Polder Dutch

Vincent J. van Heuven, Loulou Edelman, Renée van Bezooijen,
The pronunciation of /Ei/ by male and female speakers of avant-garde Dutch
(referaat TIN-dag Utrecht, 26 januari 2002)

The pronunciation of /Ei/ by male and female speakers of avant-garde Dutch

1. Introduction

In de last decade a new variety of Modern Dutch has emerged, which was first noted and commented on by the dialectologist Reker in 1993 in a press release by Radio-Noord (for details see Stroop 1998: 13, 107). Taking a cue from Reker's observation, the emerging variety has been closely monitored by Stroop, who christened it Polder Dutch. The variety was elegantly described and placed in a wider sociolinguistic perspective in a popular-scientific brochure by Stroop (1998). More recently the variety got an entire website devoted to it (www.hum.uva.nl/poldernederlands), with very thorough documentation - both in Dutch and in English - on the phenomenon.

1.1. The phonetics of the sound change
The new variety differs from the Standard language only in its phonetics. Stroop presents the change as a chain shift, whereby the tense high mid vowels /e:, 2:, o:/, which have slight diphthongization in the standard language, are somewhat lowered and more noticeably diphthongized. The high mid vowels push the low mid diphthongs /Ei, 9y, Au/ to a more open position. As a result the onset of the low mid diphthongs assumes a position very close to open /a/, so that the three diphthongs are no longer clearly differentiated in their onsets. However, the end points of the diphthongs - which may possibly be lowered as well - still differentiate adequately between the front unrounded /Ei/, the front rounded /9y/, and the back rounded /Au/.
It remains unclear from the descriptions provided whether the degree of diphthongization is affected by the sound change. If it is only the onset of the diphthongs that is more open, and the end point remains stable, then the strength of diphthongization (the size of the diphthong trajectory) should have increased. However, if the onset and the endpoint have been lowered together, then the strength of diphthongization should have remained the same.
Interestingly, Stroop (1998) transcribes the new variant of the diphthong /Ei/ as aai, which trigraph is the conventional orthographic representation of the long open vowel /a:/ followed by a consonant /j/, as in haai /ha:j/ 'shark' or maait /ma:jt/ 'mows'. The rhymes of these words are appreciably (ca. 50 to 90 ms) longer than their counterparts in hei /hEi/ 'heather' and meid /mEit/ 'maid'¹. Also, the vowel in haai and maait remains stationary for a relatively long time, and then abruptly glides off toward /i/, whilst the vowel quality in hei and meid changes from the beginning onwards ('t Hart and Collier 1983; Peeters 1991; Nooteboom and Cohen 1976, Rietveld and van Heuven 2001).

Stroop claims that the lowering of /Ei/ reflects a natural tendency. In fact, low mid diphthongs are rare in the world's languages. Cognates of Dutch /Ei/ in the neighbouring languages English and German are fully open diphthongs, e.g., English wine /wain/, German wein /wain/. The Dutch mid open diphthongs, especially /Ei/ and /{y/ are notoriously difficult for foreign learners of the language. The fully open alternative would be easier to produce on the strength of the argument that the speaker just has to open his mouth as wide as he can to get the vowel right, whilst more intricate articulatory control would be required for the low-mid variety. If it is true that the more open variety of /ei/ would also have a larger change in vowel quality from onset of offset, there is the added advantage that the diphthongal nature of the vowel would stand out more clearly, reducing potential perceptual confusion with tense /e:/, which is also slightly diphthongal, (Nooteboom and Cohen 1976, Rietveld and van Heuven 2001).
In this paper we concentrate on the phonetic details of the diphthong /ei/ in Polder Dutch. This sound has been advanced as the exponent of the new variety, as is evidenced by the following quotation from Stroop (1998: 25)² :

Het opvallendste kenmerk van het Poldernederlands is [...] de uitspraak van de tweeklank [...] /ei/, die als ei of ij gespeld wordt.

'The most conspicuous feature of Polder Dutch is the pronunciation of the diphthong /ei/, which is spelled as either ei or ij.'

The first aim of this study is to clarify the phonetics of the sound change in so far as it relates to the pronunciation of the diphthong /Ei/. Moreover, rather than using the traditional impressionistic phonetic approach, i.e. listening and transcription, we wish to settle the issue using acoustic measures of greater mouth opening and/or stronger diphthongization and lengthening.

1.2. The sociolinguistics of the change
Stroop (1998) claims that the change is typical of (relatively) young, highly educated, progressive Dutch women, who wish to make a statement through speech that they are unconventional and emancipated. The variety is often found among women with high-prestige social positions such as authors, actors, film producers, artists, left-wing politicians (either local or national), high-ranking academics, and pop-singers. It is for this reason that we prefer the use of the term 'avant-garde' Dutch for the new variety, rather than Polder Dutch, which in hindsight seems a misnomer. It should be pointed out here that the avant-garde variety is found throughout the country. It is not based on any existing dialect of Dutch, and it has no documented geographic epicentre. So, avant-garde Dutch truly qualifies as a sociolect rather than a dialect or regiolect.
Although the change is led by women, Stroop predicts that men will follow suit³. It is in fact common for sound changes to be initiated by women (see Labov 2001: 366-382 for an elaborate treatment of the issue), and the avant-garde Dutch variety is no exception to the general rule. Van Bezooijen and van den Berg (2001) have shown that young women, as opposed to older women, identify with the new variety, and clearly more so than their male counterparts. The speakers of the new variety are not aware of the fact that their pronunciation of the language differs in any linguistically relevant way from the standard variety. When accused of using the avant-garde pronunciation their first reaction is denial. The sound change has all the characteristics of what Labov (2001) calls a change from below (see also van Bezooijen, Kroezen and van den Berg, 2002).
The second aim of the present study is to test the claim that the avant-garde variety of Dutch is more widespread among women than among men of similar social status and age.

2. Experimental approach

There is agreement among experimental phoneticians and sociolinguists that vowel quality, and change of vowel quality in diphthongs, can be quantified with adequate precision and validity by measuring the centre frequencies of the lower resonances in the acoustic signal. Specifically the centre frequency of the lowest resonance of the vocal tract, called first formant frequency or F1, corresponds closely to the articulatory and/or perceptual dimension of vowel height (high vs. low vowels, or close vs. open vowels). For an average male voice, the F1 values would range between 200 Hertz (Hz) for a high vowel /i/ to some 800 Hz for a low vowel /a/. The second formant frequency (or F2) reflects the place of maximal constriction during the production of the vowel, i.e., the front vs. back dimension, such that the F2 values range from roughly 2200 Hz for front /i/ down to some 800 Hz for back /u/.
The relationship between the formant frequencies and the corresponding perceived vowel quality is not linear. For instance, a change in F1 from 200 to 300 Hz brings about a much larger change in vowel quality (height) than a numerically equal change from 700 to 800 Hz. Over the past decades experimental phoneticians and psycho-physicists have developed an empirical formula that adequately maps the differences in Hertz-values onto the perceptual vowel quality (or timbre) domain, using the so-called Bark transformation (for a summary of positions, see Hayward 2000). Using this transformation (a relatively simple mathematical formula, cf. Rietveld and van Heuven 2001: 371), perceptual distance between two vowel qualities can be computed from acoustic measurements.
The acoustic characterization of diphthongs is relatively easy, once adequate measures are available to capture vowel quality in a perceptually realistic fashion. All that is needed, then, is a comparison of the formant values computed at the onset and at the offset of the diphthong. Given that the first and last portions of any vowel, monophthongs and diphthongs alike, are strongly influenced by (the articulation place of) the neighboring consonants, it is customary to sample the formant values for the starting point of the diphthong at one-quarter of the time-course of the diphthong, and to measure the formants for the endpoint of the diphthong at 75% of its duration. The degree of diphthongization is then expressed in a straightforward fashion by computing the distance between the onset (25%) and the offset (75%) vowel quality. In terms of the traditional vowel diagrams used by impressionistic linguists and phoneticians, this procedure is the equivalent of measuring the length of the arrow that represents the diphthong.
Unfortunately, formant values measured for the same vowel differ when the vowels are produced by different individuals. The larger the differences between two speakers in shape and size of the cavities in their vocal tracts, the larger the differences in formant values of perceptually identical vowel tokens. Given that the vocal tracts of women are some 15 percent smaller than those of men, comparison of formant values is especially hazardous across speakers of the opposed sex. Numerous attempts have been made, therefore, to factor out the speaker-individual component from the raw formant values such that phonetically identical vowels spoken by different individuals would come out with the same values. None of these vowel normalization procedures have proven fully satisfactory (Adank, van Heuven and van Hout 1999; Labov 2001: 157-164; Rietveld and van Heuven 2001). Broadly, two approaches to the normalization problem have been taken. The first approach, called intrinsic normalization, tries to solve the problem by considering only information that is contained in the single vowel token under consideration, typically by computing ratios between pairs of formant values such as F1/F0, F2/F1.⁴ The alternative, extrinsic normalization, looks at tokens of all the vowels in the phoneme inventory of a speaker and expresses the position of one vowel token relative to the other tokens within the individual speaker's vowel space. Obviously, extrinsic normalization procedures will be more successful but they achieve their goal by drawing on much more information than the intrinsic procedures. There is ample evidence that human listeners apply some sort of intrinsic normalization, and do not require (much) extrinsic normalization. For the purpose of the present study, however, we will adopt a hybrid normalization procedure, which combines virtues of both intrinsic and extrinsic normalization.

In the present problem, we will not need to include the full vowel system of the speakers in the analysis. Since the study is limited to the sound change in /Ei/ - a front, unrounded vowel - we only require reference vowels that allow us to determine the individual implementation of the front region of the speaker's vowel space. All that is required, therefore, is a reliable estimation of the speakers /i/ (maximally high front vowel)⁵ and /a/ (maximally open vowel front vowel). We make the explicit assumption that the point vowels /i/ and /a/ do not participate in the sound change in progress that affects the Dutch mid vowels. We are probably correct in making this assumption as Stroop's vowel diagram does not indicate any involvement of the point vowels /i, a, u/ (1998: 28).
The second basic point to consider is the sampling of the speakers. Remember that we wish to test the hypothesis that women lead the sound change. It seems imperative, therefore, that we should compare groups of male and female speakers that are equivalent in all sociolinguistically relevant aspects, such as socio-economic status and age. The speakers should not be aware of the fact that their speech production is being recorded for linguistic analysis, and their speech should be non-scripted, i.e., unpremeditated and spontaneously produced rather than rehearsed of - even worse - read out from paper. To aggravate matters, the type of speaker we were targeting is not easily accessible. These are typically well-known public figures, celebrities who will not be persuaded to participate in a scientific study. As a feasible alternative we decided to record a televised series of weekly talkshows featuring precisely the type of speakers that we were looking for. The particular talkshow, Het Blauwe Licht, is produced by the 'high-brow' VPRO television network in the Netherlands. In each show one male and one female guest together discuss recent television programs, press photos and newspaper articles with a male-female pair of co-hosts.⁶

3. Method

3.1 Materials
During the winter season of 1998/99, 16 male and 16 female guests who appeared in the television talkshow Het Blauwe Licht, were recorded on tape. The guests were native speakers of Dutch, who do not speak with a regional accent. The mean ages of the men and women were the same (45) but the women spanned a somewhat wider range than the men (ranges 28-64 and 32-52, respectively). The recordings were interviews of some 15 minutes per speaker, containing spontaneous, non-rehearsed speech (see § 2 above).
For each speaker 10 tokens of the target diphthong /Ei/ were selected from the recordings, along with 5 tokens of /i/ and 5 tokens of /a/ (see § 2). For any token to be selected into the database, it had to optimally satisfy the following list of ordered, but violable, constraints:

1. vowels must occur in content words
2. vowels must occur in different word types (multiple tokens of the same word are banned)
3. vowels must occur on syllables with main stress (as listed in the lexicon)
4. vowels must be followed by obstruents

Note that these criteria were applied to the orthographic transcripts of the recordings. No auditory (let alone acoustic) analysis was attempted before the vowel tokens were selected - on purely textual grounds.

3.2 Acoustic processing
The audio recordings were digitally sampled (16 kHz, 16 bits) and transferred to computer disk. Using the Praat speech processing software (Boersma and Weenink 1996, Boersma and van Heuven 2002) the beginnings and end points of the target vowels were located in oscillographic displays. Formant tracks for the lowest four formants (F1 through F4) were then automatically computed using the Burg LPC algorithm implemented in Praat, and visually checked by superimposing the tracks on a wideband spectrogram. Whenever a mismatch between the tracks and the formant band in the spectrogram was detected, the model order of the LPC-analysis was changed ad hoc until a proper match was obtained between track and spectrogram. Once a satisfactory match was obtained, the values for F1 and F2 were extracted at 25, 50, and 75% of the duration of the target vowel, as well as the vowel duration as such, and stored for off-line statistical processing.

4. Results

In order to give the reader a feel of the type of data that we are dealing with, we will begin by presenting a few overviews of vowel data for individual speakers. Figure 1 represents the acoustic vowel space for one male and one female speaker, whose vowel tokens are dispersed in a perfectly regular fashion.

Figure 1. Acoustic vowel diagrams plotting five tokens of /i/, five tokens of /a/ and ten tokens of diphthong /Ei/ in the Bark-transformed F1 versus F2 plane, for one male (left) and one female (right) speaker. Formants for /i/ and /a/ tokens were measured at the temporal midpoint of the vowel; formants for /Ei/ onsets were extracted at 25% of the vowel duration.

The male and the female vowel spaces have been plotted in the same coordinate system so as to facilitate cross-sex comparison. Observe that vowel height ranges between 2 and 8 Bark for the male speaker and between 3 and 9 for the female. Similarly, the front-back values range between 10 and 14 Bark for the male as opposed to 12 and 16 Bark for the female. This is a direct consequence of the cross-sex difference in the size of the oral and pharyngeal cavities. Still, by virtue of the Bark transformation, equal distances across the male and the female vowel spaces are perceptually the same.
The data typically show that the reference point vowel tokens are tightly clustered in the left-hand top corner for /i/ and the open-central area for /a/. There is more variability for the target diphthong /Ei/ for both speakers, possibly indicating within-speaker instability for this diphthong. Also, visual inspection reveals that the cloud of /Ei/ onsets for the male speaker finds itself roughly halfway between the /i/ and /a/ clusters. For the female speaker, however, the cloud of /Ei/ onsets seems to have dropped to a relatively lower position between the /i/ and /a/ reference clusters, leaving a wide gap between the /i/ and /Ei/ onset clusters.
More problematic cases arise in figure 2, where two more female speakers have been plotted as in figure 1.

Figure 2. Acoustic vowel diagrams for two more female speakers (further see figure 1).

Figure 2 illustrates two problems. Female speaker A.M.'s /i/ tokens (left-hand panel) are far from tightly clustered; in fact, one token seems completely off target, which, incidentally, is not a measurement error but due to extreme rounding and centralization of this /i/ token (which sounds more like /y/). Such occurrences of off-target realizations vowel tokens cannot be avoided, and should not be avoided, when strictly applying the textual selection criteria for inclusion of tokens in the dataset. A similar problem manifests itself in the /a/ reference tokens of female speaker A.L. (right-hand panel). Although we would expect the female /a/ tokens to have F1 values around 9 Bark - as we did indeed find for the female speakers D.L. (figure 1) and A.M. (figure 2 left) - A.L.'s /a/ tokens never extend below 8 Bark. This, again, is probably the result of rather severe centralization of the /a/ tokens, probably due to fast, colloquial speech.
In view of the susceptibility of the reference point vowels to reduction (centralization) it seems unwise to adopt the centroid (center of gravity) of the /i/ and /a/ clusters as the reference values when defining the speaker-individual vowel height dimension. Rather we decided to select the single most extreme (i.e. front-most) token within the speaker's /i/ cluster as the high-front endpoint of the dimension, and the most extreme (i.e. most open) /a/ token as the other endpoint. Consequently, the speaker's /i/ token with the highest F2 value and the /a/ token with the highest F1 value were adopted as the extremes of the speaker-individual vowel height dimension.
This procedure allows us to express vowel height speaker-individually as a relative measure. The spectral distance between the extreme /i/ token and the extreme /a/ token is set at 100%, such that /i/ has 100% vowel height and /a/ 0%. When some /Ei/ onset finds itself exactly midway between the extreme /i/ and /a/ tokens, its relative height will come out as 50%. This measure is implemented by computing (a) the euclidian distance between the reference endpoints, (b) the euclidian distance between the /Ei/ onset and the /a/ reference value, and (c) the percentage of b relative to a. The smaller the percentage c, the lower the relative starting point of the diphthong.
By the same reasoning we define a relative spectral change measure so as to express the speaker-individual degree of diphthongization for the /Ei/ tokens. First we compute the euclidian distance in the Bark-transformed F1 by F2 plane between onset (formant measurements at the 25% temporal point) and offset (measurements at the 75% point) and then take this distance as a percentage of the total distance between extreme /i/ and /a/ of the speaker. A relative glide measure of 25% would then indicate that the /Ei/ glide extends over one quarter of the entire front edge of the speaker's vowel diagram.
These speaker-normalized measures of (relative) vowel height of the /Ei/ onset and of the magnitude of the diphthongization are shown in figures 3 and 4, respectively. In these figures the values have been plotted separately for the male and female speakers, such that the speakers are ordered from left to right in ascending order of conservatism in both figures. [note: figures will be redrawn so as to conform to this description; axes will be relabeled in English]
It is obvious from figure 3 that the female speakers, on the whole, have lower /Ei/ onsets than the males. There is one man and one woman with an extremely open /Ei/ onset of 20% vowel height. It seems that the change from [Ei] to [ai] has been completed for these two speakers. At the conservative end of the scale, there is one woman with a higher (i.e. more conservative) /Ei/ onset than the most conservative of the male speakers. For the 2 × 14 speakers remaining speakers the women consistently lead in the change from [Ei] to [ai]. The effect of sex is significant by a paired t-test, t(15) = 5.46 (p < .001, one-tail).
Figure 4 reveals the same state of affairs with respect to the (normalized) magnitude of the spectral change in the diphthongs. Clearly, the women generally have a larger difference between onset and offset of the diphthongs than the men, t(15) = 2.93 (p = .005, one-tail).
Figures 3 and 4 together indicate that the phonetics of the sound change in progress are best characterized as a combined lowering and magnification of the low-mid diphthong: the onset changes from low-mid to fully low but the offset remains more or less stationary, such that a larger spectral distance has to be covered between onset to offset, which would perceptually enhance the diphthongal nature of the vowel.

Figure 3. Relative vowel height of /Ei/ onset for 16 male and 16 female speakers. Individuals are ordered from left to right in ascending order of conservatism.

Figure 4. Relative magnitude of diphthongization of /Ei/ for 16 male and 16 female speakers. Further see figure 3. [note: x-axis has descending order of conservatism; will be corrected in final version]

Figure 5 plots the relationship between onset lowering and magnitude of spectral change in the /Ei/ diphthongs of the 16 men (in gray) and 16 women (in black). The figure, and subsequent statistical analysis, reveals that there is a moderate but significant correlation between onset lowering and strength of diphthongization for the female speakers, r = .481 (p = .030, one-tail) but not for the males, r = .168 (ins.). This finding strengthens the claim that the sound change in progress is predominantly found with female speakers.
One might still argue that the acoustic differences between the male and female diphthongs /Ei/ presented so far do not reflect a difference in phonetic vowel quality but are merely due to non-uniform scaling differences between the dimensions of the vocal tracts of men versus women. In order to bear out that the acoustic measures adopted truly reflect differences in auditory vowel quality we have regressed the acoustically defined /Ei/ onset values against a perceptual measure for vowel opening that was reported earlier in Edelman (1999). Edelman reported a perceptual index (based on narrow phonetic transcriptions of diphthongs) for the strength of the Polder Dutch impression that was made by (a random selection of) 13 out of the 32 speakers in the present study. The relationship between acoustic and perceptual strength of the avant-garde quality of these 13 speakers is plotted in figure 6 above. The correlation between acoustic measure and perceptual impression is considerable (r = .742), which clearly indicates that we have not just been measuring the acoustic consequences of differences in the shapes and sizes between male and female vocal organs.

Figure 5. Relationship between normalized /Ei/ onset and magnitude of spectral change from onset to offset of /Ei/ for 16 male and 16 female speakers. Separate regression lines have been drawn for male (gray) and female (black) speakers.

Figure 6. Relationship between normalized /Ei/ onset and perceived strength of avant-garde quality of 13 speakers, broken down by sex. Regression line in black.

Let us, finally, consider the duration issue of the [Ei] to [a(:)i] change. Figure 6 presents the duration of the target /Ei/ as well as that of the reference point vowels /i/, which is a phonetically short vowel when immediately followed by an obstruent (Nooteboom 1972), and /a/, which is a long vowel.
Figure 7 shows that the /Ei/ of both male and female speakers has the same duration as the long, tense reference vowel /a/, and that both /a/ and /Ei/ are some 50 ms longer than the short reference vowel /i/. We computed a speaker-normalized duration measure for /Ei/ by dividing the duration difference between /Ei/ and /i/ into the difference between /a/ and /i/. A paired t-test on the normalized /Ei/ durations revealed no effect of sex of speaker.

Figure 7. Duration (ms) of short /i/, long /a/ and diphthong /Ei/ for 16 male and 16 female speakers.

The duration of /Ei/ is moderately - but significantly - correlated with the size of the spectral change between onset and offset of the diphthong, for both the male and the female speaker groups, r = .626 (p = .005, one-tail) and .531 (p = .017, one-tail), respectively. However, there is no correlation between the onset height of /Ei/ and its duration, for women, r = -.270 (ins.) nor for men, r = -.168 (ins.). We must therefore conclude that the correlation between duration and spectral change is not a part of the sound change in progress; it will be found in any sample of diphthongs.

5. Conclusions and discussion

The results that were obtained from the acoustic analysis of the 320 targets diphthongs (10 tokens of /Ei/ for each of 16 male and 16 female speakers) allow us to answer the phonetic issues raised in the introduction. The phonetic characterization of /Ei/ in the emerging avant-garde variety of standard Dutch is that it has a lowered onset. The offset, or end-point of the diphthong, tends to keep its original vowel height, so that the quality change between the onset and offset of the diphthong has increased accordingly. The duration of the new variant of /Ei/ has not changed; as a result the speed of the spectral change (rate of formant change as visible in a spectrogram) must have increased as well.
The analysis bears out that the onset of the new /Ei/ variety [ai] has the phonetic quality of a low front vowel, close to or even identical to the Dutch tense monophthong /a/ that was used as a reference vowel in the present study. The phonetic quality was therefore judged correctly by Stroop (1998). The lowered /ei/ should not be equated with the loan diphthong /Ai/ that occurs in some Dutch words that were borrowed from French, such as detail 'id.', email 'enamel', canaille 'riff-raff'. The onset of the imported /Ai/ would qualify as a back vowel, pronounced close to the lax Dutch vowel /A/.
The analysis also shows that the timing properties of the lowered /i/ have remained the same before and after the change. The new variant [ai] cannot be equated to the long vowel plus glide combination /a:j/ as in haai 'shark', which combination should be some 50 to 90 ms longer than /Ei/ (see introduction). Consequently, we predict that native speakers will continue to observe a phonetic contrast in pairs such as hei ~ haai [hai ~ ha:i] and mei ~ maai [mai ~ ma:i].
Sociolinguistically, the data bear out that the avant-garde variant of /Ei/ is more strongly present in the female speaker group than in the male counterparts. Although extremely progressive and conservative speakers are found among both sexes, the women lead the change quite noticeably, especially in the middle portion of the range. This conclusion supports Stroop's (1998) observation that the avant-garde variety of standard Dutch was initiated by women in precisely the socio-economic group that we targeted in this study.
Methodologically, our study has the added advantage that the sound change in progress could be studied in more detail, and in a non-impressionistic fashion, through the use of acoustic measurement procedures. Moreover, although cross-speaker and cross-sex comparison of acoustic measures of vowel quality are hazardous in principle, the procedure that we applied in our study, i.e., recording reference vowels and performing partial extrinsic speaker normalization on Bark-transformed formant measurements) affords valid comparison across vowels produced by male and female speakers.
As far as we have been able to ascertain, we are the first researchers to have adopted this specific normalization procedure, which is really a mixture of extrinsic and intrinsic normalization. It bears a resemblance to Gerstman's (1968) so-called end-point normalization, but differs from it in two details: (i) our procedure specifically looks for the front-most /i/ and the open-most /a/ in the front vowel continuum only, while the Gerstman procedure indiscriminately adopts the lowest and highest F1 and F2 values in an entire vowel set as the end-points, and (ii) our procedure is applied after Bark-transformation (which is a form of intrinsic normalization).

References

Adank, P., Heuven, V.J. van, Hout, R. van (1999) 'Speaker normalization preserving regional accent differences in vowel quality'. Proceedings of the 14th International Congress of Phonetic Sciences, San Francisco, 1593-1596.
Bezooijen, R. van and Berg, R. van den (2001) 'Who power Polder Dutch? A perceptual-sociolinguistic study of a new variety of Dutch'. In T. van der Wouden and H. Broekhuis, eds., Linguistics in the Netherlands 2001. Benjamins, Amsterdam, 1-12.
Bezooijen, R. van, Kroezen S. and Berg, R. van den (2002) Approximant r: a new and vigorous change in Dutch. This volume.
Boersma, P. and Heuven, V.J. van (2001) Speak and unSpeak with Praat. Glot International, 5, 341-347.
Boersma, P. and Weenink, D. (1996) Praat, a system for doing phonetics by computer. Report of the Institute of Phonetic Sciences Amsterdam, 132.
Booij, G.E. (1995) The phonology of Dutch. Oxford University Press, Oxford.
Collier. R. and Hart, J. 't (1983) 'The perceptual relevance of the formant trajectories in Dutch diphthongs'. In M. van den Broecke, V. van Heuven and W. Zonneveld. eds., Sound structures. Studies for Antonie Cohen, Foris, Dordrecht, 31-45.
Edelman, L. (1999). Het Poldernederlands: een vrouwentaal. Een sociolinguistisch onderzoek [Polder Dutch: a woman's language. A sociolinguistic study]. Unpublished report, Linguistics Program, Nijmegen University.
Gerstman, L.J. (1968) 'Classification of self-normalized vowels'. IEEE Transactions Audio Electro-acoustics, AU-16, 78-80.
Hayward, K. (2000) Experimental phonetics. Pearson Education, Harlow.
Labov, W. (2001) Principles of linguistic change. Social factors. Blackwell, Oxford.
Nooteboom, S.G. (1972) Production and perception of vowel duration, a study of durational properties of vowels in Dutch. Dissertation, RU Utrecht.
Peeters, W.J.M. (1991) Diphthong dynamics: a cross-linguistic perceptual analysis of temporal patterns in Dutch, English, and German. Dissertation, RU Utrecht.
Rietveld, A.C.M. and Heuven, V.J. van (2001) Algemene fonetiek [General phonetics]. Coutinho, Bussum.
Stroop, J. (1998) Poldernederlands. Waardoor het ABN verdwijnt [Polder Dutch. What makes Standard Dutch disappear]. Bert Bakker, Amsterdam.

Notes

1 No full-scale studies are available on these durations. The differences given here are visible in spectrograms provided by Cohen and Nooteboom (1976: 62) and Rietveld and van Heuven (2001: 153-154).

2 Indeed, whenever Stroop presents illustrations of the new variety on the website or during lectures, it is the diphthong /Ei/ that is used. In fact, the website even invites the public at large to submit a recording of their own production of /Ei/ - rather than some other vowel or diphthong in the language - in order to test whether the particular speaker has already fallen victim to the Polder Dutch sound change.

3 In fact, Stroop produces some examples of male pop-singers who feature the open diphthongs, specifically /Ei/, even today.

4 When formant values are rescaled to Bark, the numerical difference (F1-F2; F2-F1, etc.) is preferred over the ratio.

5 Dutch tense /a/ is quite close to cardinal vowel 4, which phonetically qualifies as a front vowel. Interestingly, tense /a/ patterns with the back vowels in the phonology of Dutch, and is normally given the feature specification [+back], cf. Booij (1995). This analysis does not invalidate the adoption of tense /a/ as the optimal reference vowel for our study.

6 Ms. Lies Kulsdon, assistant producer of Het Blauwe Licht, characterized the guests as "people with considerable cultural payload such as film producers, authors and intellectuals, who are able to uncover the deeper layers of meaning in photographic images and who are not afraid to voice their opinions".