e Issues and Practices (EPIP)

29
L a b o r a t o i r e L a n g a g e s L i t t é r a t u r e s S o c i é t é s Langages Université de Savoie English Pronunciation: Issues and Practices (EPIP) Proceedings of the First International Conference Alice Henderson (ed.) 9

Transcript of e Issues and Practices (EPIP)

Laboratoire

Langages

Littératures

Sociétés

Langages

Engl

ish

Pron

unci

atio

n: Is

sues

and

Pra

ctic

es (E

PIP)

Pro

ceed

ings

of t

he F

irst

Inte

rnat

iona

l Con

fere

nce

Université de SavoieISBN: 978-2-915797-73-2ISSN : 1952-0891

!is book is the fruit of the "rst English Pronunciation: Issues & Practices (EPIP) conference, which took place at the University of Savoie, France, in June 2009. Researchers and teachers from sixteen di#erent countries came together to discuss: phonetic variations and phonological changes; varieties, identity and their implications for teaching; and the use of new technologies in research and in the classroom. !e twelve papers selected for publication re$ect EPIP’s prime objective: the creation of bridges between researchers and teachers from various backgrounds (EFL, ESL, EAP, ESP, language acquisition, etc.). Such sharing is essential for the development of new theories and e#ective teaching methods.

Alice Henderson is responsible for coordinating ESP teaching within the Humanities faculty at the Université de Savoie.

English Pronunciation:

Issues and Practices(EPIP)

Proceedings of the First International Conference

Alice Henderson (ed.)

20 € 9

DIRECTEUR DU LABORATOIRE

Christian Guilleré

COMITÉ DE RÉDACTIONAlice Henderson

COMITÉ SCIENTIFIQUEUna Cunningham, University of Dalarna, Sweden

Heather Hilton, University of Savoie, FranceDavid Levey, University of Cadiz, SpainJan Majer , University of Łódź, Poland

John Osborne, University of Savoie, FranceLinda Shockey , University of Reading,

Rias van den Doel, Utrecht University, #e NetherlandsEwa Waniek-Klimczak, University of Łódź, Poland

Cet ouvrage a été réalisé avec le concours de l’Assemblée des Pays de Savoie,

la Région Rhône-Alpes

SOMMAIRE

Introduction to EPIP publicationAlice HENDERSON ..............................................................................7

Pronunciation Preferences ................................................................... 11

A corpus-based study of phonological free variation in EnglishJose A. MOMPEÁN .............................................................................13

English pronunciation preferences: Research by ‘ indirect’ questionnaireMohamed BENRABAH ........................................................................37

Native & Non-native Speakers ............................................................ 55

!e perception of word stress in English and French: Which cues for native English and French speakers?Dan FROST ..........................................................................................57

!e e"ect of task on the pronunciation of English.High front vowels by Japanese learnersRika AOKI ............................................................................................75

An empirical study of individual di"erences in L2 oral pro#ciency: What makes native-like speakers special?Tanja ANGELOVSKA ..........................................................................95

Teaching Issues ................................................................................... 117

Speaking of speech: Developing metalanguage for e"ective communication about pronunciation between English language teachers and learnersHelen FRASER .................................................................................... 119

Phonology and Moodle: Enhancing pronunciation through learning platform-based training?Angela HAHN .................................................................................... 139

Temporal parameters in the implementation of the voicing contrast in English spoken by Poles: A pedagogical perspectiveArkadiusz ROJCZYK .......................................................................... 159

!e pedagogical implications of variability in transcription: !e case of [i] and [u] Sophie HERMENT ............................................................................. 177

Pronunciation teaching materials in #nnish EFL textbooksElina TERGUJEFF .............................................................................. 189

Future Developments .........................................................................207

How could English truly become a new Latin?Sylwia SCHEUER ...............................................................................209

!e PhonBank initiative and second language phonological development: Innovative tools for research and data sharingYvan ROSE ..........................................................................................223

Author contact details ........................................................................243

13

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

JOSE A. MOMPEÁN

University of Murcia

ABSTRACT

.is paper looks at phonological free variation in English. A corpus of RP English newscasts from 1999 to 2008 was compiled and 52 lexical items known to exhibit phonological free variation were analysed. .e rates of occurrences of the variants and rates of the speakers producing the variants are provided. .e results show that most of the tendencies the data reveal agree reasonably well with data from pronunciation polls carried out for lexicographic purposes. When they do not, explanations are provided for the di/erent results. It is argued that further empirical approaches through corpus analysis and/or speech elicited under experimental, controlled conditions are necessary in the study of this largely neglected phenomenon.

Keywords: phonological free variation, RP English, corpus-based analysis.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

14

Introduction

Phonological free variation is a well-known concept in phonological theory. .e concept refers to the variation in the phonemic and/or stress pattern in a given lexical item (e.g. the variation between !"#!$and !%!$in the 0rst syllable of the word economic).

At least three subcases can be distinguished within phonological free variation: a) two or more phonemes can be found in the same position in a given lexical item (e.g. again !&'(%)!$or !&'(%*)!+,$b)$a phoneme is present or missing from a given position (e.g. !-! in often+; and c) two – or more – stress patterns characterize a given lexical item (e.g. dispute n. !'.*/012#-!3!.*'/012#-!+. .ese phonemes or stress patterns can be referred to as the ‘free variants’. .e latter expression can also be used to refer to each of the pronunciations of a given lexical item.

For the variation in lexical items to be considered as phonological in the sense described in this paper, three conditions must be met. .e 0rst condition is that the potential variants should be associated with one and the same lexical item. .us, homographs with di/erent etymology, meaning and pronunciation like bow /4&5!$– a weapon – and bow /465/ – forward end in vessels – fall outside the scope of phonological free variation. .e second condition is that there should not be any change in the grammatical category of the variants. .is excludes, for instance, the so-called etymologically related word-class pairs like record (n.) !'7%89:.!$ and$ record (v.) !7"'89:.!, di/ering in stress pattern and phonemic structure – typically – but also in grammatical category. Finally, the third condition is that the variation should be a characteristic of citation forms, so it should occur irrespective of the operation of speech processes in connected speech. .erefore, weak forms of function words like and or for, due to processes like vowel or consonant elision, syllabic consonant formation, /r/-liaison, or assimilation cannot be considered as phonological free variation. .is condition also excludes changes in stress patterns due to rhythmic reasons like the so-called stress shift (e.g. academic !;<8&'.%=*8! but academic year !;<8&;.%=*8$'1*&!).

.e existence of phonological free variants is due to di/erent reasons, some of which may also interact. .ese include, amongst others, sound changes, phonetic processes, or cognitive and sociolinguistic/sociocultural factors.

To begin with, ongoing sound changes explain why, at some particular moment in the evolution of a language, two or more variants coexist. For example, the pronunciation !>9#! for sure in British English (henceforth BrE) is increasingly heard at present alongside the more common traditional !>5&! by virtue of the general replacement of the diphthong !5&! by the monophthong !9#! in that variety of English.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

15

Phonological free variants may also be the result of phonetic factors whose results become lexicalised. For example, February is traditionally pronounced !'?%472&7"! – with variations in the vowel in the second syllable – but is now increasingly heard as !'?%412&7"!, a variant likely due to dissimilation or the process by means of which nearby similar or identical consonant sounds, like two r-sounds in this case, become di/erent over time.

Finally, cognitive and sociolinguistic factors may give rise to free variants. .e cognitive process of analogy is a particularly relevant one in this respect. Phonetic analogy justi0es, for instance, why ‘irregular’ stress patterns often change in words to resemble those of phonetically similar words in the mental lexicon. As a case in point, the English su1x –able does not regularly alter the stress of the stem it is attached to (cf. 'pro!t ! 'pro!table, 'recognise ! 'recognisable). Given this, analogy justi0es why, in the word incomparable, a variant with stress on the third syllable, !*)8&='0%&7&4@&+A!, agreeing with the 0nal stress of the underlying verb, is increasingly heard in BrE. .is variant coexists with the traditional variant with stress on the second syllable, i.e. !*)'8B=0@&+7&4@&+A!.

Sound-grapheme analogy, combining cognitive and interrelated sociolinguistic factors like degree of literacy and knowledge/awareness of sound-spelling correspondences in the native language also justi0es the existence of free variation. In this respect, sound-spelling regularisation tendencies explain why variants that follow typical sound/spelling correspondences often appear in English. .us, the spelling pronunciation !'C%*/-8&5-! for waistcoat has largely replaced the former !'C%/8*-D$3&-! in BrE, and$nephew !')%?12#! has also replaced !')%E12#! since the digraph <ph> generally has the phonological value !?!. In contrast, the tendency to divert from typical sound/spelling correspondences also explains the existence of free variants. .is tendency often occurs due to knowledge of sound/spelling correspondences in foreign languages or beliefs about what foreign pronunciations of words might be despite L1 phonological and/or sound-spelling rules. .us an attempt to pronounce the word Munich in English in a manner supposedly closer to the original justi0es the variant !F!$at the end of the word, as if in German – though in German there is no voiceless velar fricative but a palatal one –, together with the more common !'=12#)*8!.

Phonological free variation has so far received little empirical attention, perhaps a consequence of its neglect in phonological theory. .e neglect is due to the fact that the phenomenon has traditionally been considered as a marginal phenomenon associated with speci0c lexical items or small lexical sets unlikely to lead to generalisations. .is may seem surprising for English given the relatively frequent number of items with free variation in this language. An exception, however, is Gimson’s analysis of a pronunciation dictionary’s entries under the A and B letters (Gimson, 1969), which aimed to 0nd out whether variation was more common in

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

16

polysyllables than in monosyllables. More recent studies consist in surveys of pronunciation preferences by written questionnaire as a data-collection device. .ese include Wells’s surveys of pronunciation preferences in BrE for the three editions of his Longman Pronunciation Dictionary – henceforth LPD – (see e.g. Wells, 1990, 1999, 2000, 2008) and the similar surveys by Shitara (1993; in Wells, 2000, 2008) and Vaux (2010; in Wells, 2008). .e motivation of the LPD surveys was to improve the traditional problem of dictionaries in deciding which variant to prioritise by providing “some kind of objective data regarding the relative prevalence of the competing pronunciations of various words” (Wells, 2003, p. 215). .ese data were mainly rates of speakers expressing preferences for the variants of the items the questionnaires contained. Other studies, like Vaux’s, are dialect surveys with the aim of 0nding out what variants people claim they use in an attempt to build up an atlas of English dialects in the US.

A corpus-based study of phonological free variation in EnglishDespite representing a considerable improvement over impressionistic

observations, research by written questionnaire (on-line or by mail) as used by Wells, Shitara or Vaux is limited as an instrument for gathering information about speakers’ pronunciation. Research by written questionnaire may reveal general tendencies but cannot guarantee a faithful re2ection of speakers’ actual usage. For example, informants may not report their linguistic usage accurately due to misperceptions or lack of awareness of their own pronunciation, a bias by their sense of linguistic norm, a misunderstanding of questionnaire information, etc. In addition, research by written questionnaire may fail to accurately pin down the informants’ accent, which may be an important variable to control for. In the LPD surveys, for instance, data were elicited from native speakers from all across Britain (Wells, 1999, p. 33), where there is a high degree of variation across accents. As a case in point, Wells (2008) suggests that although stress on the second syllable of controversy seems to be more common across British accents, in the British accent known as Received Pronunciation – henceforth RP – stress on the 0rst syllable “perhaps still predominates” (p. 182).

When the focus of pronunciation research is on actual usage, analysis of real speech data is necessary. .is can be done through corpus analysis or through speech elicited under experimental conditions. With this consideration in mind, a 200,000 word spoken corpus was compiled in order to conduct an empirical study of free variation in RP. .e aims of the study were two: a) to obtain rates of speakers producing the variants of a number of lexical items; and b) to obtain the rates of occurrence of the instances of ‘free variants’ of those lexical items.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

17

Method

DataNewscasts from the BBC Learning English website (BBC, 2010)

were used for this study. .e archive consists of a set of brief newscasts (usually from around 100 to 300 words) on di/erent topics from January 1999 to the present day. .e archives analysed for this study were from the year 1999 to 2008.

.is resource was used since most speakers are correspondents that have an RP accent, so the resource provides a large amount of spoken data on that variety. However, although the newscasts are delivered by identi0ed RP speakers employed by the BBC in most cases, not all speakers can be described as such. Given this, four criteria were used to decide whether to consider a given newscast or not for analysis. .e 0rst criterion is that the newscast should be read by a professional journalist or correspondent; some speakers were interviewees or had jobs other than the speci0ed ones. .e second criterion is that the name of the speaker should be known; a few speakers were anonymous and were therefore excluded. .e third criterion is that the text has to be available as an audio 0le, which for some newscasts was not the case because of failed links. .e fourth criterion is that the speaker should have an RP accent, which some speakers did not have. Following these criteria guaranteed relative homogeneity as far as accent and sociocultural background are concerned, if that background is de0ned by reference to the speakers’ professional activity, i.e. broadcasting. .ey also guarantee that the precise number of newsreaders considered is known. Following these criteria, 1339 newscasts were considered, amounting to a corpus of RP speech of around 200,000 words.

Since the corpus was not produced under the supervision of the author, not all lexical items exhibiting phonological free variation could be studied since many items occur very few times or not at all. Given this, words were selected as potential candidates for analysis only if they occurred ten or more times in the corpus, irrespective of the number of speakers who produced them. With this arbitrary criterion in mind, 52 words were selected after a preliminary, automatic word search of the number of instances per potential candidate word. Table 1 contains the 0nal list of words studied in alphabetic order.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

18

Table 1Lexical items studied in the BBC 1999-2008 corpus.

againagainstAsiaAsianassociate (v.)associationBeijingCaribbeancigarettecomplex (adj.)contribute (v.)

controversydecadedetaildetaileddirect (adj.)direct (v.)directlydirectordispute (n.)economiceducation

electoralFebruary0nance (n.)0nancialfortuneFridayholidayhurricaneissue (n.)issue (v.)kilometre

medicineMondayMuslimneither oftenpoor produce (v.)PutinSarkozySaturdayschedule

situationSovietSundaysure.ursdayTuesdayWednesdaywithdrawal

Speakers.e speakers in the corpus were identi0ed as male or female by their

name and voice quality. .eir accent was classi0ed as falling within or outside RP, based on features like presence or absence of rhoticity, typical segmental inventories described for RP or phonetic realisation of phonemes well within or clearly without the inherent variation described for the accent. For this study, the speech of 318 newsreaders (222 males, 96 females) was considered to be RP.

Procedure.e corpus was created with a written version and a spoken version.

.e written corpus was created by copying the written version of each newscast from the BBC Learning English website onto another document and creating a database. .e spoken corpus was available online so no further actions were necessary. As a preliminary step to the analysis, the newscasts were listened to in order to check that written and spoken versions coincided and when di/erences were found, the written version was corrected.

.e analysis involved a three-step procedure: a) identi0cation of the target words in the written corpus; b) listening to the relevant audio 0les from the website; and c) recording, for each instance of the words studied, the speaker’s name and variant produced. For the second and third steps, three judges working independently, two native speakers and a non-native speaker, participated in the task. A decision on the variant used was reached when the three judges coincided. In cases of doubt, the opinion of a fourth judge was considered and the items in question were inspected spectrographically with SFS/WASP, a program for speech research developed at University College London.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

19

Results and Discussion

In this section we present the number and rate of instances and speakers per ‘free variant’ of the 52 lexical items studied. .e number of instances and speakers vary from item to item. When rates are given, they represent the total number of speakers producing the items studied. Also, speakers who vary are also considered for both variants. .is is the reason why, when adding up the number of speakers per variant, 0gures may be slightly higher than the number of speakers whose speech is considered. It should be said that within-speaker variation, although found in the corpus, is very rare and can be ignored in the general presentation of results as it does not alter the general tendencies revealed by the data.

.e aim of this section is not to compare the results with the rates obtained in studies such as the LPD pronunciation polls, although they are mentioned when available. When results from those surveys are referred to, the expressions LPD 88, LPD 98 and LPD 08 are used, respectively, to refer to the studies carried out for the 0rst (Wells, 1990), second (Wells, 2000) and third (Wells, 2008) editions of the dictionary. .is includes not only the data regarding BrE but also data from American English (henceforth AmE), from the work of Shitara (1993) and Vaux (2010) reported in the second and third editions of the LPD.

.e items are discussed in 0ve di/erent groups: a) those exhibiting an alternation between a weak vowel and a diphthong in an unstressed syllable; b) those in which one of the variants is the result of a sound change in progress; c) those in which a phonetic factor can produce or has already produced a lexicalized variant; d) those in which there are at least two likely grapheme(s)-phoneme(s) correspondences; and e) those with variation in lexical stress patterns.

Alternation between weak vowels and diphthongs in unstressed syllables .e 0rst group of items discussed is the one exhibiting an alternation

between a weak vowel and a diphthong in unstressed syllables. .e items studied are hurricane, decade, and the lexical sets ending in –day (days of the week plus holiday), direct (v., n., adj., plus the adverb directly and director) and !nance/!nancial.1

.e item decade, stressed on the 0rst syllable, can have two free variants in the second syllable: !%*! or !&!. For these two variants, the results obtained (see Table 2) show that all speakers pronounce decade with !%*!. .is is not surprising since the schwa variant is listed in LPD 08 as the

1 .e items yesterday, direction and (to) !nance were not included since they were underrepresented in the corpus.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

20

fourth and last, and it is a variant that is considered to be associated mainly with the religious sense of ‘part of the rosary, set of ten Hail Marys’ (Wells, 2008, p. 212), a sense not found in the corpus. In contrast, the results for hurricane show that the variant with !&!$is the most common. .is variant was produced by 75% of speakers in 93% of all cases. Curiously, LPD 08 shows a preference for !%*! of 60% of respondents and for !&! of 40%. .e di/erence between the two studies could be that the schwa or ‘weak form’ variant may be more likely in close-knit expressions like Hurricane Katrina, hurricane season, etc. than in exposed positions, where !%*! or the strong form variant could be more common. In the LPD polls, words are decontextualised, so they are in exposed positions while in the corpus studied, hurricane occurs on 12 occasions in non-exposed cases. In all of these cases !&! is used, while in the two exposed positions, !&! is used in one case and !%*! in the other.

Table 2Number and rate of instances and speakers

per variant: decade and hurricane.!"#$!" !"#$%"&'#

()*+,-'%.'/#(01+2

%&''(#$)" !"#$%"&'#3(*4+

,-'%.'/#(4+

!'.%8&.! 0 56 * 06 !3&)! *0 106 0 7)6!'.%8%*.! 48 146 01 *996 !3%*)! * 76 * 2)6

A strong form – weak form relationship like the predicted one between !%*! and !&! in hurricane has also been suggested as an explanation for the alternation between the weak vowel !"! and the diphthong !%*! in compounds ending in –day. According to Wells (2000), !%*! is generally preferred in exposed positions like in I’ ll do it on Monday while the weak vowel is preferred in close-knit expressions such as Monday morning (p. 210). However, although there may be speakers for whom the diphthong-schwa alternation may be in a strong form – weak form relationship, the results obtained (see Table 3) reveal that !%*!$is far more common than !"! in both exposed and non-exposed positions for all the lexical items. In fact, /%*!$ranges from the 84% of instances in Friday to the 100% of cases in Monday or Saturday. Combining the results of the whole lexical set, 157 instances out of a total of 171 were pronounced !%*!$(92%) and 14 !"! (8%). Of those 14 instances, however, 0ve are in exposed positions while nine are in close-knit expressions. In the 156 instances of the diphthong, there are 109 exposed positions and 47 non-exposed ones.

2 One speaker used both variants so he was considered for both variants.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

21

Table 3Number and rate of instances and speakers per variant: -day lexical set.*+)!$, !"#$%"&'#3

(27+,-'%.'/#3(22+

-&".!$, !"#$%"&'#3(09+

,-'%.'/#3(2)+

!3."! 9 96 9 96 !3."! 0 *96 0 *26!3.%*! 27 *996 22 *996 !3.%*! 27 196 20 126

/"!)".!$, !"#$%"&'#(29+

,-'%.'/#(*5+0

-%&'.!$, !"#$%"&'#3(*5+

,-'%.'/#(*)+

!3."! * )6 * 56 !3."! * 56 * 76!3.%*! *1 1)6 *5 *996 !3.%*! *) 146 *) *996

0'(!$, !"#$%"&'#(2)+

,-'%.'/#3(22+

1$2&'!$, !"#$%"&'#(*9+

,-'%.'/#(*9+

!3."! 4 *56 4 *86 !3."! 9 96 9 96!3.%*! 2* 846 *8 826 !3.%*! *9 *996 *9 *996

1&)!$, !"#$%"&'#(27+

,-'%.'/#3(22+

%+3(!$, !"#$%"&'#3(*5+

,-'%.'/#3(*2+

!3."! 4 *)6 4 *86 !3."! * 56 * 86!3.%*! 20 8)6 *1 856 !3.%*! *) 146 ** 126

Moving now on to the lexical set direct, it should be mentioned that it was studied since it is di/erent from word-class pairs of the type abstract (n., adj.) vs. to abstract (v.), since in the latter each member has a speci0c pronunciation associated with it while in direct (adj.) and direct (v.) both items exhibit free variation. Given this, the$ results of our study show that the adjective direct is pronounced by most speakers (69%) and in most cases (71%) with !*!!or !&!!rather than with the diphthong !6*!$(see Table 4).$.e reverse occurs, however, in the case of the verb (to) direct, pronounced by 55% of speakers with the diphthong and in 61% of all instances. .e results obtained show then that there seems to be a higher preference for a short vowel in the 0rst syllable of direct (adj.) and for the diphthong in direct (v.). .e !6*!$variant is also the most common in the adverb directly (61% of speakers, 58% of cases) and the noun director (55% of speakers, 61% of cases).

3 One speaker produced the two variants.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

22

Table 4 Number and rate of instances and speakers per variant: direct lexical set. !('"#2(%:;<+

!"#$%"&'#(*7+

,-'%.'/#(*0+4

!('"#2(=<+

!"#$%"&'#3(*2+

,-'%.'/#(**+

!.*3D$.&3! *2 7*6 1 516 !.*3D$.&3! ) 426 ) 4)6!.6*3! ) 216 ) 086 !.6*3! 7 )86 5 ))6

!('"#23, !"#$%"&'#3(*1+

,-'%.'/#3(*8+

!('"#2+' !"#$%"&'#3(00+

,-'%.'/#(22+

!.*3D$.&3! 8 426 7 016 !.*3D$.&3! *0 016 *9 4)6!.6*3! ** )86 ** 5*6 !.6*3! 29 5*6 *2 ))6

Finally, looking at the items !nancial and !nance (n.), the results (see Table 5) show that the weak vowel variants !*!$or$ !&! are infrequent for !nancial since they are pronounced by 24% of speakers and in 14% of all instances; moreover, !*!$or$ !&! are not found at all in the case of !nance (n.). .ese data are in accordance with data from LPD 98, which shows a 79% preference for the diphthong in !nancial (the LPD does not show data regarding the noun !nance).

Table 5Number and rate of instances and speakers

per variant: !nance and !nancial. 4)$)#($3 !"#$%"&'#

(*9)+,-'%.'/#3(4*+

4)$)#"("<+

!"#$%"&'#3(4)+

,-'%.'/#3(25+

!?*3D$?&3! *) *46 *9 246 !?*3D$?&3! 9 96 9 96!?6*3! 19 856 0* 756 !?6*3! 4) *996 25 *996

Sound changes in progress.e second group of items discussed in this section is that in which one

of the free variants is the result of a synchronic and systematic sound change, and that variant coexists with a traditional variant or variants. As far as RP is concerned, two of the current sound changes a/ecting this accent are: a) the decline and disappearance of the diphthong !5&! and its replacement by the monophthong !9:!; and b) yod coalescence, or the process by means of which an alveolar plosive (!-! or !.!) and a following palatal approximant (!1!) merge to create a post-alveolar a/ricate (!G! or !H!).

As far as the change from !5&! to !9:!$is concerned, the lexical items poor and sure were investigated (see Table 6). In the 0rst case, LPD 88 and LPD 98 show preferences of 43% and 45% for !5&!$and of 57% and 55% for

4 One speaker used both variants so he was considered for both variants.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

23

!9:! respectively. .e increasing prevalence of /9:! is even more noticeable in LPD 08, where the preference for !5&! drops to 26% and that for !9:!$rises to 74%. .e results obtained from the corpus are in line with the tendency revealed by the LPD polls. In the corpus, poor is pronounced !5I!$by 32% of all speakers (in the same rate of instances), con0rming the decline of the diphthong. Regarding sure, the results obtained reveal that the diphthong seems to be declining faster in this item than in poor, since !5&! is only used by 7% of speakers and in 6% of all instances. .e decline of the diphthong in sure shows that one and the same sound change may occur in di/erent lexical items at a di/erent pace. .is is more remarkable since LPD 98 showed !5&! to be preferred by 54% of respondents, although the percentage was lower (40%) for younger respondents, born after 1973.

Table 6 Number and rate of instances and speakers per variant: sure and poor.

poor Instances (62)

Speakers(47)

sure Instances(18)

Speakers(15)

!05I! 20 32% 15 32% !>5I! 1 6% 1 7%!09#! 42 68% 32 68% !>9#! 17 94% 14 93%

.e second sound change in English investigated in this study is that of yod coalescence. .is change has been at work over many centuries in non-stressed syllables, although it also started to spread to stressed syllables in the second half of the twentieth century in BrE (Wells, 1997). In the corpus investigation described in this paper, the items studied for yod coalescence in non-stressed syllables were situation, fortune, schedule and education.

Regarding items where the plosive involved in yod coalescence is voiceless, i.e. situation and fortune, the results show that the a/ricate variant is the most common one (see Table 7). .us, in situation, 78% of speakers used /G!$in 83% of all instances of the item; in fortune the rate rises up to 92% speakers, who produced the a/ricate in 94% of all instances. .e data obtained therefore show a clear predominance of the a/ricate as opposed to the biphonemic plosive + approximant consonant cluster in both situation and fortune. For situation, this is an interesting result given that LPD 98 showed a preference for !-1! of 64% for this item.

Considering the next items where the plosive is voiced like schedule and education, the tendency of yod coalescence to gain ground in non-stressed syllables is also observed (see also Table 7). In schedule, the a/ricate !H! is used by 70% of all speakers and in 70% of all instances. .e data from the corpus con0rm then that yod coalescence is now widespread in schedule. .is is in contrast with data from LPD 98, which showed !.1!$to be preferred by 79% of respondents. In education, however, the data reveal that !H!$is pronounced by 50% of all speakers in 44% of all instances. Although

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

24

the LPD does not provide any data for education, the three editions list !.1! as the 0rst option, accurately capturing the fact that yod coalescence is still not clearly predominant in that item, at least in RP.

Table 7Number and rate of instances and speakers per variant:

yod coalescence in non-stressed syllables..(2&$2(+) !"#$%"&'#3

(47+,-'%.'/#3(02+)

5+'2&)" !"#$%"&'#3(04+5

,-'%.'/#3(25+

!;/*-123! 8 *76 8 2)6 !'?9#-12#)! 4 56 2 86

!;/*G23! 01 806 2) 786 !'?9#G2#)! 02 146 24 126

.#%"!&3" !"#$%"&'#(*9+

,-'%.'/#(*9+

"!&#$2(+) !"#$%"&'#3(27+

,-'%.'/#(22+

!3.13! 0 096 0 096 !;%.123! *) )56 ** )96!3H3! 7 796 7 796 !;%H23! *2 446 ** )96

After looking at yod coalescence in non-stressed syllables, the same process was investigated in the stressed syllables of two items: Tuesday, and the verb (to) produce (see Table  8 for results). In the case of Tuesday, the consonant cluster !-1! is more common than the a/ricate /G/ in the 0rst syllable of Tuesday, where it is pronounced by 60% of all speakers and in 60% of all instances. .e cluster is also the most common variant in (to) produce, where !.1! is pronounced by 70% of speakers, in 60% of all instances. .ese results indicate that yod coalescence continues to widen its scope, extending at present to stressed syllables even in carefully delivered RP speech. However, the fact that biphonemic clusters are still more common in stressed syllables than in non-stressed ones suggests that yod coalescence may still be somehow stigmatized in stressed syllables. An increase can be expected, however, in the use of yod coalescence in stressed syllables in the not too distant future if yod coalescence gains more acceptance amongst speakers.

Table 8Number and rate of instances and speakers per

variant: yod coalescence in stressed syllables-&".!$, !"#$%"&'#3

(09+,-'%.'/#3(2)+

6'+!&#"7(=<+ !"#$%"&'#(79+

,-'%.'/#(41+7

!'-12#J3! *8 596 *) 596 !3.13! 42 596 04 796!'G2#J3! *2 496 *9 496 !3H3! 28 496 *1 016

5 One speaker produced the two variants.6 Two speakers produced the two variants.7 Four speakers produced both variants.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

25

Phonetic factors producing lexicalized variants.e third group of items discussed is that in which one of their free

variants represents the lexicalization of the outcome of a phonetic process other than a systematic sound change. Amongst these processes we can consider: a) dissimilation, by means of which nearby similar or identical consonant sounds become di/erent over time; b) assimilation, by means of which nearby di/erent sounds become similar or identical; c) compression, or the reduction of the number of syllables in a word for articulatory ease; and d) /r/-liaison, or the insertion of an r-sound in between vowels to prevent hiatus. In our corpus, one case of dissimilation (February) was analysed, as well as four cases of assimilation (issue –n.–, issue –v.–, associate, association), one case of compression (medicine) and one case of /r/-liaison (withdrawal),

.e tendency to avoid two r-sounds in adjacent syllables in the lexical item February has led to the appearance of the dissimilation-driven variant !1!$ as an alternative to$ !7! after !4!. .e data obtained reveal that !1! is used by 36% of speakers and in 42% of the cases (see Table 9). .ese data are similar to those from LPD 98, which reveals that !1! was preferred by 39% of respondents (!7! by 61%). Interestingly, according to LPD 98 again, dissimilation seems to have gone further in AmE, where !1! is preferred by 64% of speaker and !7! by 36%. One reason why the !1!$does not have a higher rate in BrE may be that it is somehow stigmatized. Wells (2008, p. 301) claims that !1! in February is sometimes criticized in BrE. .is author also claims that !1! can be heard from educated speakers, particularly in AmE which might explain why the !1!$variant is more common in that accent.

.e opposite process to dissimilation, that is, assimilation, justi0es the existence of a variant in the items issue (n.) and issue (v.). .ese can be pronounced with a voiceless alveolar fricative !/! before palatal !1!, or with a voiceless post-alveolar fricative !>! due to assimilation of the fricative to the adjacent palatal. In the latter case, the palatal may follow the fricative (!>12#!) or may have disappeared altogether (!>2#!). .e data obtained (see Table 9) show, for both the noun and the verb, that the assimilated variant !>!$is$the most common (74% in the noun and 81% in the verb) and that it is also pronounced by most speakers (79% in the noun, 87.5% in the verb). .ese results coincide with the data from LPD 88, which show the !>!$variant to be preferred by 70% of respondents$(!>2#!, 49%, !>12#!: 21%) as opposed to the 30% of !/12#!.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

26

Further cases of variants probably due to assimilation are the items associate and association. In this respect, it should be borne in mind that the grapheme <c> can have the values /s/ and />/, the latter being the rule in the non-stressed syllables of words ending in <-cient> (e.g. pro!cient), <-cial> (e.g. crucial), <-cian> (e.g. musician), <-cious> (e.g. vicious), or <-cion> (e.g. suspicion). .e same may be said of items ending in <-ciate>, where !>! is often the only option (e.g. excruciate) or, apparently, the main variant (e.g. appreciate, depreciate) alongside a second variant with !/!. .is should also be the case with associate, in which both fricative variants are also possible. However, data from LPD 98 show that, in BrE the option with !/! was preferred by 69% of respondents and !>! by 31%. .is may be due to long-distance progressive place assimilation. .e corpus study described in this paper con0rms this tendency (see table 10) as !/! is used by 92% of the speakers in 93% of the instances of associate. Curiously, the presence of an adjacent post-alveolar fricative in the last syllable of the item association does not prevent the long-distance progressive assimilation mentioned above from applying. Dissimilation may also be working in this item, it could be claimed, to prevent the third and 0fth syllables from containing a post-alveolar fricative each.

Table 10 Number and rate of instances and speakers per

variant: associate (v.) and association$..+#($2"7!"#$

!"#$%"&'(*4+

,-'%.'/#(*2+

$..+#($2(+) !"#$%"&'#(*5+

,-'%.'/#(*2+

!I'/I5>"3! * 76 * 86 !I;/I5>"'%*>&)! 9 96 9 96!I'/I5/"3! *0 106 ** 126 !I;/I5/"'%*>&)! *5 *996 *2 *996

Looking now at withdrawal, this lexical item can be pronounced with or without !7! after the long back vowel !9#!. If pronounced, this r-sound, known as ‘intrusive r’, is an epenthetic intervocalic segment due to analogy with the well-known process of ‘linking /r/’. .e results of the corpus (see Table 11) show that the variant with /r/ is only produced in 10%

8 One speaker produced the two variants.

Table 9Number and rate of instances and speakers per

variant: February, issue (n.) and issue (v.).0"8'&$',7 !"#$%"&'#

(*2+,-'%.'/#(**+

(..&"("<+

!"#$%"&'#(*97+

,-'%.'/#(52+8

(..&"(=<+

!"#$%"&'#(25+

,-'%.'/#(24+

!'?%413! ) 426 4 056 !'*/12#! 28 256 *4 206 !'*/12#! ) *16 0 *2<)6!'?%473! 7 )86 7 546 !'*>@1+2#! 71 746 41 716 !'*>@1+2#! 2* 8*6 2* 87<)6

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

27

of cases, so its lexicalisation is very weak. .e reason why intrusive /r/ in an item like withdrawal seems to be so infrequent could be that the intrusive r-sound is unetymological and not justi0ed by the spelling. Given this, it has traditionally been regarded as a vulgarism, its use – but not that of linking /r/ – being more strongly stigmatized word-internally than across word boundaries (Wells, 1982, p. 225). .is may be the reason why intrusive /r/ is less common than linking /r/ in corpus studies (e.g. Mompeán & Mompeán, 2009) when, phonetically speaking, they are identical processes.

.e last item studied in this group is medicine, which can be pronounced with two main vowel variants in the second syllable, !*!$or !&!,$between$ !.! and$ !/!$or with no vowel. .e consequence of the free variant with no vowel between !.! and !/! results in the compression or the reduction, in this case, of a trisyllabic word to a bisyllabic one. In the corpus analysed (see Table 11), the bisyllabic variant is produced by 67% of speakers and also in 67% of all instances, showing that compression is common in this lexical item. .e editions of the LPD show the compressed variant to be the 0rst option although no data are provided from pronunciation polls.

Table 11Number and rate of instances and speakers per

variant: withdrawal and medicine.withdrawal Instances

(10)Speakers

(8)medicine Instances

(12)Speakers

(12)!C*K'.79#7&A!$ 1 10% 1 12.5% !'=%.*/3D$3.&/3! 4 33% 4 33%!C*K'.79#&A! 9 90% 7 87.5% !'=%./3! 8 67% 8 67%

Grapheme(s)-phoneme(s) correspondences .e fourth group of items discussed in this section is that in which a

particular grapheme or graphemes can be pronounced in two di/erent ways given sound-to-spelling correspondences in the language. However, typically, one variant is less frequent as a sound value for the grapheme(s) while the other variant is more frequent. In this section we look at seven items where the variation involves vowel sounds and seven items where the variation involves consonants. All items involve free variation in stressed syllables.

Looking 0rst at vowels, the items discussed are two pairs of etymologically related items (again-against, either-neither) and the items economic, Soviet and Muslim. Regarding the 0rst pair of etymologically related items, again-against, the variation is to be found in the second syllable of both items, where the vowel can be !%!$or !%*!. .e results obtained show that !%!$is the most common variant, pronounced by 85% of speakers and in 85% of all cases (see Table 12). .ese results are in line with data from LPD 88, which reported a preference of 80% for !%! and of 20% for !%*!$in

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

28

BrE, and of 97% for$!%!$and 3% for$!%*!$in AmE. Given that the spelling forms <ai> and <ay> can have the value !%! (e.g. said, says) but most typically !%*!$(e.g. pain, raid, day, ray, etc.), the results obtained may be taken as a resistance to a more regularized, spelling pronunciation. Regarding against, the results of the corpus study show that !%! is also the most frequent variant, as it is pronounced by 71% of the speakers in 66% of instances. Although to a lesser degree than in again, the same explanation can be o/ered for this item. .e LPD does not provide data for against, although it lists the variant with !%! as the 0rst one.

Regarding either and neither, which have the variants !":! and !6*! in the 0rst syllable, the results reveal that the diphthong is the most common variant (see Table 12). In either, 83% of speakers used !6*!, in 85% of all instances, and in neither, 77% of speakers (also in 77% of the cases). .ese results mirror those in LPD 88, where !6*!$in either was preferred by 88% of speakers in BrE (no data are provided for neither). In terms of sound-spelling correspondences, it could be claimed that both sound values are infrequent for the digraph <ei>, which typically has the value !%*!. .ere are English lexical items where <ei> has the value !6*!, mainly words of foreign origin like zeitgeist or Heinz, but also words with !":! like receive, conceive, perceive, etc. Assuming !":! for <ei> is more ‘regular’ in the sense that it is not perceived as foreign and is found in lexical items not perceived as recent borrowings, the predominance of !6*! in (n)either may be regarded as a resistance to sound-spelling regularization. However, data from LPD 88 show that in AmE, !":! was preferred by 84% of speakers, indicating a stronger tendency to sound-spelling regularization in that variety of English.

Table 12Number and rate of instances and speakers per

variant: again-against, either-neither.$9$() !"#$%"&'#

(57+,-'%.'/#(48+

$9$().2 !"#$%"&'#(*)*+

,-'%.'/#3(12+1

!&'(%)! )7 8)6 4* 8)6 !&'(%)/-! *99 556 5) 7*6!&(%*)! *9 *)6 8 *76 !&'(%*)/-! )* 046 21 026

"(2%"' !"#$%"&'#(27+

,-'%.'/#(20+

)"(2%"' !"#$%"&'#(*0+

,-'%.'/#3(*0+

!'":L&! 4 *)6 4 *76 !')":L&! 0 206 0 206!'6*L&! 20 8)6 *1 806 !')6*L&! *9 776 *9 776

Next we look at the cases of economic, Muslim, Soviet and Putin (see Table 13 for results). Firstly, in economic, the two variants in the 0rst syllable, !":! and !%!, represent common sound values of <e> in word-initial stressed syllables (e.g.$!":! in even; !%! in ever). .e results obtained reveal that !%!$is pronounced by 56% of the speakers, and in 53% of all instances. .ese

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

29

results di/er from the LPD 88 data, in which 62% of respondents showed preference for !":!, and 38% for !%!. .e di/erence might be one of accent in the sense that /e/ might be more common in RP while, across British accents, /":! might predominate. Alternatively, !%! may have been gaining ground since the late 80’s, when the LPD 88 was conducted.

Secondly, in Muslim, the letter <u> in the 0rst syllable can have the values /5/ or /M/, two common values of <u> in stressed syllables. Of these, /M/ is the most common (e.g. cut, much), while /5/ is found in fewer lexical items, although some of which are very frequent, like push, put, etc. .e results indicate, however, that /5!$is the only variant found in Muslim. .e LPD 98 polls showed that /5/ was chosen by 70% of speakers, and /M/ by 30%. It seems though that /M/ in Muslim has probably disappeared in BrE as a common pronunciation, which explains why, in the third edition of LPD, /M/ is no longer o/ered as a variant. Curiously, in AmE !M! seems to be the preferred variant, according to LPD 98.

.irdly, Soviet can have !B! or !I5!$as a vowel variant in the 0rst syllable. Both are typical$ pronunciations of <o> when followed by a consonant in two or three-syllable words (e.g. !B! in soccer, !I5!$in sober). .e results show that the diphthong was pronounced by 67% of all speakers and in 62% of all instances. .e data obtained are in line with the data from LPD 88, which revealed that 73% of respondents chose !I5!$– and 27% !B!$–.

Finally, in Putin, the letter <u> can have the values !12#! or !2:!. Although, from a phonological point of view, the di/erence between the two variant pronunciations is typically considered to be the presence or absence of an approximant consonant in the head of the 0rst syllable, the variation is considered here as a phonetic vowel alternation with a di/erent phonological interpretation: presence of a monophthong or presence of a rising diphthong. .e results show that the !2#! variant is pronounced by all speakers on all occasions. One explanation for the results obtained is that, since BBC correspondents are supposed to know about the world a/airs and the culture they talk about, the former may try to pronounce proper nouns as closely as possible to the original, in this case, a pronunciation in Russian that has no yod.

9 Two speakers produced the two variants.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

30

Table 13 Number and rate of instances and speakers per

variant: economic, Muslim, Soviet and Putin."#+)+:(# !"#$%"&'#

(*)2+,-'%.'/#(79+

*&.3(: !"#$%"&'#(04+

,-'%.'/#(*8+

!;"#83! 7* 476 0* 446 !M! 9 96 9 96!;%83! 8* )06 01 )56 !5! 04 *996 *8 *996

1+;("2 !"#$%"&'#(21+

,-'%.'/#(*8+

<&2() !"#$%"&'#(04+

,-'%.'/#(*)+

!B! ** 086 5 006 !12#! 9 96 9 96!I5! *8 526 *2 576 !2#! 04 *996 *) *996

Looking next at consonants, seven lexical items were considered. In four cases, the items involve a variation in feature speci0cations of the consonant segment involved in free variation, like voicing (Asia, Asian, and Muslim) or manner of articulation (Beijing). .e variation in the other three items involves the presence of a single consonant vs. that of a two-consonant cluster (schedule) or the presence or absence of an epenthetic, spelling-motivated consonant (often).

For items where the variation implies a change in one of the features (see Table 14), the results show that in Asia, where there is variation between a voiceless post-alveolar fricative !>! and its voiced counterpart !N!, the latter is the most common variant as it is pronounced by 89% of speakers in 89% of all instances. .ese data support the tendency revealed by the LPD surveys for the use of !N! to increase since, for Asia, both LPD 98 and LPD 08 show !N! was preferred by 51% and 64% of respondents respectively. Moreover, in the case of Asian, for which LPD does not present data, !N!$is the only variant used. One reason for the prevalence of the voiced fricative variant in Asia (and Asian by extension) may be analogy with words where <s> represents !N! in non-stressed, word-0nal syllables ending in schwa, i.e. /N&/, like erasure, pleasure, leisure, closure, etc., and with which Asia rhymes.

Another voicing variation can be found in the item Muslim, which can be pronounced with !/! or !J!. .e data obtained reveal that the voiced fricative is predominant, since it is produced by 89% of speakers and in 91% of the instances. .ese 0ndings are in line with the results of the LPD 98 which revealed a preference for !J! of 89%. Finally, Beijing can be pronounced with a voiced post-alveolar a/ricate !H!D$following typical sound-grapheme correspondences in English, or fricative !N!D$as if foreign, although there is no justi0cation from Chinese. .e data reveal that 56% of speakers, and in 62% of the instances, pronounce the a/ricate variant, but the foreign-sounding fricative is also widespread.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

31

Table 14Number and rate of instances and speakers per variant: voicing alternation in Asia, Asian, Muslim and Beijing.

=.($ !"#$%"&'#())+

,-'%.'/#(28+*9

=.($) !"#$%"&'#(28+

,-'%.'/#(*7+

!>! 5 **6 4 *46 !>! 9 96 9 96!N!$ 41 816 2) 816 !N! 28 *996 *7 *996

*&.3(: !"#$%"&'#(04+

,-'%.'/#(*8+

>"(?()9 !"#$%"&'#3(07+

,-'%.'/#(*8+

!/! 0 16 2 **6 !N!$ *4 086 8 446!J! 0* 1*6 *5 816 !H! 20 526 *9 )56

Moving next to an item where free variation involves the presence, word-initially, of a single consonant vs. that of a two-consonant cluster, i.e. schedule, the data obtained (see Table 15) show that a voiceless post-alveolar fricative !>!$is more common than the biphonemic consonant cluster$!/8!. .e data obtained, which show that !>! is pronounced by 80% of speakers in 80% of all instances, agree reasonably well with those from LPD 98, where 70% of respondents preferred !>!.$Curiously, the LPD 98 data also showed that !/8!$was chosen by 60% of those born after 1973. However, since the age of the BBC correspondents is not known, it is not possible to tell whether the variant !/8! in the corpus is less frequent because the speakers studied are older or because !>! is simply more common in RP at present irrespective of the speaker’s age.

Finally, considering an item like often, where free variation implies the presence or absence of an epenthetic, spelling-motivated consonant, the results show that although 67% do not pronounce a t-sound in the second syllable, the spelling pronunciation with /t/ is relatively common: 36% of speakers and in 42% of all cases (see Table 15). .is is so despite the fact that, traditionally, the variant with !-! has been considered as a vulgarism. .is variant has a higher rate than in the data from LPD 88 (27% of speakers), which indicates that this variant is gaining acceptance, even in RP and in careful, broadcast speech.

Table 15Number and rate of instances and speakers per variant: schedule and often.

schedule Instances(10)

Speakers(10)

often Instances (55)

Speakers(39)11

!>3! 8 80% 8 80% !'B?-&)! 23 42% 14 36%!/83! 2 20% 2 20% !'B?&)! 32 58% 26 67%

10 One speaker produced the two variants.11 One speaker produced the two variants.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

32

Variations in stress pattern.e last group of items discussed is that in which items exhibit stress

pattern variation. Typically there are two patterns although some items may even have three patterns (e.g. Sarkozy). .e eleven items studied in this group were contribute, controversy, cigarette, kilometre, Caribbean, dispute (n.), complex (adj.), detail, detailed, electoral, and Sarzoky.

We look 0rst at words of three or more syllables where the variation involves word-initial vs. non-word-initial stress. .e items studied are contribute, controversy, cigarette and kilometer (see Table  16 for results). Firstly, the results show that in contribute, stress on the second syllable is more common than stress on the 0rst syllable (82% of speakers in 78% of all cases). .is parallels data from LPD 88 and LPD 08, which reveal a preference for stress on the second syllable of 73% and 59% respectively. .e LPD 08 data suggests, though, that word-initial stress in contribute may be gaining ground across Britain, although RP may be resisting it. Secondly, regarding controversy, the data show that word-initial stress is the most common variant: 73% of speakers in 75% of all instances. .ese data di/er from those of LPD 88 and LPD 98, where word-initial stress was preferred only by 44% and 40% respectively, and they con0rm the impression in Wells (2008) that although stress on the second syllable is currently more widespread in BrE, word-initial still predominates in RP (p. 182). .irdly, in cigarette, the data show that 56% of speakers produced stress on the 0rst syllable in 50% of the cases and the same applies to word-0nal stress, the predictable stress according to the type of su1x. .e data from the LPD 88 show word-0nal stress to be preferred by 85% of respondents in BrE. .erefore, the data obtained from the corpus suggest antepenultimate stress in cigarette has gained ground in recent times and is now the predominant variant, like in AmE, where word-initial stress seems to be predominant – 65% – according to LPD 98. Finally, in kilometre, the data reveal that 77% speakers used stress on the second syllable in 74% of all instances of the word. .is agrees with the increase revealed by the LPD polls in the use of stress on the second syllable in BrE, preferred by 48% (LPD 88), 57% (LPD 98), and 63% (LPD 08). Finally, word-initial stress in Sarkozy is the least common variant, with stress on the second syllable and the last syllable, resembling the original French pronunciation, as the most common ones: 50% and 42% of all instances respectively.

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

33

Table 16Number and rate of instances and speakers per variant: contribute, controversy, kilometre, cigarette and Sarkozy.

#+)2'(8&2"!"#$

!"#$%"&'#3(*8+

,-'%.'/#(*7+*2

#+)2'+;"'., !"#$%"&'#3(*2+

,-'%.'/#(**+

!'8B)-7*412#-! 4 226 4 246 !8&)'-7BE&/"! 0 2)6 0 276!8&)'-7*412#-! *4 786 *4 826 !'8B)-7&EO#/"! 1 7)6 8 706

@(3+:"2'" !"#$%"&'#3(0*+

,-'%.'/#(25+

#(9$'"22" !"#$%"&'#3(*2+

,-'%.'/#(1+*0

!'8*A&="#-&! 8 256 5 206 !;/*(I'7%-! 5 )96 ) )56!8*'AB=*-&! 20 746 29 776 !'/*(I;7%-! 5 )96 ) )56

1$'@+A, !"#$%"&'#3(*2+

,-'%.'/#(7+

!'/P#8IJ"!$ * 86 * *46!/P#'8I5J"!$ ) 426 0 406!/P#8I5'J"#! 5 )96 0 406

We consider next various items in which free variation exists between word-initial and word-0nal stress in bisyllabic words like dispute (n.) or detail. .e data for these are displayed in Table 17. In the case of dispute (n.), stress on the second syllable is the most common variant: 96% of the instances, pronounced by 95% of speakers. In contrast, all speakers, in all cases, use word-initial stress in detail. .e related adjective detailed, also shows word-initial stress to be the most common variant: 88.5% of speakers and cases. .e case of detail also reveals an interesting fact: some of the variants studied in the LPD pronunciation polls or listed in dictionaries are extremely uncommon in RP or simply do not occur in this accent. .us, items like electoral, which LPD 08 shows elicits a 17% preference for penultimate stress in BrE, is never pronounced with that stress pattern in the corpus. Similarly, the adjective complex never occurs with 0nal stress and Caribbean is never found with stress on the second syllable. LPD 08 shows this option to be uncommon anyway, with 9% of respondents. .e four instances of Caribbean with stress on the 0rst syllable !'8<7&;4"&)!3following stress shift were treated as cases of underlying stress on the third syllable<

12 One speaker produced the two variants.13 One speaker produced the two variants.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

34

Table 17Number and rate of instances and speakers per variant: dispute

(n.), detail, detailed, electoral, complex (adj.), Caribbean. !(.6&2"!%#$

!"#$%"&'#3(28+

,-'%.'/#(22+

!"2$(3 !"#$%"&'#3(21+

,-'%.'/#(22+

!'.*/012#-! * 46 * )6 !.*'-%*&A! 9 96 9 96!.*'/012#-! 27 156 2* 1)6 !'."#-%*&A! 21 *996 22 *996

!"2$(3"! !"#$%"&'#3(8+

,-'%.'/#(8+

"3"#2+'$3 !"#$%"&'#3(*2+

,-'%.'/#(*2+

!.*'-%*&A.! * *2<)6 * *2<)6 !3A%8'-9#3! 9 96 9 96!'."#-%*&A.! 7 88<)6 7 88<)6 !3'A%8-3! *2 *996 *2 *996

#+:63"B7(%:;<+

!"#$%"&'#3(*)+

,-'%.'/#(*0+

C$'(88"$) !"#$%"&'#3(**+

,-'%.'/#(*9+

!8I='0A%8/! 9 96 9 96 !8&'7*4"&)! 9 96 9 96!'8B=0A%8/! *) *996 *0 *996 !;8<7&'4":&)! ** *996 *9 *996

Conclusion

Phonological free variation, or the variation in the phoneme make-up and/or stress patterns of a lexical item, has been the focus of study in this paper. For English, this phenomenon has been investigated mainly through written questionnaire research as a data-collection device, in an attempt to obtain objective data regarding the relative prevalence of free variants for lexicographical or dialectal survey purposes. However, it was claimed in this paper that analysis of real speech data is necessary in the study of phonological free variation since this provides more objective evidence about the relative prevalence of free variants. It was argued then that this empirical approach could be made through corpus analysis or through speech elicited under experimental, controlled conditions.

.is paper has adopted corpus analysis, or the study of language through samples (corpora) of ‘real world’ texts, as its empirical approach. For this, a relatively large corpus of RP English was compiled in order to obtain rates of speakers producing the free variants of speci0c lexical items as well as rates of instances per free variant in those items.

.e results obtained agree with observations in the specialized literature of phenomena like the increasing use of yod coalescence in stressed and non-stressed syllables, the disappearance and replacement of !5&! by !9:!, etc. Furthermore, comparing the results with those found in the three editions of the LPD, most of the data and the tendencies revealed by the corpus agree reasonably well with those found in the three editions of the LPD. When di/erences are observed, they can be attributed, for instance, to real speech observations that have so far passed unnoticed

A CORPUS-BASED STUDY OF PHONOLOGICAL FREE VARIATION IN ENGLISH

35

(e.g. the widespread nature of !%*! in the –day lexical set irrespective of the position of the item in the utterance) or to diachronic changes in the space of two decades. Di/erences can also be attributed to cross-accent di/erences since, it should be remembered that the data from the corpus study reported here represent the performance of speakers from a very speci0c professional background and accent while the LPD data re2ect opinions of speakers of various social backgrounds from all over Britain.

Despite the evidence gathered, this study has certain limitations due, essentially, to the limited control the researcher could have over the type and number of items that could be studied. For instance, the corpus analysed did not permit investigation of certain lexical items which were underrepresented or not found at all. Another limitation is that diachronic change could not be investigated since the year of birth of the speakers is not available on the BBC Learning English website, so conclusions about associations between pronunciation variants and newsreaders’ age cannot be made. Contacting the BBC or speci0c newsreaders for this information is a measure that could be taken to include the age variable in further interpretations of the data provided.

All these limitations suggest directions for future research, which could include further corpus studies, expanding the corpus used in this study and/or using other corpora. .is may be useful to study new items and obtain stronger results for those discussed in this paper. Further action could include the design of experimental studies that control variables like age, gender, accent, social background, etc. and their e/ect on the use of the di/erent free variants of lexical items. Sentences could be constructed, for instance, to elicit pronunciations from speakers or words might be elicited in other ways too. Also, to study di/erences between accents, for instance, a corpus of AmE speech from some private or public broadcasting company or a set of sentences elicited from AmE could be used, the results of which could then be compared with similar data from BrE, just as Wells does in the second and third editions of the LPD. .ese future empirical studies will shed further light on the largely neglected phenomenon of phonological free variation.

ENGLISH PRONUNCIATION: ISSUES AND PRACTICES

36

REFERENCES

BBC. (2010). BBC Learning English website. Retrieved March 25, 2009 from http://www.bbc.co.uk/worldservice/learningenglish/index.shtml

Gimson, A. G. (1969). A note on the variability of the phonemic components of English words. Brno Studies in English, 8, 75-79.

Mompean, J. A., & Mompean-Guillamón, P. (2009). /r/-liaison in English: An empirical study. Cognitive Linguistics, 20/4, 733–776.

SFS/WASP. (Version 1.41). (2009). [Computer software]. Retrieved March 25, 2009 from http://www.phon.ucl.ac.uk/resource/sfs/wasp.htm

Shitara, Y. (1993). A survey of American pronunciation preferences. Speech Hearing and Language, 7, 201-232.

Vaux, B. (2010). Dialect survey. Retrieved March 25, 2009 from http://www4.uwm.edu/FLL/linguistics/dialect/maps.html

Wells, J. C. (1982). Accents of English. Cambridge: Cambridge University Press.Wells, J. C. (1990). Longman Pronunciation Dictionary (1st ed.). Harlow: Pearson

Education.Wells, J. C. (1997). Whatever happened to Received Pronunciation? In C. Medina-

Casado & C. Soto-Palomo (Eds.), II Jornadas de Estudios Ingleses (pp. 19-28). Jaén, Spain: Universidad de Jaén.

Wells, J. C. (1999). British English pronunciation preferences: a changing scene. Journal of the International Phonetic Association, 29(1), 33-50.

Wells, J. C. (2000). Longman Pronunciation Dictionary (2nd ed.). Harlow: Pearson Education.

Wells, J. C. (2003). Pronunciation research by written questionnaire. Proceedings of the 15th International Conference of Phonetic Sciences. Barcelona, Spain. 215-218.

Wells, J. C. (2008). Longman Pronunciation Dictionary (3rd ed.). Harlow: Pearson Education.

Laboratoire

Langages

Littératures

Sociétés

Langages

Engl

ish

Pron

unci

atio

n: Is

sues

and

Pra

ctic

es (E

PIP)

Pro

ceed

ings

of t

he F

irst

Inte

rnat

iona

l Con

fere

nce

Université de SavoieISBN: 978-2-915797-73-2ISSN : 1952-0891

!is book is the fruit of the "rst English Pronunciation: Issues & Practices (EPIP) conference, which took place at the University of Savoie, France, in June 2009. Researchers and teachers from sixteen di#erent countries came together to discuss: phonetic variations and phonological changes; varieties, identity and their implications for teaching; and the use of new technologies in research and in the classroom. !e twelve papers selected for publication re$ect EPIP’s prime objective: the creation of bridges between researchers and teachers from various backgrounds (EFL, ESL, EAP, ESP, language acquisition, etc.). Such sharing is essential for the development of new theories and e#ective teaching methods.

Alice Henderson is responsible for coordinating ESP teaching within the Humanities faculty at the Université de Savoie.

English Pronunciation:

Issues and Practices(EPIP)

Proceedings of the First International Conference

Alice Henderson (ed.)

20 € 9