Results: attitudes to voices

How do people evaluate accents when listening to real speakers?
Our survey on accent labels indicated that a long-standing hierarchy of accent prestige is still in place. But do we evaluate an accent in the same way when we hear it spoken? In this study using audio recordings, we found a similar accent hierarchy - but we also demonstrated that respondents are sensitive to context: accent bias exists, but effects may be attenuated when evaluating an actual speaker in a social context.

Overview

Our study examined how listeners around the UK rated job interview candidates speaking in the five main accents of interest. The findings pointed to a number of important patterns of behaviour.

The most prominent of these was the effect that listener age has on evaluations. Listeners who were over 45 show a dispreference for Southern vernacular accents such as Estuary English (EE) and Multicultural London English (MLE), rating them significantly lower than the other accents. Younger listeners, however, make no such distinction and rate all 5 accents approximately equally. 

A second important factor is whether speakers were providing answers to ‘expert’ questions (demonstrating technical knowledge) or to ‘non-expert’ questions (about general skills and behaviour).

As expected, answers to questions that suggested a level of expertise were rated more highly. Yet even when listening to these expert responses, older listeners in the South, and to a lesser degree the Midlands, still rated EE and MLE significantly lower than any of the other accents.

We also find that overall, listeners who have a higher Motivation to Control a Prejudiced Response (MCPR) rate speakers more highly.  This effect trumps any accent bias: for older Southern and Midlands listeners, it is only those listeners with low levels of MCPR who show a dispreference for EE and/or MLE.

Are our attitudes to accents shaped by age?

As shown in Fig. 1 it is clear that older listeners (over 45 years of age) rate Estuary English (EE) and Multicultural English (MLE) significantly lower than the other accents tested. Younger listeners make no such distinction and rate all the accents tested approximately equally.

This result could mean one of two things: it could be the case that attitudes to accents are actually changing, such that older people are biased against Southern non-standard vernaculars but younger people are not. Or, it could be that bias against Southern non-standard varieties emerges later in life; that when we are young, we are more tolerant of (accent) diversity, but become more critical as we get older and enter the workforce. 

The latter pattern is called age-grading and is commonly found in linguistic and psychological research.  Based just on the findings of the nationwide survey, we are unable to choose between these two possible interpretations, but our study of accent labels found the same age-graded pattern not only in our data but also as far back as 50 years ago (in Giles 1970). This points toward age-grading during the lifetime as the most likely explanation.

Fig. 1 - Age-grading

Older listeners (over 45 years of age) rate EE and MLE significantly lower than the other accents tested.

Graph showing age-grading of accents

Does expertise overcome bias?

A second trend we noted related to whether speakers were providing answers to ‘expert’ questions (answers that demonstrated technical knowledge of the law) or to ‘non-expert’ questions (answers about general skills and behaviour).

Answers given to ‘expert’ questions were rated more highly by listeners, as you can see in Fig. 2. 

Yet even when listening to these responses, older listeners in the South, and to a lesser degree the Midlands, still rated Southern vernacular accents (EE and MLE) significantly lower than any of the other accents (see Fig. 3).

This means that expert content can reduce the effects of accent bias to a certain extent, but that for older Southern listeners, Southern non-standard accents are still heard as less favourable overall.

Fig. 2 - Expert vs. Non-expert answers

Answers to ‘expert’ questions were rated more highly than those to ‘non-expert’ questions.

Graph showing difference between non-expert and expert answers

Fig. 3 - Older southern responsdents

Even when listening to expert answers, older southern respondents still rated EE and MLE significantly lower than any of the other accents.

Graph showing older southern respondents accent ratings

How important is a desire not to appear prejudiced?

As shown in Fig. 4, we see a similar pattern for listeners’ reported Motivation to Control a Prejudiced Response (MCPR) – the measure of a psychological characteristic that corresponds to an individual’s desire to be perceived as not acting in a prejudiced fashion. (This is different than measuring whether a person holds biased or prejudiced beliefs; MCPR measures how important it is to someone not to appear prejudiced to others.) 

We find that overall, listeners who have a higher MCPR – that is, a stronger desire not to appear prejudiced – rate speakers more highly. 

This effect trumps any accent bias: for older Southern and Midlands listeners, it is only those listeners with low levels of MCPR who show a dispreference for EE and/or MLE. 

Fig. 4 - High vs. low MCPR

For older southern respondents, it is only those with low MCPR who show a dispreference for EE and MLE.

Conclusion

Overall, the results of the nationwide audio survey show a similar accent hierarchy to that observed in the accent labels survey, though the patterns are somewhat weaker and are specific to particular groups of listeners (e.g. older Southern listeners). 

Together, the two nationwide surveys (accent labels and audio recordings) provide strong evidence of systematic bias against certain accents in England (particularly Southern working-class varieties) and, conversely, a general belief that the national standard accent (RP) is the most prestigious and the most suitable for professional employment.

Nevertheless, the two surveys also indicate that listeners are sensitive to context: we see less bias overall in the audio survey. This seems to show that while evaluative accent ideologies exist, their effects may be attenuated when evaluating an actual speaker in a specific social context (like a job interview).

Our work with lawyers to examine attitudes in the workplace was designed to examine this context effect further.