Exploring norms and variation in L1/L2 pronunciation Entre écarts et normes de prononciation : acquisition, variation, apprentissage

et Maelle AMAND 

Mots-clés : écarts et normes, perception L1/L2, prononciation L1/L2, variation socio-phonétique

Keywords : L1/L2 perception, L1/L2 pronunciation norms and deviation from the norm, sociophonetic variation

Texte intégral


1Today, many tools are available to enable researchers to conduct in-depth studies of pronunciation characteristics: acoustic analysis tools like Praat (Boersma & Weenik, 2007) showing precise phonetic detail, statistical analysis tools allowing researchers to map patterns and draw relevant conclusions from their results. In this volume, contributions include a wide range of methodologies to address the question of norms and variation, focusing on both native languages (L1) and foreign language learning (L2). The languages under study are English, both as L1 and L2, French as L1 and L2, Japanese L1, and L1 Qeltu, a variety of Arabic. A qualitative analysis approach is favoured with data collection methods varying from lab experiments to cohorts of students and official corpora (the DECTE corpus), the focus being mainly on speech production but one study also includes speech perception. Several dozen participants were involved in each study representing different age groups, from primary school learners to adults (mainly young), both male and female.

Note de bas de page 1 :

English as a Foreign Language.

2Our question “norms and variation” reflects a gradient rather than any absolute value. As teachers, we are faced with a paradox: the necessity of presenting learners with a coherent pronunciation system, and our awareness of its true complexity. We may even speak of the kaleidoscopic nature of variation in language. But as researchers, we are free to consider language in all its diversity. We shall therefore begin from the perspective of variation. Indeed, phonetic analysis immediately highlights the diversity of realizations for apparently similar utterances and when these occur regularly (as opposed to performance-related differences), we may speak of “variation”. As Ogden (2012, p. 49) states, “it is a truism to say that speech–especially normal conversational speech–is highly variable”. Variation in speech can be studied from a number of perspectives: phonetic, phonological, social… (Chambers et al., 2013; Thomas, 2011). Moreover, if the term “variation” implies diversity, and perhaps individuality, there is also an implicit comparison with some sort of general “norm”, raising the question of the degree of acceptability (a social concept) and even intelligibility of such variation (cf. Nyberg, J., & Strömbergsson 2021 in SLT, Gooskens 2017 on dialect intelligibility, Henderson 2021 on intelligibility in the EFL1 context).

1. Atypical profiles in L1: questioning norms

Note de bas de page 2 :

“La Richesse des écarts à la norme” was quoted in the title of the PhonLim 2022 symposium (Limoges University, 18-19 March 2022: la “richesse des écarts à la norme” : individus atypiques en prononciation L1 / L2). It was translated into English as “Outliers, a goldmine for research”.

3Intelligibility is particularly significant in a medical context where speech impediments or defects mean that language may deviate from what is considered standard, possibly to the extent that communication is undermined. In such cases, it appears relevant to establish what the norm/standard is. Some researchers suggest that individual deviation may also be considered of interest for research. In his article “La Richesse des écarts à la norme”2, Bullinger (2002) analyses child language development and argues that what lies outside the standard also needs to be taken into account. This is also the standpoint of Cattu Alves, Ode & Strömbergsson in their article entitled, “Dealing with the unknown–addressing challenges in evaluating unintelligible speech” (2020) where they question the way standard speech is currently evaluated: “[w]hen investigating the interaction between speech production and intelligibility, unintelligible speech portions are often of particular interest” (Cattu Alves et al. 2020), proposing additional metrics beyond the common Percentage of Consonant Correct (PCC) which allow us to measure the proportion of intelligible speech (consonants and syllables).

Note de bas de page 3 :

The original quotation: “Prendre en compte ces particularités, ne pas les référer à la norme, nous fait comprendre la richesse et la subtilité de ces conduites. Faire l’économie de cette décentration nous diminue et mutile les personnes qui les produisent. Tenter d’en connaitre les mécanismes permet de mettre en place des moyens de soin adaptés.” (Bullinger, 2002, p. 103)

4Likewise, Bullinger (2002, p. 12) suggests that even if the performances of atypical children do not match the norm, they still do not necessarily perturb measures. He further argues that trying to understand the specificities rather than approaching them via a norm is the best way to set up adapted communication strategies3.

Note de bas de page 4 :

Sofi Strömbergsson was key speaker at the PhonLim 2023 symposium held in Limoges in June 2023.

5Sofi Strömbergsson developed this idea in her paper, “Navigating the search for “normal” in children’s speech / language disorders” (2023)4, confronting two perspectives: on the one hand, intelligibility as a linguistic concept, leading to potential misinterpretation; on the other, acceptability as a social one, highlighting the need for efficient communication with others within a social network. Like Bullinger, she suggests that, when evaluating norms, measures should include non-verbal aspects in narratives. She also suggests that different perspectives should be combined so as to embrace diversity because extensive language variation seems to be increasingly the “norm” rather than the exception.

6This diversity clearly represents quite a challenge for speech and language pathologists who need to adapt their appreciation of what the norm is (e.g. Klein, 2011; Jauer-Niworowska, 2022). But if norms need to be questioned, the goal for a speech therapist is still to be able to assess the extent to which deviant pronunciation hinders intelligibility and therefore communication, and to help the patient overcome the difficulties in their L1 where possible. As Jauer-Niworowska puts it: “[a] message which is not fully functional may regain its functionality after the causes of difficulties in its reception and interpretation and/or correct construction are removed” (2022, p. 9). She defines this strategy as a “reinstatement of normative functioning of the communication” (Ibid.).

Note de bas de page 5 :

“Teachers and speech therapists need to deal with norms as professionals” (our translation).

Note de bas de page 6 :

“Part of their job involves assessing and correcting and so this social role, when interacting with novice partners (unless otherwise stated), means establishing an implicit didactic contract” (our translation).

7We cannot but draw a parallel between their objectives and those of L2 teachers who aim to promote communication skills in L2: “[e]nseignants et orthophonistes sont en quelque sorte des professionnels de la norme”5, remarks Py (2000). He adds: “leur métier consiste en partie à évaluer et à corriger et que l’activation de ce rôle social dans des interactions avec des partenaires novices entraîne (sauf décision contraire) l’établissement implicite d’un contrat didactique”6. This raises the question of what “norm” to target as a reference model for learners. Angelovska and Hahn remind us that “in the traditional view of foreign language learning and teaching, the native speaker criterion is taken as a measure of success in learning, as well as a role model for language teaching and as a measuring stick in research” (2009, p. 164). But what is meant by a “native speaker criterion”?

2. Norms, variation and reference

8If “nativeness” in EFL generally refers to an aimed target associated with standard pronunciation, native speakers of the same language are far from sharing identical phonetic features. When features are shared and accepted by a recognizable (social) group (Eckert, 1989), they may constitute a new variety (Hickey, 2003) which is therefore regarded as a “norm” by that group of speakers, be it covert or overt (Trudgill, 1972). As research in sociolinguistics has shown, beyond regional variation, there are many layers to the language a native speaker uses, influenced by social environment, register and phonostyle (Léon, 2000), among others.

9Despite such diversity, the choice of an L1 reference model for L2 tends to be fairly restricted. It may simply depend on the teachers’ own norm which is then passed on to the learners who consciously or unconsciously adopt it (Dalton-Puffer et al., 1997). Reference works tend to refer to a limited number of models, often just one (cf. Michari, 2023, for French textbooks for EFL). In this volume where L2 French and L2 English are studied, reference is made to two varieties of the language and not just one (Kamiyama et al., Capliez), thereby including the concept of variation when dealing with second language learning. For L1 French, these are standard Parisian French, aka. Reference French (Detey et al., 2016), and Canadian French (used here for research purposes rather than as a teaching model). For English, there have been several ways to refer to British English (Received Pronunciation (RP), BBC English, Standard or Southern British English (SBE)–used as a teaching model by Vaissière & Exare in the present volume. The other variety is generally referred to as General American (GA or GenAm).

2.1. Reference models: what’s in a name

10The names chosen for English are worthy of note. The term General American (GA) is used in British English pronouncing dictionaries which propose a single American variant corresponding to the British English entry (or entries) and designates what aims to be a commonly acceptable pronunciation that does not single out any social or regional variety (Jones, 2006, p. vi). Yet, in America, the term has been subject to scrutiny, particularly as regards what it covers (Van Riper, 1986). An alternative to GA is proposed by Kretzschmar who considers that the term General American “implies there is some exemplary state of American English from which other varieties deviate” (2008, p. 42). He contests this and proposes to use the term Standard American English (StAmE) to refer to an unmarked pronunciation which “designates the level of quality (here of pronunciation) that is employed by educated speakers in formal settings” (Kretzschmar, 2008, p. 37).

Note de bas de page 7 :

This is the terminology used in recent Board of Examiner’s reports for the Agrégation Externe d’anglais, a French teaching exam, henceforth referred to as AEA.

Note de bas de page 8 :

Today, in both the Longman Pronouncing Dictionary (Wells, 2008) and the English Pronouncing Dictionary (Jones, 2018), systematic reference to both British and American English is made (and a few variants are included, but these cannot be exhaustive). Fifty years earlier, there was no systematic reference to American English in the 12th edition (Jones, 1963).

11For British English, a number of names have been used and there has been much discussion as to how to refer to a “standard” and what the standard represents. Unlike Kretzschmar’s interpretation of “standard” for American English, the association of the term to British English may suggest some sort of recommendation, potentially implying that other varieties are not recommendable. By contrast, Southern British English7 suggests that varieties are categorised based on a geographical scale rather than a hierarchical one. The discussion is reflected in the explanations given in pronunciation dictionaries for the model described in the entries. The lexicographers’ objective is to provide learners with a generally acceptable model; they may also include alternatives and even refer to another variety8, but their focus is on one (or two) reference models. To designate the variety described in the 12th edition of the English Pronouncing Dictionary, Jones refers to Received Pronunciation (RP) “for want of a better term” (Jones, 1963, p. xvi) but qualifies his use of the term as follows: “I should like it to be understood, however, that RP means merely ‘widely understood pronunciation’ and that I do not hold it up as a standard which everyone is recommended to adopt.” (Ibid.) As he says, this variety is widely understood, but it is not in fact widely used, and he concludes: “There exist countless other ways of pronouncing English, some of them being used by large communities.” (Ibid.). Nearly fifty years later, in the reference section of the 18th edition of the English Pronouncing Dictionary, Catherine Sangster in her contribution entitled “The BBC, its Pronunciation Unit, and ‘BBC English’” again refers to RP and to the term ‘BBC English’, which she considers inappropriate:

There is not (and never was) an official BBC pronunciation standard; its broadcasters speak English with a range of accents. In spite of this, the term ‘BBC English’ is often used as a synonym for Received Pronunciation (Roach, 2004) and for the sort of accent described in dictionaries like this one. (Jones, 2011, p. xxix)

Note de bas de page 9 :

She explains that the variety described in the dictionary did indeed correspond to BBC newsreaders in the past but she feels: “this would be better viewed as a by-product of the restricted social group from which BBC employees was drawn at that time, rather than a matter of deliberate policy.” (Jones, 2011, p. xxix)

12She shows how far broadcasting has evolved9, embracing a very broad approach to pronunciation which reflects current BBC policy as regards diversity: “[…] many accents can now be heard. This is felt to be important for the BBC because it means that the voices that the BBC’s broad audience hear in its broadcasts are as diverse as their own voices.” (Jones, 2011, p. xxix).

13The debate on standard varieties reflects changing attitudes to variation itself (Baratta, 2023). Yet, in a learning environment, whatever the name, teachers will promote one or more reference models.

2.2. Norms and reference models for L2

Note de bas de page 10 :

Vaissière & Exare were guest speakers at the PhonLim 2022 symposium (Limoges University, 18-19 March 2022) where they presented CleanAccent ©.

Note de bas de page 11 :

For them, this refers to “Standard British English”.

Note de bas de page 12 :

« La variété choisie pour les leçons en classe est l’anglais britannique standard pour deux raisons. Tout d’abord, l’anglais britannique est encore, par rapport à l’anglais américain, la variété la plus enseignée en France à l’heure actuelle (Michari, 2023). De plus, les deux enseignantes filmées pour l’expérimentation maîtrisent l’anglais britannique standard. »

14As there are significant differences in the vowel systems of General American and Southern British English, Vaissière & Exare10 choose to use a single model as their reference for L2 English since they are promoting a method for learning pronunciation, thereby exposing young learners to one coherent pronunciation system. Furthermore, they justify their choice of SBE11 as the standard for pragmatic reasons12 because they consider, following Michari (2023), that this is the most widely taught variety of English in France today, and secondly, that the two speakers who participated in the videos speak this variety.

Note de bas de page 13 :

Capliez refers to Southern British English.

15Likewise, Capliez adopts SBE13 as the norm but since he focuses on the pronunciation (or not) of <h> and the glottal fricative /h/, SBE and General American share the same characteristic–he does nonetheless identify specificities of General American /h/. But he points out that h-dropping does occur in some varieties (accents) of English, but it is not regarded as a model for learners because h-dropping indexes marked social identities. Indeed, sociolinguistic research has confirmed the correlation between h-dropping and social factors (Wells, 1982, p. 254). This is probably still very marked in terms of social acceptability despite a much more inclusive approach to diversity.

16Reference models are generally used to establish a norm for learners’ productions. This is the perspective for much teaching and research, as we see in the contributions of Vaissière & Exare and Capliez. But in one of their experiments, Kamiyama et al. take a different approach and try to establish possible correlations between listeners with different L1 varieties and their perception of L2 learners’ speech. In their experiment, they ask subjects who speak different reference varieties of French (Quebec and Parisian French) to assess the performances of Japanese learners of L2 French, thereby measuring the impact of the variety on perception.

2.3. Norms and atypical profiles in L2

17Research in foreign language learning has focused mainly on groups of individuals that have similar second language perception and production strategies rather than on individuals. Studies of L2 often propose to measure the influence of factors that may contribute to explain or improve second language perception and production. Learning models are developed with regard to the learning strategies of the majority. Yet this type of analysis also raises the question of outliers–the speakers who do not correspond to the pattern–that are also worthy of interest and it is by considering what deviates from the norm that we can fully appreciate second language production and perception in all its diversity.

18Brevik and Hellekjaer (2017) focused their study on L2 readers with atypical profiles. They studied a statistically marginal group of readers who performed significantly better when reading their L2 than their L1, thus challenging the premise that L1 and L2 reading competences are correlated. Their work questions the way learners acquire reading strategies and they conclude that:

Outliers show that we need to rethink L1 and L2 reading instruction, not only to understand why these students read markedly better in English as L2 than in their L1, but also to develop these students as readers […] taking students’ interests into account when choosing reading tasks and materials. (Brevik & Hellekjaer, 2017, p. 89)

19Likewise, Cauvin (2017) in her work on learners and prosodic patterns identifies an atypical profile type which she calls the “pianist profile”, students who perform both very well and very poorly as compared with a native model. She hypothesizes that these students have very specific learning strategies, particularly influenced by written English, and that diagnosing the profile and providing them with relevant tools should help to improve performance.

20In all of these studies, researchers are trying to bridge the gap between L1 and L2 and, so far, we have considered norms and references with regard to an L1 model for L2 learners. If some learners present atypical characteristics, they all have in common the fact that they go into the learning process with their own L1 norm, namely cross-linguistic phenomena. We shall now consider what the implications of this are.

3. L1 norms and L2 challenges

21Learners not only have to address the challenges raised by the native speaker model chosen as the benchmark, they also need to deal with their own L1 norm which will necessarily influence how they perform in L2. Flege (1995) considers that, as this is their primary reference, it will act as a filter influencing both perception and production of L2.

3.1. L1 norm / speaking L2

22As regards production, the characteristics of a learner’s L1 will trigger fairly predictable variation with respect to any native model. In his Speech Learning Model, Flege (1995) posits that on a segmental level, three categories may be identified: phonemes that are shared between L1 and L2; phonemes which are absent from the L1 inventory; phonemes that are similar to those in L1. He argues that the similar-but-different phonemes will be most tricky to produce as they tend to be articulated as for L1, unlike new phonemes where there is no interference.

23This is one of the hypotheses explored by Capliez in his contribution to this volume since the glottal fricative /h/, absent from the L1 inventory of the French students he is analyzing, does not correspond at all to their L1 code. Kamiyama et al. go a step further by considering not one but two varieties of L1 Japanese in order to see how the L1 norm of the Japanese learners affects their pronunciation of L2 French.

3.2. L1 norm / interpreting L2

Note de bas de page 14 :

Moore Mauroux (2022) demonstrates how learning the English spelling-sound code can benefit learners.

24If it is true that the spelling-sound codes of a learner’s L1 will influence their pronunciation of written L2 until they integrate the L2 code14, it is surely listening that is most challenging because, so often, what they see is not what they expect to hear.

25Indeed, phonetic variation in L1, and by extension in the L1 model for L2 learners, can be divided into two categories: (1) word internal variation and (2) variation that occurs across word boundaries in connected speech. In the first case, phonetic processes such as assimilation result in pronunciations that are far from the way words are written: bonfire, when pronounced /ˈbɒmfaɪə/may be misinterpreted as bomb fire, which is far from the actual meaning. Likewise, sandwich is frequently pronounced /ˈsæmwɪʤ/ (like Sam widge) with elision of the intermediate consonant in the /ndw/ cluster followed by an anticipatory assimilation of the labialisation in /w/ and voicing of the final affricate. For both <nd> and <ch>, the general reading rules for English do not apply. The native pronunciation will be even less recognizable for learners because, as the word has been borrowed in many languages including French, they are used to saying and hearing it according to their own spelling-sound code. In the second case, as Huart (2010) points out, the process frequently affects unstressed grammatical words (but not only) in connected speech. Phonetic variation occurs especially in spontaneous speech, in interactive situations, where tempo is rapid. Vowels tend to lose their identity (pronounced as the central vowel schwa), consonants may be elided, making the connection between the written word and its pronunciation unrecognizable, which presents a real difficulty for some learners, e.g./jɔːʃɔːtətraʊzəz/for your shorter trousers or you’re short of trousers. Such realizations are readily intelligible for a native speaker who has an innate understanding of these processes. It is often, however, challenging for learners who do not possess the tools to decode the signal.

Note de bas de page 15 :

The 2018 AEA report suggests that awareness of the specificities of the two reference accents (SBE & GA) should be beneficial to candidates / teachers in acquiring a coherent variety of English (AEA 2018, p. 82).

Note de bas de page 16 :

« La maîtrise phonologique du locuteur natif idéalisé a longtemps été un objectif de l’enseignement des langues, l’accent représentant un indicateur de mauvaise maîtrise de la phonologie. L’insistance mise sur l’accent et son exactitude au détriment de l’intelligibilité a nui au développement de l’enseignement de la prononciation. Les modèles idéalisés qui ignorent les accents, ne prennent pas en compte les contextes, les aspects sociolinguistiques ni les besoins des apprenants. » (CECRL, 2018, p. 140)

26Yet, beyond this type of predictable context-dependent variation, there is increasing awareness of the diversity of varieties and the necessity for learners to hear them. Even if emphasis in teaching exams is still laid on the coherence of the teacher’s pronunciation15 (and by implication that of the learner), in recent directives of the Common European Framework for Languages, trying to imitate a native norm is considered to be counter-productive16; learners simply need to approximate to a chosen standard, not reproduce it. Angelovska and Hahn challenged the premise that learners’ proficiency should be measured by “the native speaker criterion” (2009, p. 164) and today the trend is clearly shifting away from this towards a focus on communication and intelligibility (Cook, 1999; Bøhn & Hansen, 2017). Current policy in language teaching in France reflects this approach, and, just as importantly, French official instructions also identify the cultural importance of being exposed to different varieties and not just the sole reference standards (Bouvet, 2021).

4. L1 norms and variation

27What lies outside standard L1 norms can contribute to a greater understanding of language and communication. This is what Ogden (2012, p. 54) demonstrates in his paper entitled “Making Sense of Outliers” where he studies apparently atypical phonetic cues (at least in statistical terms) in native American English speakers, arguing that these in fact contribute to intensifying emphasis (IE). His findings show that segmental cues combine with prosodic cues to convey emphasis and, just as connected speech phenomena are perfectly decoded by native speakers, he explains that even though rare tokens do not occur frequently in the language, they remain meaningful, and should not be disregarded in language studies.

28Likewise, much research in sociolinguistics has been devoted to studying different pronunciations as well as lexical / grammatical specificities in different varieties of languages, that is to say, where groups of speakers share common pronunciation characteristics. Researchers may choose to explore variation from a number of different perspectives: ethnic, regional, geographical, social, gender… The amount of data available for analysis varies from one language to another, which necessarily has an impact on the methodologies used, ranging from statistical analyses of large corpora to lab experiments on a smaller scale, as we see in the contributions to this volume.

5. The volume: L1/L2 corpora and analysis methodologies

29Where for English, large corpora have been set up in order to analyze variation, in other languages, there has been much less background work. This is certainly the case for Qeltu, a variety of Arabic, which is the focus of Yaseen’s article. He starts by questioning an established “norm” for explanations of vowel lowering in Qeltu–that the process is triggered by proximity to guttural sounds–and sets out to explore the process further by setting up a number of sociolinguistic interviews involving some thirty respondents. His analysis brings to light hitherto unreported patterns of vowel lowering, suggesting that there are external factors involved in vowel lowering in Qeltu that go beyond the system–internal ones that have been traditionally identified.

30Amand’s contribution is based on a well-established corpus of Tyneside English, the DECTE corpus (Corrigan et al., 2012). She works from the original transcriptions of some forty-four adult speakers aged between 15 and 80, both male and female, who are mainly from working-class backgrounds. Her analysis is based on variation in four key vowels of Tyneside English. Using advanced statistical tools, combining Multiple Factorial Analysis and cluster analysis (Husson, 2010), to identify sub-groups of speakers, she investigates to what extent the speakers correspond to the norm for this variety of English and how coherent their variety is with regard to these 4 vowels. Her results reveal significant class and gender disparity.

31For Kamiyama et al., L1 variation is studied in relation to learning L2. Their analysis is based on a series of lab experiments targeting L2 French which were realized with Japanese speakers from two distinct regions of Japan: Tokyo (Kanto) and Osaka (Kansai). They measure the formant values of Japanese vowels, then study the perception and pronunciation of French vowels, especially /u/. The objective of the study is to establish how far the variety of L1 Japanese affects performance in French. Moreover, production was assessed by native French speakers using different varieties (Quebec / Paris), adding an interesting layer of complexity to the question of norms and variation.

32Likewise, Capliez has set up his own experimental corpus to explore how French L1 learners cope with phonemes in the L2 reference norm that do not exist in their own L1. He focuses on the pronunciation of /h/ and analyzes ninety-four subjects, forty-two men and fifty-two women, aged between 18 and 23, measuring their performance in three different recorded tasks (word reading, sentence reading, and spontaneous speech). From a detailed comparison of the different speakers, he shows how performance is related to the type of speech; his findings also highlight distinct gender differences. This exploration of the relation between L1 and L2 raises questions about teaching methods in general, including how best to introduce new phonemes.

33This is very much the concern of Vaissière & Exare who have set up an innovative system for early learning of English in French primary schools. Their approach draws on a great many references in the fields of both linguistics and didactics. The method is based on the premise that introducing a new language–so a new norm–will be more successful if it is done very early, before vocabulary acquisition and before commonly observed errors have had a chance to stick. The method focuses mainly on the sub-segmental and segmental features that are potentially challenging to French learners of English (vowels, consonants or aspiration). The experiment, so far carried out in one school, highlights the need for researchers, primary school teachers and also parents to work together.


34The present volume was largely inspired by the papers presented during the PhonLim 2022 symposium where we began investigating variation in language acquisition, language learning and native languages, with a special interest in outliers and atypical pronunciation. The theme was further developed at the PhonLim 2023 symposium, from the perspective of speech therapy. Papers and the ensuing discussions led us to take a slightly different angle for this volume, addressing both norms and variation in language.

35The volume includes studies relating to four languages (French, English, Arabic and Japanese); the question of norms and variation is addressed from two perspectives: on the one hand, foreign language (L2) teaching / learning where the target language is the reference norm, and on the other, variation within a native language (L1 varieties). The first two articles address the challenges of speaking L2 English, beginning with Vaissière & Exare’s account of their early learning experiment with young L1 French learners. The method involves the English pronunciation system as a whole, concentrating particularly on what is potentially tricky, so where the L2 reference norm is different from a French learner’s L1 norm. This is followed by Capliez’s article which highlights a specific challenge in English (pronouncing /h/) for young French adults. The third contribution, Kamiyama et al., also deals with learning a foreign language (this time, L2 French), from both production and perception perspectives. They also introduce a different angle on norms / variation, questioning how the Japanese learner’s L1 variety affects performance in L2. The last two articles look at variation in L1, first with Yaseen’s work on Qeltu with a corpus he has set up to investigate traditional perspectives on vowel lowering, questioning generally accepted views and developing understanding of a little researched variety. Finally, Amand looks at variation within a variety of English in the DECTE corpus of Tyneside English, identifying linguistic sub-groups within the variety that are mostly gender defined, and collectively form a continuum from localised to less-localised speech.

36The wide range of contributions to this volume underscores the significance of diversity as a social construct reflected in language. As learners, teachers and researchers we can only accept the need for constant questioning of our values and recognise their relativity: one person’s norm is another person’s variety.