LINGUIST List 30.1465
Tue Apr 02 2019
Review: Dutch; Spanish; Language Acquisition; Phonetics: Burgos (2018)
Editor for this issue: Jeremy Coburn <jecoburnlinguistlist.org>
***************** LINGUIST List Support *****************
Fund Drive 2019
29 years of LINGUIST List!
Please support the LL editors and operation with a donation at:
Robert Squizzero <rsquizz
Non-native pronunciation: Patterns of learner variation in Spanish-accented Dutch E-mail this message to a friend Discuss this message
Book announced at https://linguistlist.org/issues/29/29-2715.html
AUTHOR: Pepi Burgos
TITLE: Non-native pronunciation: Patterns of learner variation in Spanish-accented Dutch
SERIES TITLE: LOT Dissertation Series
PUBLISHER: Netherlands Graduate School of Linguistics / Landelijke (LOT)
REVIEWER: Robert Squizzero, University of Washington
“Non-native pronunciation: Patterns of learner variation in Spanish-accented Dutch” is a doctoral dissertation that examines the production and perception of second language (L2) Dutch used in the Netherlands by adult first language (L1) Spanish speakers at different levels of Dutch proficiency. Chapter 1 provides background information for the dissertation, particularly on the phonological systems of Spanish and Dutch, theory surrounding perception and production of L2-accented speech in general, and general background regarding Dutch spoken by L1 Spanish speakers in the Netherlands. Chapter 2 presents an auditory analysis that identifies rates of common segmental errors in Dutch produced by 23 adult native Spanish speakers of four different proficiency levels, then provides hypotheses for L1 interference as a cause of these errors. In Chapter 3, a comparative acoustic analysis of Dutch vowels produced by L1 Dutch and L1 Spanish speakers reveals similarities and differences in spectral quality and duration of the vowels in the speakers’ systems. Chapter 4 presents a perceptual study analyzing the accuracy of crowdsourced native Dutch listeners in orthographically transcribing Spanish-accented Dutch words. Chapter 5 is a follow-up perceptual study in which listeners were carefully selected to expand on Chapter 4’s crowdsourced study and provide a baseline against which the effectiveness of the crowdsourcing method could be evaluated. Chapter 6 takes the classifications of Spanish-accented Dutch vowels by human listeners obtained in Chapter 5’s study and compares them with machine-based classifications using multinomial logistic regression. Chapter 7 provides a summary and discussion of the dissertation’s results, limitations, and potential for societal impact.
The book begins with the statement of a problem regarding Spanish-accented Dutch: many Spanish learners of Dutch find the pronunciation difficult and often believe that Dutch listeners do not try enough to understand their speech, while many Dutch listeners, conversely, believe that Spanish learners of Dutch do not put enough effort into their pronunciation. Chapter 1 shows that this is a problem with a past, briefly summarizing the history of two major waves of migration from Spain into the Netherlands. Following that is a general comparison of the Dutch and Spanish languages, including morphosyntax in addition to phonology. The focus of this book is segmental phonology, with a detailed overview of differences in vowels and consonants between the two languages and a brief discussion of existing studies of Spanish learners of other languages. Next is an examination of speech perception models used to understand how L1 Spanish speakers acquire Dutch as a second language. Flege’s (1995) Speech Learning Model is chosen as the primary model employed by Burgos in understanding this acquisition process, with the occasional use of the Second Language Linguistic Perception Model (Escudero 2005). The last points in the introduction detail Burgos’s perspective on perceptual adaptation of native Dutch listeners to Spanish-accented Speakers.
Chapter 2 is a study of the pronunciation of the Dutch of 23 L1 Spanish speakers, divided into four proficiency levels from A1 (lowest) to B2 (high intermediate) on the Common European Framework Reference (CEFR) scale. The author compiled a corpus of running speech from the oral sections of standardized Dutch proficiency exams for these 23 speakers. The corpus was then transcribed and its segmental errors notated. The analysis revealed that the primary difference in pronunciation errors is between A1 and A2 levels, as A1 speakers had a significantly higher level of pronunciation errors than A2, B1, and B2 speakers, which had comparable error rates between each other. The study also identified that the most frequent and most persistent segmental errors were of vowels, specifically vowel height, length, and rounding (of front vowels). The identification of these errors proves to be valuable in laying the groundwork for experiments described in later chapters. Burgos also points out that the errors are consistent with the acquisition models described in Chapter 1 in that the confusions seem to stem from multiple L2 vowel phonemes occurring within the acoustic space of a single L1 phoneme, for example, /i/ and /ɪ/, which contrast in Dutch but not Spanish.
Chapter 3 acoustically examines the production of the fifteen vowels of Dutch by 28 native Spanish speakers as compared to 20 native speakers of standard Dutch. Dutch’s fifteen vowels are classified into subsets: three diphthongs /ɛi œy ɔu/, three long mid monophthongs /eː øː oː/, and nine (other) monophthongs /i y u ɪ ʏ ɔ ɛ ɑ aː/. This contrasts with Spanish, which has only five monophthongs /i u o e a/ and no phonemic diphthongs. Each participant read 29 monosyllabic Dutch words ending in /t/ or /s/, and their recordings were acoustically analyzed. A particularly clear and interesting result of the comparison can be seen in the normalized vowel plots (in first formant (F1) x second formant (F2) space) of the nine Dutch monophthongs presented on p. 94, which contrast the Dutch monophthong spaces of native Dutch and native Spanish speakers. The plots show that the tense-lax vowel pairs /aː/-/ɑ/, /i/-/ɪ/ and /y/-/ʏ/, entirely separate in F1 x F2 for L1 Dutch speakers, have very high rates of spectral overlap for the L1 Spanish L2 Dutch speakers, and /y/-/ʏ/ are realized as central rather than front vowels. Pillai’s trace measures confirm the low rates of separation for these vowel pairs produced by the L2 Dutch speakers. This study also shows that the L2 Dutch speakers realize the differences between these vowel pairs by producing each tense vowel with a longer duration than its lax counterpart. Like the L2 speakers, L1 Dutch speakers do pronounce the tense vowels /i/ and /a:/ with a longer duration than /ɑ/ and /ɪ/, but the discrepancy is not as great as the L2 speakers, and the L1 Dutch speakers actually produce /y/ with a shorter duration than /ʏ/.
Chapter 4 is a perceptual study investigating whether or not Spanish-accented L2 Dutch words differ enough from the expectations of a non-expert native listener to impact intelligibility. Nearly 200 native Dutch-speaking participants completed a web survey, which had a game component and could be shared on Facebook. The listeners heard the 29 monosyllables from the 28 speakers recorded for the experiment in Chapter 3 and transcribed them using Dutch orthography. Participants could choose how many words to transcribe, i.e. when to quit the game, with a minimum of 10 and a maximum of 833 words. The study’s results indicate that the Dutch monophthongs closest acoustically to the Spanish monophthongs, /i u ɔ ɛ aː/, were recognized correctly at the highest rates. The predicted confusions (based on the acoustic study in Chapter 3) largely occurred as anticipated, with /ɪ/ frequently confused for /i/, and /ɑ/ frequently confused as /aː/. The central quality of /y/ and /ʏ/ also appears to have interfered with listeners, as they were both frequently misheard as /u/.
A similar perceptual study was conducted and reported on in Chapter 5. This experiment used the same set of stimuli as the experiment in Chapter 4, but with the addition of two sets of the 29 monosyllables recorded by two native Dutch speakers as “anchor points.” The other major difference from the previous experiment was that participants were recruited via snowball sampling instead of crowdsourcing. Differences in rates of vowel identification between the studies were minimal, which implies that crowdsourcing may very well be a reliable method for subject recruitment in speech perception studies. A point of interest as pertains to the results of this study is that the tokens produced by L1 Dutch speakers were not correctly identified any more often than those produced by L2 Dutch speakers. Burgos attributes this result to the idea that the raters shifted their perceptual boundaries towards the L2 productions, which make up 93% of all stimuli.
Chapter 6 is a comparison of auditory vowel classification by native Dutch listeners with machine-based statistical classification based on the vowels’ acoustic properties using a multinomial logistic regression model. The same recordings of Dutch monosyllables analyzed in Chapter 3 are under investigation in Chapter 6; these were recorded by 20 native Dutch speakers and 28 Spanish L1 Dutch L2 speakers, ranging from CEFR A1 to B2 in proficiency. The machine classification was done multiple times using different sets of data: all 48 speakers’ data at once (‘Total’), native and non-native data separately (‘Group’), and each L2 Dutch speaker’s data alone (‘Individual’). The ‘Group’ classification was conducted in order to further explore Chapter 5’s idea that perceptual boundaries of native vowel categories can shift when non-native data is included, and the ‘Individual’ classification was done so that proficiency level and other individual learner variation could be captured. All of these conditions were categorized twice each, once based on the vowels’ F1 and F2 values, and once based on F1, F2, and duration. The results confirm that vowel height, length, rounding, and diphthongization impacted machine classification, and that confusions of L2 Dutch tended to occur for vowel contrasts not present in Spanish. Human and machine classification rates were comparable for the five Dutch vowels /i u ɔ ɛ aː/ that most closely resemble the five Spanish monophthongs /i u o e a/ but differed for the remaining ten Dutch vowels. Burgos’s explanation for this is that human listeners are likely attending to elements in the acoustic signal other than F1, F2, and duration, and also that human listeners have a chance to shift their perceptual boundaries as they listen to more stimuli, whereas machine classification takes in all of the data at once. The latter point is supported by the ‘Individual’ condition, in which classification rates more closely matched the intended production; the variation within a speaker’s vowel system can be attended to one speaker at a time, and classification boundaries can shift accordingly. Lastly, the analysis of individual speakers in this study shows that acquisition of more nativelike phonology does not necessarily proceed at the same rate of an increase in higher language proficiency, since the more proficient speakers in this study did not necessarily have higher rates of intelligibility than less proficient speakers.
The final chapter summarizes the work and discusses its findings. Burgos outlines future research that should be undertaken based on this book’s findings, particularly on evaluative judgments of L2 Dutch speakers and their corresponding social consequences. Burgos also concludes that the information learned from these studies can be used to inform L2 Dutch pronunciation pedagogy, help in developing Computer-Assisted Pronunciation Training, and raise awareness among Spanish learners of Dutch.
On the whole, Burgos’s work is a well-designed examination of the topic under investigation, and each chapter builds on the previous in a logical and consistent way. Chapters 2 through 6 would also be perfectly comprehensible as independent works. This book investigates internet crowdsourcing of subjects and machine classification of vowels, modern methods that are likely to be of interest to speech perception researchers. Burgos’s contrastive analysis is overall sound with regard to both the production and perception studies.
This work is somewhat challenging to understand for readers unfamiliar with Dutch. A more consistent use of the IPA (International Phonetic Alphabet) instead of Dutch orthography in referencing particular vowels, both in prose and in charts, would have made this work more accessible. While it makes some sense to use Dutch orthography when referring to transcriptions actually typed by listeners, those of us less familiar with Dutch must flip back and forth to interpret the confusions made by listeners in the perceptual studies. Also unclear is the possibility of frequency effects – the reasoning based on acoustics is sound, but could the listeners have confused certain vowels for each other simply because they are more frequent in Dutch? This certainly seems to be the case based on the frequency tables in Chapter 2 but is not accounted for in the perceptual studies.
All in all, this work adds value to the existing body of literature on L2 speech production and perception, and its relevance is clear to applied and theoretical linguists working on second language studies of any language. The potential benefits brought by this research, mentioned in the conclusion, are sure to be relevant to teachers of Dutch as a second language. Burgos also appropriately identifies the need for future work on Spanish-accented Dutch that takes into account social biases of L1 Dutch listeners. This call is particularly welcome in the context of some of the statements made throughout the work, such as one advanced (CEFR B2) learner’s comment that “she was fired because customers could not understand her Dutch” (p. 187).
Escudero, P. (2005). Linguistic perception and second language acquisition: Explaining the attainment of optimal phonological categorization. Doctoral dissertation. Utrecht, the Netherlands: Utrecht University.
Flege, J. E. (1995). Second-language speech learning: Theory, findings, and problems. In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-language Research (pp. 233-277). Timonium, MD: York Press.
ABOUT THE REVIEWER
Rob Squizzero is a doctoral student in linguistics at the University of Washington. His main research interests include articulatory and acoustic phonetics, sociolinguistics, and second language speakers. He also holds an M.A. in TESOL from the University of Washington and spent several years teaching English language to adults at the university level.
Page Updated: 02-Apr-2019