The influence of familiarity and attitudes on
the intelligibility of native English accents
Oosterhuis, J.N
s2061651
Master thesis European Linguistics 10-01-2020
University of Groningen
Supervisor 1: dr. C.S. Gooskens, University of Groningen Supervisor 2: dr. R.M. Knooihuizen, University of Groningen
Abstract
In this study, two factors that could influence the degree of intelligibility of native English accents for non-native speakers of English were examined: familiarity with and attitudes towards native English accents. In the first part of the experiment, twenty five Dutch students, who were non-native speakers of English, were presented a list of simple true/false statements, uttered by five native English speakers with different accents (American, Australian, British, Canadian and Irish). The participants were asked to identify the truth value of each statement as quickly as possible. Reaction times were measured to indicate the degree of intelligibility. In the second part, participants were asked to give scores for different attitude categories for each native English speaker and to try identifying the correct English accent. In the consent form, participants were then asked for the amount of daily contact they had with the different native English accents. Response time data indicated that differences in the degree of intelligibility occur. The most important factor for this difference is familiarity with the accent. No clear evidence was found that attitudes and degree of intelligibility are related. The results are discussed to indicate the importance of the examined factors for the perception of different English accents and the effects of these factors for teaching English as a second language.
Table of contents Introduction ... 3 Background ... 4 Method ... 9 Results ... 13 Discussion ... 20 Conclusion ... 24 References ... 26 Appendix ... 28
Introduction
“A language is a dialect with an army and a navy”. This statement was once used by Max Weinreich to make a distinction between a language and a dialect. In an article by Harbeck (2016), this statement was used to poetically explain why a language changes. Separation (e.g. geographical distance, political independence and cultural divergence) of the original language makes it possible to create a changed version of the original language and also to create different accents. Imagine someone having a British accent, for instance Received Pronunciation; when this person would leave Great Britain for a year to study abroad in America, one would notice that his accent will have slightly changed after a year.
The English language has spread around the globe over centuries and, as a result, many different accents of English were created (e.g. British English, American English, Australian English, Canadian English, and Irish English). It is these five accents of English, which were first spread around the world1, forming the inner circle of Kachru (1992) and whose countries are now native English speaking countries. Kachru (1992) formed three circles of English: the inner circle, the outer circle and the expanding circle. The inner circle was formed by the first emigration of British people. The outer circle was created when countries from the inner circle, especially Great Britain, were expanding their influence to Africa and Asia. Countries in the outer circle (e.g. India, Nigeria or Philippines) do not have English as native tongue, but it is used as a lingua franca. The last circle is called ‘the expanding circle’. It contains countries (e.g. China, European countries or Russia) where English is used as a language of international communication. In contrast to the inner and outer circle, imperialistic factors are not the basis in this expanding circle.
One of the countries in the expanding circle is The Netherlands. According to the Education First English Proficiency Index (EF EPI) (2018), Dutch people are the second best worldwide non-native speakers of English (behind Sweden) and, since the start of the EF EPI in 2011, always found in the top three best non-native English speaking countries. Although the proficiency of English is very high in The Netherlands, little is known about differences in intelligibility of different native English accents. It might be the case that Dutch people are more proficient in the British or American accent than, for instance, the Australian accent, because English television, movies or series are mostly in these two accents. This assumption suggests that the more exposure one has to a particular (English) accent, the easier it is to understand (Gass & Varonis, 1984). This study will examine possible differences in the
1
New Zealand English, South African English and some Caribbean English accents are also in the inner circle of Kachru, but will not be included in this study.
intelligibility of different English accents by testing Dutch students, using a reaction time test and asking for exposure to the different English accents. Furthermore, attitude scores for the different English accents are collected by using a survey. Attitude scores are considered to investigate if there is a relationship between differences in attitudes and intelligibility. For instance, one might prefer to listen to a British accent over an Irish accent, rating the British accent as more competent or possessing more status (Milroy & McClenaghan, 1977) and therefore make a greater effort to understand the British accent. In the following chapter, background information about intelligibility, familiarity and attitudes will be provided to give an overview of research that has already focused on these topics.
Background
Intelligibility
Intelligibility might roughly be defined as the extent to which a listener understands the message of the speaker. However, researchers have not always measured the term ‘intelligibility’ in the same way. For instance, the number of correct paraphrases (Brodkey, 1972), the number of words transcribed correctly (Lane, 1963) and the rating of intelligibility on a Likert scale in a survey (Fayer & Krasinski, 1987; Palmer, 1976) have all accounted for the level of intelligibility.
Previous research has mainly put the focus on differences in intelligibility of native English speakers compared to second language learners of English. However, no research has yet focused on the intelligibility of different native accents of English (e.g. American English, Australian English, British English, Canadian English and Irish English). Two of the three ways, mentioned above, to compute intelligibility are considered in this study. First, accuracy scores are measured on both correct responses (true/false) and correct transcriptions of a true/false task with simple statements, recorded by different native speakers of English with an accent typical for their country (e.g. American, Australian, British). Second, listeners are asked to rate intelligibility on a 5-point scale. In addition to the first two ways, the degree of intelligibility is measured as to the extent to which a listener understands the message of the speaker in the form of reaction times on binary (true/false) statements.
Familiarity
A factor that might play a role for the intelligibility of and attitudes towards different native English accents is the familiarity with the particular native accents. Previous research has
mainly focused on differences in the intelligibility and degree of accentedness of L2 speakers of English, but none have yet examined differences in the relationship of familiarity and intelligibility of native English accents for non-native speakers of English. However, results of previous research are described below to outline the framework in which familiarity with and intelligibility of non-native speech is examined, so that results of this study can be compared with results of previous research.
A research of Munro and Derwing (1994) gives an insight in the relationship between recognition of a non-native accent and the familiarity with this accent. They were interested in the difference between narrative speech and extemporaneous speech when measuring the familiarity with a non-native English accent. Native speakers of English had to give scores for the degree of accentedness of Mandarin L2 speakers of English and a control group, consisting of native speakers of English from Canada. They found that there was no difference in the perception of foreign accentedness in material that was read by L2 speakers or extemporaneous speech by that same group. More importantly, they did obtain findings that non-native accents are more often recognized as non-native, when the familiarity with the non-native accent was greater.
This finding is supported by Gass and Varonis (1984). They were interested in the intelligibility of non-native speech in native speakers and what factors determined the comprehension. The variable that most influenced the comprehension among native speakers was familiarity with the topic of discourse. Furthermore, they found that exposure to not only non-native speech in general, but also to a particular non-native speech, lead to a greater comprehension. This means that if one knows certain characteristics of a non-native accent, the accent will be judged as non-native more often. These findings suggest that the same might happen with the identification of different native English accents, i.e. that a higher extent of familiarity to the different native English accents will result in a higher percentage of correct identification, when measuring intelligibility by means of accuracy scores.
Although familiarity with non-native accents play a role in the comprehension of native speakers, a study by Munro and Derwing (1995), however, revealed that this not automatically means that a sentence produced by a non-native speaker is less intelligible, when accuracy scores are considered. In their study, native speakers of English listened to English sentences produced by native English speakers as well as native Mandarin speakers. Listeners were asked to judge statements as true or false in a sentence verification task. They found that sentences produced by the Mandarin speakers resulted in longer processing times, but did not account for differences in accuracy scores with the native English produced
sentences. This means that it took longer for listeners to evaluate the message of the non-native speaker, but still both the accented and unaccented utterances were perfectly intelligible.
The same findings are reported in Munro and Derwing (1999) when testing extemporaneous English speech of native Mandarin speakers. The Mandarin speakers were asked to describe some illustrations, without the possibility to prepare the story. Then, native English speakers were asked, in a first session, to give scores for intelligibility on each non-native spoken description of the illustrations, and, in a second session, to rate the degree of accentedness of each Mandarin speaker. Although many speech samples were rated as highly accented, the transcriptions revealed that these samples were perfectly intelligible. Taking Munro and Derwing (1995) and (1999) together, it can be concluded that different accents might be perfectly intelligible, but differences in processing times (i.e. the time it cost to process the meaning of the sentence) might still occur, which means there can be differences in the degree of intelligibility of an accent.
Attitudes
This study aims to get an insight in the role of attitudes for the intelligibility of different native English accents as well. Schüppert, Hilton and Gooskens (2015) have investigated the relationship between intelligibility and attitudes for Danish and Swedish children and adolescents. They found that Danish children held little more positive attitudes towards Swedish than vice versa, and their word recognition of Swedish was significantly better than their peers. However, the correlation between attitudes and intelligibility was low, indicating that the variables do not really influence each other. Although there is almost no relationship between intelligibility and attitudes in a setting with two linguistic related languages, a study by Boets and De Schutter (1977) reported that there is a relationship between intelligibility and the appreciation in different accents of the same language. In this study, inhabitants of a Belgium village judged geologically different Dutch accents of Flanders on their pleasantness and were also tested on intelligibility. They found that the more intelligible an accent is, the more it was appreciated and vice versa.
Two hypotheses are put forward by Giles et al. (1974) and Giles et al. (1979): the inherent
value hypothesis and the imposed norm hypothesis. The inherent value hypothesis argues that
a dialect or accent has certain aspects (e.g. harmonious speech) that are always equally positive or negative evaluated by people, whether they know the characteristics or social context of the accent or not. The imposed norm hypothesis, on the other hand, argues that a
dialect or accent is not inherently better than another, but that cultural norms play a role in what people judge as the most pleasing or most sophisticated dialect or accent. Support for the latter hypothesis is found in a study by Giles at al. (1974). In this study, English adolescents with no experience of the Greek language were tested on two varieties of Greek accents: the accent with most prestige, Athenian, and the accent which sounds less intelligent and sophisticated to Greek people, Cretan. Results revealed that subjects rated both varieties equally in terms of beautiness, prestige, etc. Similar results were found when testing subjects from Wales with no knowledge of French on different accents of French (Giles et al., 1979). These findings are in favor of the imposed norm hypothesis, i.e. that cultural factors do play a role in rating an accent as more pleasing and sophisticated.
However, in a study by Ladegaard (1998), results in contrast with the studies supporting the imposed norm hypothesis and in favor of the inherent value hypothesis were presented. Ladegaard (1998) tested Danish participants on speech samples of five different English accents (American, Australian, Cockney, Received Pronunciation and Scottish). Results of a quantitative analysis on attitudes were similar to stereotypes of English accents found in previous studies (see also Giles, 1970; Giles and Powesland, 1975; Stewart, Ryan and Giles, 1985), although the majority of the Danish participants could not identify the origin of the accent.
More arguments against the imposed norm hypothesis and in line with Boets and De Schutter (1977) are presented by Van Bezooijen (1994), who tested attitudes of children and adults towards different accents of Dutch. Van Bezooijen (1994) concluded that rural accents were rated as least attractive, followed by the urban accent and finally Standard Dutch as most attractive in all age groups, despite the fact that the group of children did not have the social geographical knowledge to discriminate between the accents. Van Bezooijen (1994) assumes that children were able to determine differences in beauty in terms of intelligibility or broadness of the accent. On the other hand, and this finding is supporting the imposed norm hypothesis again, adults in the study of Van Bezooijen (1994) were perfectly able to assign an accent to its geographical location and also rated the accents in the same order as the children (i.e. rural accents - urban accent - Standard Dutch). In this study, the imposed norm hypothesis will be examined in terms of the relationship between familiarity with and attitudes towards the accent. In other words, when participants can discriminate between the native English accent and show more positive attitudes towards the accents where they have had more contact with (i.e. which accent is more standard to them), the imposed norm hypothesis is supported.
Research question and hypothesis
This study first tries to get an insight in the degree of intelligibility of different native English accents. When a difference in degree of intelligibility is found, the relationship between intelligibility and familiarity with, and intelligibility and attitudes towards these accents are considered. In terms of familiarity, it is expected that the Dutch participants in this study will report to have more contact with the American and British accent, compared to the Australian, Canadian and Irish accent, because television, movies and series contain mostly the American or British accent. When linking this to the imposed norm hypothesis, the accents reported of having the most contact are considered as the more standard native English accents for this group of participants and therefore more positively evaluated.
In the present study, the first part of the experiment consists of a sentence verification task (based on the experiment of Munro and Derwing (1995)), in which Dutch non-native speakers of English have to judge sentences, recorded by different native speakers of English (American, Australian, British, Canadian and Irish), as true or false. Processing times are conducted to measure the degree of intelligibility. Participants will also be asked to write down the sentences they heard. Correct transcriptions indicate whether the sentences are, in the end, perfectly intelligible for the participants. In the second part of the experiment, the accents of each native English speaker will be judged by Dutch students on different attitudes by using a survey. Finally, the participants will receive a consent form, found in Appendix 3, in which they are asked to give a score from 1 (no contact) to 5 (contact on a daily basis) for the amount of contact they have with the five different native English accents used in this study. Participants were not told that these five accents were used in the study, until they filled in the consent form.
The main question then asked in this study is whether (1) there is a difference in the degree of intelligibility of different native English accents for non-native speakers of English and, if so, whether (2a) this is influenced by the familiarity with the accents or (2b) attitudes towards the accents. Results of degree of intelligibility, familiarity and attitude scores are conducted and then put into a correlation analysis to check for possible relationships. In this study, the following hypotheses are assumed:
(1) A difference in degree of intelligibility measured by reaction times occurs. (2) This difference is influenced by:
a) the familiarity with the native English accent (i.e. the more familiar an accent is, the faster the response times will be or/and
b) the attitudes one has toward the native English accent (i.e. the more positive scores an accent will receive, the more intelligible an accent is).
In addition to hypothesis (1), no differences in intelligibility are expected when measuring accuracy scores on the true/false task combined with the transcriptions, since all accents are native English accents. When considering hypotheses (2a) and (2b), it is expected that participants will report to have more contact with the British and American accent. In that case, following the imposed norm hypothesis and the results of Van Bezooijen (1994), it is hypothesized that:
(2)
c) British and American will receive higher attitude scores, as they are considered to be the more standard accent for the group of participants.
Degree of accentedness of the speakers might also be a factor in the degree of intelligibility for each native English accent, but it is beyond the scope of this study to examine this factor. In the following chapter, the material, participants and procedure are explained.
Method Material
Speakers
Five native male speakers of English between 21 and 29 were selected for this study. They all lived in one of the traditional English speaking countries in the inner circle of Kachru (1992) (Great Britain, United States of America, Australia, Ireland and Canada). They were asked to give a score on a scale from 1 to 5 for the typicality of their accent for their country (i.e. 1 is very not standard, 5 is very standard). Four of them scored themselves a 5, which means their accent was very typical or standard for their country, and one scored himself a 4. The speakers were all bachelor or master students.
Stimuli
A total of 30 statements were designed by the researcher. The mean length of the statements was 5.9 words, with a range from five to eight words. Each statement consisted of high frequent lexical items, which are all listed in the 5000 most frequent words by Davies (2008), except for two well-known common nouns (America and Euro in the statement In America,
the currency is the Euro). The truth/false-value of each statement lies within the last word of
each sentence. The latter is an important factor when measuring reaction times, as one could already identify the truth/false-value of a statement, before the sentence ends if it does not lie on the last word. The statements were pretested to ensure that none of them was potentially difficult to understand or ambiguous. Five native Dutch students, who did not participate in the actual experiment, were presented the statements during a pretest. They had to judge the statements as either true or false. If their response was not in line with the expected response, the statement would be excluded from the test. Based on their judgements and comments, four true and five false statements were excluded, because they did not run well or tended to be either ambiguous or incorrect (e.g. the statement Milk is a product of dogs was judged as true two out of five times, because, besides cows, dogs can produce milk too). Although one statement (Cancer is horrible disease) was judged five times as true during the pretest, the sentence was left out in the study, because it contains a widely accepted opinion, instead of it being a fact. This resulted in a total of twenty statements with an equal number of true and false sentences. In Appendix 1, the list of statements used after exclusion, is presented.
Recordings
The speakers were asked to read and record a list of twenty statements as naturally as possible. They were instructed to read the full list of statements attentively before recording them. Recordings were made in a silent room, to minimize the effect of any background noise. Each statement was recorded twice, to ensure that factors as, for example, hesitation could be excluded from the stimuli presented to the listeners. The recordings of the twenty statements were then split into separate recordings, using Adobe Audition CC 2019, for each of the stimuli at 20 kHz (with a 16-bit resolution), so they could be used in the test. For each individual statement, the recording with the best sound quality out of the two was used. The quality was depending on loudness, the number of background noise (e.g. breathing before or after the statement), the clarity of the speaker's voice and pitch (e.g. one full recording was eliminated due to the fact that, at the end of each statement, the speaker had an unnatural pitch rise. The speaker then made two new recordings). All individual recordings were
normalized to -4.4dB, so that the volume was distributed equally across each audio file. Finally, each recording was cut off after the last sound of the sentence, so that response times measured after each sentence were as accurate as possible.
Utterance durations
Before it was possible to analyze the response time data, it was important to determine whether the duration of the statements was equal across the different accents. The mean sentence duration for the American (M = 1750 (in ms), sd = 293), the Australian (M = 2198,
sd = 352), the British (M = 1960, sd = 265), the Canadian (M = 2058, sd = 320) and the Irish
accent (M = 2008, sd = 261) were put into a one-way ANOVA. A significant shorter duration was found for the American accent compared to the Australian (p < 0.001) and the Canadian (p < 0.05) accent. The Tukey HSD post-hoc test showed that other comparisons between accents failed to reach a significant difference in length of the audio. Because of the difference in utterance duration of the American statements compared to the Australian and Canadian accent, two ways of analyzing the reaction times will be presented below; one analysis without the recordings of the speakers and one including the recordings. For the first analysis, where the recording is excluded, the duration of each individual recording was subtracted from the reaction times, measured by E-prime. Therefore, the reaction times reported in the results without the audio of the speakers’ recordings beneath are measurements of the time between the end of the recording and the moment the participant pressed the corresponding button. Both ways of analyzing are considered, since utterance durations between other accents do not occur and might result in different reaction times, as Munro and Derwing (1995) reported that one might be able to already process the sentence while listening.
Truth value duration
Besides the utterance duration of all accents, there was also a test performed on the duration of truth and false statements. The mean sentence duration of the truth statements (M = 1946,
sd = 344) and false statements (M = 2044, sd = 308) looked slightly different with 98 ms.
However, a t-test on the two conditions revealed that there was no significant difference between true and false statements across all accents, t (98) = 1,489, p = 0.140. This result means that there is no effect of true/false-value that should be considered when measuring reaction times.
Participants
Listeners
Twenty five native Dutch students, 7 male and 18 female, with an age ranged between 19 and 27 participated in this study. They all followed a bachelors or master’s program at the University of Groningen at the time of testing. To rule out above average effects of exposure to English, in particular British, no students doing a study in English language and cultures were tested. None of the participants knew anything about the list of statements used in this study. Participants were asked to indicate the amount of contact they had with English speaking situations in daily life on a 1 to 5 scale, where 1 is ‘no contact’ and 5 is ‘contact every day’, for each accent used in this study. This was asked after the test in the consent form (presented in Appendix 3), so that the question would not give away a clue for the accents used in the test. The participants reported to have more contact with the American (M = 4.04) and British (M = 3.88) accent than with the Australian (M = 1.84), the Canadian (M = 1.84) and the Irish accent (M = 1.64). This was mainly due to the fact that most of the participants watch American or British television or series.
Procedure
The experiment took place at the University of Groningen in a silent room. In the first part, participants were asked to decide as quickly as possible whether a statement was true or false. Each of the 20 statements was presented to them in a random order. Prior to hearing the statements, a fixation point and the text for the two possible buttons for true and false were shown on the screen for three seconds. Also, the buttons on the laptop for true and false, respectively “w” (waar (true)) and “o” (onwaar (false)), were colored green for true and red for false. Reaction times were measured by the time it took the participants to press the true or false button in E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA). To ensure that the participants had understood the stimuli and not just pressed the truth or false button randomly, they were asked to write down the statement after each decision task.
The second part of the experiment contained an attitude test towards the different native English accents. A questionnaire was presented to them, consisting of seven different categories (intelligibility, clearness, friendliness, richness, normality, intelligence and beauty), which the participants had to rate on a 1 to 5 scale. They were also asked to guess the accent of the speaker. The questionnaire is presented in Appendix 2. Before they continued with this part, they were first instructed to read the questionnaire, so that it was clear what they had to
give scores for, and then to listen carefully to the audio, because the statements could only be heard once. The listeners heard the same sentence (A week contains seven days) five times (each English accent one time) in a fixed order: Australian, American, Irish, British and Canadian. After each individual sentence, the listeners gave scores for the different attitude categories.
Results
Reaction times
Before analyzing the results, all reaction times falling above 2 standard deviations (sd = 1458 ms) from the mean (M = 1403 ms) were excluded to minimize the effects of outliers. These outliers were, for example, created due to the fact that a participant was trying to comprehend a statement that was difficult to understand for them, rather than immediately pressing the corresponding button. After pressing the correct button, they also produced a correct transcription of the sentence. A total of ten sentences were deleted from the data.
Analysis for truth value
As concluded above, no difference was found in the utterance duration of the speakers for true and false statements. To examine if the listeners responded differently to true and false statements, a Mann-Whitney U test was performed on the reaction times with truth value as factor. The test revealed that there was no significant difference in reaction times between true and false statements, W = 23248, p = 0.78. Thus, in the results below, no effect of truth value is expected and therefore not included in the analysis.
Intelligibility
Accuracy and transcription scores
The accuracy rate of each listener pressing the correct button for each statement was automatically collected by E-Prime. Only the responses that matched the corresponding outcome of the pretested statements were considered as correct. To verify that listeners pressed the correct button on purpose, the transcriptions of each statement were analyzed by the researcher. The transcriptions had to match the statements identically to be judged as correct, with the exception of a few minor mistakes (i.e. listeners were allowed to write definite article where the correct answer would be an indefinite article (and vice versa) and they were allowed to write a singular noun as plural (and vice versa)). Out of the 500
statements, 88% of the sentences were verified correctly. A difference between accents was found in the accuracy scores: the Australian sentences were verified 73% correctly, the Canadian 95%, the British 89%, the Irish 89% and last, the American 92%. A Kruskal-Wallis test revealed that this difference was significant, X2 (4, n = 500) = 21.76, p < 0.05. There was a difference between the Australian accent and the other accents (p < 0.05). The other accents matched in their accuracy scores.
Response times without speakers’ recordings
The mean reaction times of each listener, and only their correct responses plus correctly transcribed sentences, were put into a Kruskal-Wallis test. The mean reaction times, presented in figure 1, for the American (M = 1243, sd = 496), the Australian (M = 1288, sd = 798), the British (M = 1231, sd = 609), the Canadian (M = 1283, sd = 570) and the Irish (M = 1299, sd = 474) accent showed almost no difference and this was confirmed by the test, X2 (4, n = 125) = 1.07, p = 0.898.
Figure 1. Mean reaction times for the different accents, excluding speakers’ recordings.
Response times with speakers’ recordings
As stated above, only the American utterance duration of the speakers’ recordings differed from the Australian and Canadian. This means that for the other accents, the recordings could
be included in the analysis. Although the truth/false-value of each statement lies within the last word of each sentence, it was suggested that one can already process the rest of the sentence while listening (Munro and Derwing (1995)). This means that one might be able to press faster after the last word, when a sentence is produced slower by the speaker, because the listener has more time to process the rest of the sentence. The mean scores including the speakers’ recordings are presented in figure 2 below. Means for the total reaction times of the American (M = 2865, sd = 407), the Australian (M = 3685, sd = 765), the British (M = 3100,
sd = 431), the Canadian (M = 3363, sd = 670) and the Irish accent (M = 3263, sd = 511) were
put into a new Kruskal-Wallis test. A post-hoc Dunn’s test with Benjamini-Hochberg correction revealed that reaction times for the British accent were significantly lower than for the Australian accent (p < 0.05). Furthermore, a quicker response time was found for the American accent compared to the Irish (p < 0.05), the Canadian (p < 0.05) and the Australian accent (p < 0.001). Because the latter was already significantly slower than the British accent, the difference between the American and Australian accent is valid too, regardless the difference in utterance duration. No valid conclusion can be reported between the Canadian and American accent on the basis of this analysis. Thus, the degree of intelligibility of the American accent is higher than the Irish and Australian accent. Furthermore, the British accent has a higher degree of intelligibility than the Australian accent.
Familiarity
Contact and correctness
Besides rating the amount of daily contact with each native English accent in the consent form, participants were also asked to guess the accent they heard in the second part of the experiment to examine familiarity. A clear difference in correct identification of the five accents was found and presented in table 1 together with the mean contact. A Chi-Square analysis revealed that there was a significant association between the familiarity with an accent and correctly identifying the accent, X2 (4, n = 125) = 48.83, p < 0.001. The strong effect of the association confirmed the interpretation, Cramer’s V: 0.625.
Table 1
Accent correctness and contact in part two of the experiment
Accent Correct guess Incorrect guess Mean contact
American 15 10 4.04
British 15 10 3.88
Australian 1 24 1.84
Canadian 0 25 1.84
Irish 2 23 1.64
In table 2, a more detailed overview of the guess for each accent is presented. Some remarkable findings are found: the Australian accent is 13 times judged by the listeners to be a British accent. The listeners, on the other hand, guessed that the British accent was an Australian accent 5 times. Furthermore, the Canadian accent is perceived mostly as American accent (7 times), followed by British (4 times). And finally, the listeners guessed that the Irish accent was British 8 times. The column with other contains mostly “no guess” and no more than 2 similar guesses (e.g. Irish was perceived as Scottish 2 times).
Table 2
Guess per accent
Guess: American British Australian Canadian Irish Other
Accent: American 15 5 0 0 1 4 British 1 15 5 0 0 4 Australian 3 13 1 0 0 8 Canadian 7 4 1 0 0 13 Irish 2 8 1 0 2 12 Attitudes Analysis on attitudes
In the second part of this study, participants had to rate each speaker on different attitude categories after hearing the sentence A week contains seven days. The mean scores for all attitudes are presented in figure 3 to visualize any potential differences. Next, for all attitude categories, a Kruskal-Wallis test with post-hoc Dunn’s test was performed to indicate which speakers’ accents scored different in each category. The results of this analysis are shortly presented below.
Figure 3. Mean scores for the different attitudes on each speakers’ native English accent on a
scale from 1 (negative) to 5 (positive).
Intelligibility
Although the Canadian speaker seemed a little bit less intelligible than the other speakers, no significant difference was found between the speakers’ accents in intelligibility scores. This means all speakers sounded equally intelligible to the listeners.
Clearness
The test revealed that the British speaker sounded clearer than the Canadian (p < 0.01) and the Irish (p < 0.05) speaker. Furthermore, the American speaker sounded clearer than the Australian (p < 0.05), the Canadian (p < 0.01) and the Irish (p < 0.05) speaker. The difference in clearness might suggest two things. First, listeners might have rated the quality of the recordings, although this is plausible since each speaker sounded equally intelligible to the
participants. Second, this result might suggest that the listeners rated each speaker on their degree of accentedness of English. Unfortunately, no qualitative reports are collected.
Friendliness
The Australian speaker looked a bit less friendly than the other speakers, but this difference was not significant.
Richness
The Dunn’s test revealed that the British speaker sounded richer than the Canadian (p < 0.05) and the Irish (p < 0.05) speaker. Furthermore, the Australian speaker, thirteen times reported as British in guessing the accent, sounded richer than the Canadian (p < 0.05) and the Irish (p < 0.05) speaker too.
Normality
The Kruskal-Wallis test revealed that there was a significant difference between the speakers in normality (p = 0.03). However, due to the Benjamini-Hochberg correction in the Dunn’s post-hoc test, all individual comparisons failed to reach significance.
Intelligence
In addition to the results on richness, both the British and the Australian speaker were rated significantly higher as having an intelligent accent compared to the Canadian (p < 0.05) and the Irish (p < 0.05) speaker.
Beauty
The participants judged the British speaker as having a significantly more beautiful accent than the Canadian speaker (p < 0.01). The difference between British and the other three speakers failed to reach significance.
Average attitude
Finally, an average attitude score of all seven categories was conducted. A significant more positive average attitude was found for the British speaker compared to the Canadian (p < 0.01) and the Irish speaker (p < 0.05). Furthermore did the listeners held, on average, more positive attitudes towards the American speaker compared to the Canadian (p < 0.05) and Irish (p < 0.05) speaker.
Analysis on relationships
To examine hypothesis (2a), i.e. the possible relationship between intelligibility and familiarity, a correlation analysis was performed with mean reaction times per participant per language as one variable and contact as the other. The correlation analysis showed that the amount of contact with an English accent and the speed of correctly responding on the statements was significantly negatively related (rs = -0.382, p <0.001). This middle strong relationship suggests that the more contact one has with an English accent, the easier it is to understand the spoken message.
Next, it was hypothesized (2b) that attitudes would have a significant effect on the intelligibility of the accents too. Therefore, a second analysis was performed with mean reaction times and average attitude scores as variables. Although a plot suggested that higher attitude scores resulted in faster response times, the correlation analysis failed to show a significant relationship (rs = -0,171, p = 0.057). The weak negative relationship confirms this outcome.
A third correlation analysis was conducted to examine if higher familiarity is related to higher attitude scores (2c). The analysis revealed, again, no significant relationship between contact and attitude scores (rs = 0.154, p = 0.087). This weak positive relationship confirms the outcome that if one is more familiar with an accent, it will not result in a more positive attitude towards the accent.
Discussion
In the following chapter, the results will be discussed to provide an answer for the main question of this study: is there (1) a difference in the degree of intelligibility of native English accents and does (2a) the familiarity with or do (2b) the attitudes towards native English accents influence the degree of intelligibility for non-native speakers of English? Therefore, the results for intelligibility, familiarity and attitudes and their relationship will be discussed below and, where possible, compared to previous research.
Intelligibility
High accuracy scores were found for pressing the corresponding button and correctly transcribing each statement. Out of 500 statements, 88% was both verified and transcribed
correctly. This result is only slightly lower than the result found in Munro and Derwing (1995), where 94% out of 800 statements were verified and transcribed correctly. This study suggested that no differences in accuracy scores would be found, since all speakers were native English speakers. However, only 73% of the sentences by the Australian speaker were correctly verified, which turned out to be a difference with the other English accents. It is possible, as was mentioned by some of the participants, that the ‘relaxed’ sound of the Australian speakers’ voice formed a certain difficulty for the participants when listening the sentences. The fact that participants indicated that the Australian speaker sounded ‘relaxed’, is supporting the inherent value hypothesis, because only one participant was able to identify the speaker as Australian. Furthermore, the comment by the participants in this study are in line with previous research (Ladegaard, 1998), who found that Danish participants characterized the Australian accent as ‘laid-back’, while almost all of them were not able to identify the origin of the speaker.
In this study, two ways of analyzing the response times were conducted; one without and one with the speakers’ recordings. No differences in reaction times were found if analyzing solely the reaction times of pressing the corresponding button after the end of the recording, as was performed by Munro and Derwing (1995). However, in response to the recommendations of Munro and Derwing (1995), it was suggested that a sentence can already be processed while listening, and therefore might lead to quicker reaction times when a speakers’ recording is longer. Although the recording by the American speaker was significantly faster than the Canadian and Australian speaker, significant faster reaction times were found for the American accent compared to the Australian and Irish accent when including the speakers’ recording. Furthermore, listeners had quicker response times when judging the British statements compared to the Australian statements. So, hypothesis (1) is confirmed, as the results indicate that there is a difference in the degree of intelligibility between native English accents, with the American accent being more intelligible than the Australian and Irish accent, and the British accent being more intelligible than the Australian accent. It is possible that the above mentioned difficulty of the Australian accent (i.e. the difference in accuracy scores) might have resulted in slower reaction times too. This means that processing the Australian accent in general is more difficult than processing, for example, the British and American accent. The difference in degree of intelligibility could be caused by the higher familiarity scores for British and American, which will be discussed in the following paragraph.
Familiarity
The results of listeners’ self-reported exposure to the five different native English accents showed a clear distinction between on the one hand the American and British accent, and on the other hand the Australian, Canadian and Irish accent. The listeners clearly reported to have more contact with the British and American accent, due to the fact that most of the participants are watching television or series with these particular English accents. Besides giving a score on exposure to the different accents, listeners were asked to guess the accent they heard during the second part of the experiment. A significant strong correlation was found between the self-reported data and the correctness on guessing the accent. This means that the listeners were able to recognize the accents they were more exposed to (Munro & Derwing, 1994; Gass & Varonis, 1984). The study has presented results that were partly in line with hypothesis (2a), i.e. that higher familiarity leads to faster response times: both the American and the British accent were both more intelligible and more familiar to the listeners than the Australian accent. Furthermore, findings showed that the American accent was more intelligible and more familiar to the listeners than the Irish accent. Overall, a middle strong negative relationship was found when conducting a correlation analysis. This means that reaction times decreased when familiarity with the accent was greater. This result confirms hypothesis (2a) that higher familiarity with a native English accent leads to a higher degree of intelligibility (i.e. quicker response times).
The results of this study on familiarity contradict the results presented by Munro and Derwing (1995), who found no effect of exposure to accented speech on reaction times, but support the findings of Gass and Varonis (1984), who reported that exposure to non-native speech lead to a greater comprehension of that speech. The findings presented in this study might confirm additionally the idea of Munro and Derwing (1995) that speakers take accentedness into account. However, as mentioned before, this study has not tested the difference in degree of accentedness and what effect this could have on reaction times. In the present study, listeners might not have been as aware as the listeners in the study of Munro and Derwing (1995) on the differences in accentedness, since the speakers in this study are all native English speakers, instead of L2 speakers with a foreign accent in English. It is then suggested that accentedness in L2 speech might be a more important factor than accentedness between different L1 speakers. Further research should carry out research on the importance of degree of accentedness of different L1 speakers for the perception of non-native speakers of the language.
When examining the outcome of guessing the accent in the second part of the experiment, an interesting finding to consider is the fact that the Australian accent was perceived 13 times as a British accent. This would suggest that, when a listener thinks the Australian accent is British, it is expected that reaction times would be faster for the Australian accent, which was not the case here. There are two possible explanations for this contradictory result. First, accuracy scores of the Australian accent were significantly lower than the British. Accuracy scores might therefore have a higher impact on reaction times than was suggested so far. On the other hand, the Canadian accent was also perceived 7 times as an American accent, although the accuracy scores of the Canadian and American accent did not differ from each other. This leaves room for a second possible explanation: there was a fixed order in which the listeners had to give attitude scores and guess the accent in the second part of the experiment, starting with the Australian accent. Listeners might not have been able to recognize the accent as Australian, since their mean contact with this accent is very low, and, since it turned out to be rated equally rich and intelligent as the British accent, it is likely that people perceived this first recording as a British accent and did not change their guess after hearing the British speaker. It would therefore be better if the recordings in the second part of the experiment were presented to the listeners in random order, to examine if participants would guess the Australian accent as British too, if the accent is presented after the British accent.
Attitudes
The third topic this study was interested in, was the participants’ attitudes towards the different native English accents and the relationship between attitudes and intelligibility. As mentioned above, the British and Australian accent received higher scores on both richness and intelligence compared to the Irish and Canadian accent. Furthermore, the British accent received a higher score on beauty compared to the Canadian speaker. The high attitude scores for British on richness, intelligence and beauty are supporting the imposed norm hypothesis, since British was more standard to the listeners, and, moreover, in line with previous research mentioned in Ladegaard (1998), who stated that speakers of Received Pronunciation (which is the accent the British speaker in this study has) are more competent and possessing more status compared to other dialects, for example Irish (Milroy & McClenaghan, 1977). Furthermore, the British and American accent received on average both higher attitude scores than the Canadian and Irish accent. However, the statistics revealed no significant relationship between familiarity and attitude scores. This means for hypothesis (2c) that, although the
accent might be more standard to the listener, it is not automatically receiving higher attitude scores. It is suggested that the Australian accent might have influenced this relationship, since it did not receive significantly lower attitude scores than British and American, due to the fact that the Australian accent could have been perceived as British, as mentioned above.
Furthermore, Boets and De Schutter (1977) reported that there was a relationship between intelligibility and attitudes on different accents of the same language. It was suggested in hypothesis (2b) that similar results would be found in this study. However, there was no evidence here that degree of intelligibility was influenced by attitude scores. Statistically, the correlation analysis just failed to reach significance. This finding is more in line with the result of Schüppert, Hilton and Gooskens (2015), who reported that attitudes and intelligibility were only weakly linked, when examining two closely related languages such as Danish and Swedish.
Conclusion
The findings reported here show a difference in the degree of intelligibility between native English accents. Coming back to the research question of this study (i.e. whether (1) there is a difference in the degree of intelligibility of different native English accents and, if so, whether (2a) this is influenced by the familiarity with the accents or (2b) attitudes towards the accents for non-native speaker of English), this difference is most likely caused by the amount of exposure one has to the different native English accents. Although attitudes like richness and intelligence (i.e. higher social status) are in line with previous research, on average, attitudes do not seem to really influence the difference in degree of intelligibility reported in this study. However, the hypothesis (2c) that British and American are seen as more ‘standard’ and will therefore be more appreciated is partly confirmed, due to the fact that Australian was often identified as British. The finding is in line with the imposed norm hypothesis and previous research of Van Bezooijen (1994), who reported that Standard Dutch was most attractive to all groups of participants. The fact that British and American are seen as the more standard accents is likely the effect of watching series, social media, etc. Most of the series include American accents. It is therefore not surprising that the participants in this study were able to recognize the American accent quite easily. Besides American, the British accent was identified correctly equally often. Ladegaard (1998) already noted that in case of Danish participants, British (RP) is seen as a form of ‘standard English’. It is not only in Denmark, but also in the Netherlands, where the British accent is the main accent taught in schools and
university as the ‘correct’ form. The findings indicate that familiarity with a particular accent will result in a better understanding of the accent and, furthermore, result in slightly more positive attitudes. This might also result in the other way around, i.e. that the learner of a native English accent might be more willing to put effort in learning the accent, when the accent becomes more familiar. The suggestion remains that, if, for example, the Irish accent would be taught more often in schools, the accent would become more appreciated and more intelligible to non-native speakers of English.
The finding that slower utterance duration resulted in higher intelligibility, as was reported by Munro and Derwing (1995), is not confirmed in this study. In fact, the Australian accent was equally intelligible with the other accents, when, like Munro and Derwing (1995), utterance duration was not taken into account in measuring reaction times. It should be noted, though, that the perception of foreign accentedness for L1 speakers might have a stronger impact on intelligibility than differences in the perception of accentedness in native English for L2 speakers. There is, thus, a strong need to examine what role degree of accentedness play for the degree of intelligibility of a native English accent. A better understanding in the role of accentedness of L1 speech for the perception of L2 speakers might have benefits for future English teaching programs. It is speculated that the Australian, Canadian and Irish accent would receive higher scores for degree of accentedness in this experiment, since familiarity with the British and American accent was higher and was seen as more standard to the listeners. This suggestion is in line with the findings reported by Gass and Varonis (1984), who reported that familiarity with a particular speech resulted in higher comprehensibility. It is, moreover, interesting to examine what linguistic features, in particular phonetics, play a role for not only the perception of degree of accentedness, but also for familiarity and attitudes. For example, listeners in this study reported that the Australian speaker had a ‘relaxed’ sound in his speech, which might be the result of the slower utterance duration. However, because familiarity with the Australian accent was low, the participants were not able to link the ‘relaxed’ sound with an Australian accent, but identified the accent as being British. This suggests that, for the perception of L2 speakers, some linguistic features between British and Australian are the same. Furthermore, this study reported findings that the Canadian accent was perceived as an American accent. It would be interesting to find out which linguistic features result in the high perception of British for Australian and American for Canadian.
References
van Bezooijen, R. (1994). Aesthetic evaluation of Dutch language varieties. Language &
Communication, 14(3), 253-263.
Boets, H., & De Schutter, G. (1977). Verstaanbaarheid en appreciatie. Nederlandse dialekten uit België zoals inwoners van Duffel die ervaren. Taal en Tongval, 29, 156-177.
Brodkey, D. (1972). Dictation as a measure of mutual intelligibility: A pilot study. Language
Learning, 22(2), 203-217.
Davies, M. (2008). The corpus of contemporary American English. BYE, Brigham Young University. https://corpus.byu.edu/coca/
Education First English Proficiency Index (EF EPI). (2018). Sweden Returns to Top Spot in World Ranking of English Proficiency Reveals EF’s Annual Proficiency Index.
Retrieved from: https://www.prnewswire.com/news-releases/sweden-returns-to-top- spot-in-world-ranking-of-english-proficiency-reveals-ef-s-annual-proficiency-index-852505467.html
Fayer, J. M., & Krasinski, E. (1987). Native and nonnative judgments of intelligibility and irritation. Language Learning, 37(3), 313-326.
Gass, S., & Varonis, E. M. (1984). The effect of familiarity on the comprehensibility of nonnative speech. Language learning, 34(1), 65-87.
Giles, H. (1970). Evaluative reactions to accents. Educational review, 22(3), 211-227.
Giles, H., Bourhis, R., & Davies, A. (1979). Prestige speech styles: the imposed norm and inherent value hypotheses. Language and society: Anthropological issues, 589-596. Giles, H., Bourhis, R., Lewis, A., & Trudgill, P. (1974). The imposed norm hypothesis: A
validation. Quarterly Journal of Speech, 60(4), 405-410.
Giles, H., & Powesland, P. F. (1975). Speech style and social evaluation. Academic Press. Harbeck, J. (2016, August 11). How English gave birth to surprising new languages.
Retrieved from: http://www.bbc.com/culture/story/20160811-how-english-gave-birth-to-surprising-new-languages
Kachru, B. B. (Ed.). (1992). The other tongue: English across cultures. University of Illinois Press.
Ladegaard, H. J. (1998). National stereotypes and language attitudes: The perception of British, American and Australian language and culture in Denmark. Language &
Communication, 18(4), 251-274.
Lane, H. (1963). Foreign accent and speech distortion. The Journal of the Acoustical Society
of America, 35(4), 451-453.
Milroy, L., & McClenaghan, P. (1977). Stereotyped reactions to four educated accents in Ulster. Belfast Working Papers in Language and Linguistics, 2(4), 1-11.
Munro, M. J., & Derwing, T. M. (1994). Evaluations of foreign accent in extemporaneous and read material. Language Testing, 11(3), 253-266.
Munro, M. J., & Derwing, T. M. (1995). Processing time, accent, and comprehensibility in the perception of native and foreign-accented speech. Language and speech, 38(3), 289-306.
Munro, M., & Derwing, T. M. (1999). Foreign accent, comprehensibility, and intelligibility in the speech of second language learners. Language Learning, 49, 285-310.
Palmer J. (1976). Linguistic accuracy and intelligibility. In Proceedings of the 4th
International Congress of Applied Linguistics, 505–513. Stuttgart, Germany: Hockschul
Verlag.
Psychology Software Tools, Inc. [E-Prime 2.0]. (2016). Retrieved from
http://www.pstnet.com.
Schüppert, A., Hilton, N. H., & Gooskens, C. (2015). Swedish is beautiful, Danish is ugly? Investigating the link between language attitudes and spoken word recognition.
Linguistics, 53(2), 375-403.
Stewart, M. A., Ryan, E. B., & Giles, H. (1985). Accent and social class effects on status and solidarity evaluations. Personality and Social Psychology Bulletin, 11(1), 98-105.
Appendix
Appendix 1 True statements
1. A hand is part of the body. 2. Our country has a government. 3. Summer is between spring and fall. 4. A researcher can work at a university. 5. The sun can melt ice.
6. A car is a vehicle.
7. A week contains seven days. 8. Each hour has sixty minutes. 9. Our planet has clouds in the sky. 10. A weapon can be dangerous. False statements
1. The ear is part of your leg.
2. The majority of people live on a farm. 3. Young children are always silent. 4. Most people drink beer for breakfast. 5. Grass always has the colour red. 6. People often cycle on water. 7. Smoking cigarettes is very healthy. 8. In America, the currency is the Euro. 9. A tooth lies within the nose.
Appendix 2
The questionnaire was originally presented to the listeners in Dutch. A translation of the questionnaire is provided below.
Geef voor ieder fragment aan hoe je de spreker vindt klinken.
Onbegrijpelijk 1 2 3 4 5 Begrijpelijk Onduidelijk 1 2 3 4 5 Duidelijk Onvriendelijk 1 2 3 4 5 Vriendelijk Arm 1 2 3 4 5 Rijk Raar 1 2 3 4 5 Normaal Onintelligent 1 2 3 4 5 Intelligent Lelijk 1 2 3 4 5 Mooi
Welk Engelse accent denk je gehoord te hebben? ……… Indicate for each fragment how you think the speakers sounds.
Non-intelligible 1 2 3 4 5 Intelligible Unclear 1 2 3 4 5 Clear Unfriendly 1 2 3 4 5 Friendly Poor 1 2 3 4 5 Rich Strange 1 2 3 4 5 Normal Unintelligent 1 2 3 4 5 Intelligent Ugly 1 2 3 4 5 Beautiful
Appendix 3 Toestemmingsverklaring Naam: ___________________________________ Geslacht: M / V Leeftijd: ___________________________________ Opleiding: ___________________________________
Geef hieronder aan in welke mate (1 = nooit, 5 = iedere dag) je in je dagelijkse leven in contact komt of bent gekomen met de volgende Engelse accenten:
Eventuele opmerkingen: Amerikaans 1 2 3 4 5 ………... Brits 1 2 3 4 5 ………... Australisch 1 2 3 4 5 ……….. Canadees 1 2 3 4 5 ……….. Iers 1 2 3 4 5 ………..
Door het ondertekenen van dit formulier verklaar ik vrijwillig deel te nemen aan dit onderzoek. Dit houdt in dat ik me te allen tijde mag terugtrekken uit het onderzoek zonder opgave van reden. Ik heb vragen mogen stellen en deze zijn beantwoord. Aan het onderzoek zijn geen risico’s verbonden. Mijn gegevens zullen vertrouwelijk en anoniem behandeld worden. Ik heb dit formulier gelezen en
begrepen en ga hiermee akkoord. Datum: ___________________________________ Handtekening: ____________________________
