written form with a group of 25 legal professionals otherwise unrelated to the project. Each answer was recorded by two speakers of each of our five accents.
Legal professionals (lawyers and graduate recruiters) were played 10 randomised answers, never hearing the same question or the same speaker twice. After listening they rated each candidate using the same ‘hireability’ criteria as in the national survey. In this experiment, we investigated whether accent interfered with an employer’s ability to hear whether an answer was objectively ‘good’ or ‘poor’.
After rating the speakers, participants were asked to complete the same short questionnaires about their attitudes and exposure to accents and and their beliefs about diversity as were used in the national survey.
In separate work, we also gathered testimonials of personal experience from lawyers in order to access domains of professional life beyond junior hiring, and we also tested the efficacy of a range of anti-bias interventions.
Below you can listen to the sample responses to the interview question; “What is the difference between contract and tort?”
Listen to a sample response in contemporary Received Pronunciation (RP)
Listen to a sample response in Estuary English (EE)
Listen to a sample response in General Northern English (GNE)
The lawyer survey adopted the design of our national survey using audio recordings as a starting point. However, that survey only identifies a simple preference for one accent over another. While this is informative, it cannot tell us the extent to which those preferences interfere with the judging of objective competence.
It could be the case, for example, that preference for a highly prestigious accent (such as RP) override judgments of objective quality, so that even if an answer is objectively “poor”, it is heard as competent when spoken with an RP accent. Or, it could be that professional recruiters can still distinguish between “good” and “poor” responses, even in spite of a general preference for some accents over others.
To address this issue, the study with professionals in the legal sector used mock answers that also varied in their objective level of quality (good or poor). This was done in order to move beyond simple accent preferences and to enable a more precise measurement of the strength and degree of distortion of any accent bias on judgements of job suitability. Including this quality manipulation in our experiment allowed us to distinguish between a number of different potential effects of bias.
Participants: who took part in our survey?
Participants in the lawyer survey were all legal professionals (lawyers and graduate recruiters) employed in major corporate law firms in the UK. The majority of our data was collected in firms in London, York and Leeds, with additional data collected via an online survey. All participants in the lawyer survey had experience with hiring in the UK legal sector.
The majority of participants were recruited through the generous participation of six major international corporate law firms, and the remainder through further individual networks. We worked closely with colleagues in the law schools at Queen Mary University of London and at the University of York to establish contact with law firms and legal professionals.
A total of 61 legal professionals participated in the study. They represent the full range of employees, with an average number of years working in the legal profession of 13.7 years. The participants were closely balanced for gender and included representatives of all major ethnic groups in the UK.
Audio stimuli: what recordings did we use?
“Good” and “poor” quality answers were developed for 10 mock interview questions. As in the national survey, answers were devised by the researchers in cooperation with senior legal professionals. Answers were pre-tested in written format with a group of 20 legal professionals to ensure that they represented objectively “good” and “poor” responses to the questions, and were all in a formal register using standard grammar.
The 20 answer texts (2 quality levels x 10 questions) were recorded by the same 10 speakers (2 native speakers per accent) that were used in the national survey, resulting in 200 recordings. The content of the answers was sufficiently varied so as to distract listeners from focussing solely on accent differences.
Procedure: how was the survey carried out?
The procedure was identical in the North and the South. Each participant heard a selection of 10 mock answers, pseudo-randomly sampled from the 200 recordings . All participants heard each of the five accents exactly twice, and they heard no mock answer more than once and no speaker more than once. This created a natural-sounding sample of ten different extracts from different job interviews. The answer text and quality level was randomised.
After hearing each recording, participants were asked to rate the candidate’s overall performance, knowledge, suitability, and hireability, as in the national survey. They used a 10-point Likert scale to answer the following questions:
Once participants finished rating the answers, they provided information about their personal background (including gender, ethnicity, age, region of origin, highest level of education, occupation, English accent, languages spoken). They then completed a short questionnaire about their exposure to different UK accents, the diversity of their own social networks, their beliefs about bias in Britain, and a set of psychological measures such as their level of concern about being perceived as prejudiced.
Sharing our results with the public
We provide a short summary of our main results of this study on this website. Detailed results will appear in scholarly publications. Our project is also creating detailed training materials and audio samples that can be used by a range of beneficiaries, including university students in law or other professional career paths, HR professionals, law firms, and the general public.
These materials can help people become familiar with how accent bias works, what current attitudes to accent in the UK look like, how legal professionals of different seniorities perform in a hiring task, and what the consequences of unconscious accent bias can be. We provide hands-on audio exercises to give users a real-life experience of the challenge of ignoring accent preferences and ensuring fair hiring practices.