LINGUIST List 23.2780

Wed Jun 20 2012

Review: Phonetics; Sociolinguistics: Van der Harst (2011)

Editor for this issue: Rajiv Rao <>

Date: 20-Jun-2012
From: Chiara Meluzzi <>
Subject: The Vowel Space Paradox
E-mail this message to a friend

Discuss this message

Announced at
AUTHOR: Van der Harst, SanderTITLE: The Vowel Space ParadoxSUBTITLE: A Sociophonetic Study on DutchsSERIES TITLE: LOT dissertation seriesPUBLISHER: LOT (Landelijke Onderzoekschool Taalwetenschap)YEAR: 2011

Chiara Meluzzi, University of Pavia-Free University of Bozen, Italy


This book deals with a sociophonetic analysis of the Dutch vowel system. Thebook is divided into eight chapters: Chapter 1 is a general introduction to thework and its aims; Chapter 2 offers a wide review of the state of the artconcerning both speech styles and the acoustic analysis of the Dutch vowelsystem; Chapter 3 presents the data set; Chapter 4 deals with four mainmethodological problems in the acoustic analysis of vowels; Chapters 5 and 6present the analysis of the vowel space of Dutch in word list data andspontaneous speech data, respectively; Chapter 7 compares the results of the twoprevious chapters; and Chapter 8 sums up the main findings of the work.

In Chapter 1, van der Harst situates his study in the field of sociophonetics,defined as being “at the interface of phonetics and sociolinguistics” (p. 1).The author illustrates the seven main aims of his work, which are all classifiedas being either general, theoretical, methodological or descriptive. The generalaim of the book is to demonstrate “the fruitfulness of integratingsociolinguistics and phonetics” (p. 4). The theoretical aim is to solve theso-called Vowel Space Paradox that concerns acoustical vowel space variation indifferent speech styles. This paradox emerges in the acoustical analysis ofvowels in different speech styles because “whereas the variation between vowelsdecreases in spontaneous speech, the variation within vowels increases, whichdoes not lead to large increases in vowel confusions” (p. 5). The fourmethodological aims are sociolinguistic (one) and phonetic (three) in nature: onthe one hand, the author proposes a method to elicit standard language in threedifferent speech styles; on the other hand, the author wants to find the bestmethod for formant measurements, normalization and temporal representation ofvowels. Finally, the descriptive aim of the book is to describe thesociogeographic variation of Standard Dutch, in particular, between Netherlandicand Flemish Dutch.

Chapter 2 is a wide review of the state of the art concerning both speech stylesand the analysis of Dutch vowels. Three main definitions of “style” have beenfound in the literature: a continuum from formality to informality (e.g. Labov2001); audience design (Bell 1984); an act of identity of the speaker (Meyerhoff2006) relating to the context of the interaction and the speaker’s communicativegoals. Van der Harst decides to follow a classic Labovian approach to style,since it has been demonstrated that intraspeaker variation found with Labov’smethod corresponds with interspeaker variation. The review of literatureconcerning vowel space reduction leads to the formulation of three researchhypotheses (pp. 48-52): (1) In spontaneous speech, the vowel space is reduced;(2) In spontaneous speech, the sociogeographic variation within a vowelincreases; (3) Vowels which show large sociogeographic variation in F1 or F2dimension in formal tasks are not supposed to reduce in that dimension inspontaneous speech. The three hypothesis are tested in Chapter 7 in order tosolve the Vowel Space Paradox.

Chapter 3 presents the data set, collected in 1999-2000, for a widersociolinguistic project on the pronunciation of Standard Dutch in theNetherlands and in Flanders. For his work, van der Harst analyzed the speech of160 high school teachers of Dutch. The corpus was stratified for two communities(the Netherlands and Flanders), eight regions (four for each community), two agegroups (22-40 and 45-60 years old), and finally, for gender. In this chapter,the author also discusses the first methodological problem of the book, i.e.,how to elicit different speech styles of standard language in a sociolinguisticinterview. For this purpose, each speaker was asked to perform different tasksin order to elicit different speech styles; the more formal tasks were logatome(i.e. monosyllabic pseudoword) reading and word list reading, while the informaltask was a short interview. Logatome reading and word list reading elicited4,800 and 4,640 tokens, respectively. The data from the logatome reading taskwere previously analyzed by Adank (2003), whose findings are summed up and usedin van der Harst’s analysis in Chapter 5. The word list data consisted of 319monosyllabic Dutch lexemes containing the 15 full vowels and the threediphthongs of Dutch in two opposite phonological contexts, i.e., with the targetvowel followed by /s/ (the /s/-context) and by /t/ (the /t/-context). After themore formal tasks, spontaneous speech was collected through 15 minute interviewson various topics. These more informal data were analyzed only for a subset of20 speakers, producing a total of 1,426 tokens.

Chapter 4 deals with the other three methodological aims of the book, since theauthor wants to define an acoustical analysis which “is to be considered theproper one for the seg(socio)phonetic study of vowel variation” (p. 65). Even ifother approaches are available for the acoustic study of vowels (e.g. the wholespectrum approach), the author prefers a formant analysis approach. The firstmethodological problem is to define the best method to obtain valid formantvalues. Formant analysis is based on Fant’s (1960) Source-filter Theory,according to which the speech signal is the product of independent components,i.e., the source and the filter; in the production of vowels, vocal foldvibration is the source and the vocal tract acts as a filter. Formants areusually estimated using Linear Predictive Coding (LPC) with a fixed number ofLPC coefficients (i.e. the default method) or by varying the number of LPCcoefficients according to both speaker and vowel (i.e. the adapted method).Although the adapted method yields better automatic measurements, in particular,for back vowels, van der Harst states that the default method equally gives goodresults (p. 89). The second methodological issue is to find the best formantnormalization procedure for the sociophonetic analysis of vowels. Such aprocedure should indeed minimize the anatomical variation related to gender andage of the speaker, while preserving phonemic and sociolinguistic variation. Theauthor statistically compares 17 normalization procedures. He concludes thatLOBANOV (Lobanov 1971) is the most effective at removing anatomical differencesand preserving sociolinguistic variation (p. 121), and is therefore used in theremaining part of the book. The author also investigates the best temporalrepresentation of vowels in sociophonetic analyses of vowel variation. Theproblem is defining the appropriate number of time points in which to analyzeformant values in order to properly describe a vowel. Two main approaches areevaluated by the author, i.e., target approaches and dynamic approaches. Thefirst is the traditional approach used in many phonetic and variationiststudies; formant values are analyzed at single points in time (namely, themiddle of a formant transition) for monophthongs and at two time points fordiphthongs. The dynamic approach is mainly used in forensic linguistics andincorporates more information about vowel dynamics. The author analyzes twodifferent dynamic approaches: the time points approach measures monophthongs anddiphthongs at more than one or two time points, respectively, while theregression approach uses regression coefficients to describe the development offormants over time. The author demonstrates that the dynamic approach, and inparticular, the time point approach, improves the sociophonetic analysis offormants of both monophthongs and diphthongs. However, in different phonologicalcontexts, the effect of coarticulation on the dynamic vowel representationremains unclear. Due to this possible effect of phonological context ondifferent vowel tokens, the author decides to follow the traditional (i.e.default) method by analyzing the midpoint of monophthongs and the onset andoffset of diphthongs.

In Chapter 5, the analysis of data begins with the tokens collected from the twoformal tasks, which are supposed to focus the attention of the speaker onhis/her production. Two different kinds of variables are taken into account inthe statistical analysis of data: a phonological variable, considering the roleof the coda of the syllable; and social variables such as gender, age andcommunity of the speaker. The analysis of phonological context reveals that “ins-words the vowel was considerably longer than in t-words” (p. 149). A followingstatistical analysis was conducted on formants (F1 and F2), considering bothphonological and social variables and the different interactions between them.For F2 dimension, the vowel space of Flemish speakers appears to be smaller thanthat of Netherlandic speakers; for the author, this means that “the Netherlandicspeakers pay extra care to their speech” (p. 199), as has been previously notedin literature. Another community difference is the lower onset of Netherlandicdiphthongs, which, according to Jacobi (2009), could be referred to as a PolderDutch characteristic, because in this variety of Dutch, diphthongs arepronounced with a wider mouth opening. Even if it is commonly claimed thatPolder Dutch features are used mainly by young, highly-educated women (Stroop1998), no relation between the two variables of gender and community was foundin the author’s data set. At the regional level, vowel data showed a differencebetween Flemish and Netherlandic Dutch. The two regions seem, however, “to crossthe border” (p. 201); Netherlands-South vowels are indeed similar to Flemishones, while Flander-Brabant shows similarities with Netherlandic Dutch.

Chapter 6 analyzes data from spontaneous speech in two Netherlandic communities:the central community Netherlands-Randstad (N-R), and the peripheral regionNetherlandic-South (N-S). Spontaneous speech is defined as “speech that isunscripted and unprepared to a large extent” (p. 203), which means thatattention paid to speech is assumed to be lower than in reading tasks. Thestatistical analysis confirms that phonological context (i.e. the coda ofsyllables) influences vowel quantity but that these differences do not clearlycorrelate with social variables. For /ε/ and /ʏ/, an important communitydifference has been found; N-S vowels are longer and more open than N-R vowels,even if N-S vowel space tends to be more symmetrical than that of N-R. Theauthor gives a sociolinguistic explanation of his findings, claiming that “N-Svowels may shift in the direction of the vowels of the standard variety culturaland economic centre of the Netherlands (i.e. N-R)” (p. 229). However, no age orgender differences were found for these vowels. Only the diphthong /œy/ showed agender difference within the N-R region; women showed a more open onset thanmen, and the author interprets this fact as confirmation of the rise of PolderDutch vowels being led by women.

Chapter 7 compares the values and findings of the different vowels in the threespeech styles in order to solve the Vowel Space Paradox. The three hypothesesoffered in Chapter 2 are discussed according to the statistical analysis of datain the two previous chapters. The first hypothesis (i.e. in spontaneous speech,the vowel space is smaller than in the reading tasks) fits more for monophthongsand for F1 dimension, but is not supported by diphthongs in the logatome readingtask (p. 253). The second hypothesis (i.e. in spontaneous speech, vowels showlarger sociogeographic variation than in the two reading tasks) is valid onlyfor /u/ among monophthongs and for /œy/ among diphthongs, whereas /ε/ and /ʏ/showed opposite behavior. A lack of interaction between speech style and regionwas also found. These results demonstrate that the second hypothesis does notclearly describe and explain sociogeographic vowel space variation. The thirdhypothesis (i.e. vowels with large sociogeographic variation in F1 or F2dimension do not reduce their space in that dimension in spontaneous speech) isviolated by /ε/, /ʏ/ and /ɑ/ (p. 285). Since the three hypothesis have beenrejected, the author proposes a new solution for the paradox. In his opinion, itis important to consider the local structure of the vowel space of the targetvowel, i.e., of the vowel which increases or shifts in variation among thedifferent speech styles. Since speakers always try to avoid phonemic confusions,a vowel could move or increase its space in two cases: (1) If the localsurrounding structure is empty; (2) If the two vowels maintain a contrastivedifference in at least one dimension other than F1 and F2 (e.g. duration, onsetposition), when there is a neighboring vowel in the vowel space. For van derHarst, this should be considered a proper solution for the Vowel Space Paradox(pp. 304-5).

Finally, in Chapter 8 the author sums up the main findings of his work,according to the seven aims proposed in the first chapter. A final sectionillustrates a wide range of proposals for further research in sociophonetics.


At a general level, the book is well structured and the arguments concerningboth research hypotheses and results are very clearly illustrated. The authorclearly made a great effort to guarantee cohesion throughout the differentchapters, which is very helpful for the reader. The review of the state of theart in Chapters 1 and 2 is very valuable because it sums up the main approachesto a key term in sociolinguistic research, i.e., style. Chapter 3 offers a goodexample of how a corpus should be structured in this kind of research. Chapter 4is a precise and rigorous analysis of three main methodological problems insociophonetic research; the observations and suggestions given by the author inthis chapter deserve to be carefully considered by scholars in both phonetic andvariationist areas of research.

The only weakness of the book is found in the analysis of spontaneous speechdata (Chapter 6). The author’s aim “to obtain a set of vowel tokens that ismaximally balanced” (p. 205) was problematized by a lack of tokens in everyphonological context. In particular, /ʏ/ does not show any occurrences in somespeakers’ speech production. In these cases, the author used the mean values of/ʏ/’s F1 and F2 detected in the word list and logatome data “in order to avoidbias of the unbalanced data set” (p. 209). However, this mere quantitativeproblem might have important consequences in the analysis, since one of the mainaims of the book is to compare different speech styles. It should be the casethat this choice did not modify the statistical results, but the author shouldhave considered and explained this problematic issue. Moreover, thesociolinguistic observations exposed at the end of the same chapter deserved awider explanation, in particular regarding the observed shift of N-S vowels toan N-R language variety.

These few problematic points aside, van der Harst’s book should be considered animportant contribution to sociophonetic research. This book could also beparticularly interesting and useful for both “classical” sociolinguists and moreskilled sociophonetic readers: the former scholars would find importantmethodological and bibliographical suggestions, while the latter will find ideason critical reflections concerning the sociophonetic study of variation. Inconclusion, this book is a clear example of “the usefulness of the integrationof sociolinguistics and phonetics” (p. 12).


Bell, Allan. 1984. Language Style as Audience Design. Language in Society 13:145-204.

Fant, Gunnar. 1960. The Acoustic Theory of Speech Production. The Hague: Mouton.

Jacobi, Irene. 2009. On Variation and Change in Diphthongs and Long Vowels ofSpoken Dutch. Doctoral disseration, University of Amsterdam.

Labov, William. 2001. Principles of Linguistic Change. Social Factors. London:Blackwell.

Meyerhoff, Miriam. 2006. Introducing Sociolinguistics. New York/London: Routledge.

Stroop, Jan. 1998. Poldernederlands. Waardoor het ABN verdwijnt. Amsterdam:Uitgeverij Bert Bakker.


Chiara Meluzzi is a PhD student at the University of Pavia and the Free University of Bozen (Italy). After a Master’s thesis on female language in Ancient Greek comedy at the University of Eastern Piedmont (Vercelli), for her PhD dissertation, she is now working on the sociolinguistics of the Italian variety spoken in Bozen (South Tyrol). Her primary research interests include sociolinguistics, pragmatics, dialectology, language contact and historical linguistics.

Page Updated: 20-Jun-2012