Acoustic Properties and Regional Variation of Saterland Frisian Vowels

Acoustic properties and regional variation of Saterland Frisian vowels

Heike Schoormann Wilbert Heeringa J¨orgPeters

Institute of German Studies Carl von Ossietzky University, Oldenburg, Germany

University of Konstanz October 9, 2014

1 / 45 Introduction

Germany District of Cloppenburg

2 / 45 Introduction

• Saterland Frisian is spoken in the municipality of Saterland (Str¨ucklingen,Ramsloh and Scharrel) by 1500 to 2000 people. • Speakers are trilingual: they speak Saterland Frisian, Low German and High German. • Saterland Frisian is the last East Frisian language still spoken today. • It is one of the smallest minority languages in Europe.

3 / 45 Vowel system Monophthongs according to Fort (1980):

See also Fort (1971, 2001), Kramer (1982, 1991), Tr¨oster-Mutz (1997, 2002)

4 / 45 Vowel system

Diphthongs according to Fort (1980, 2001):

See also Kramer (1982, 1991), Bussmann (2004)

5 / 45 Research questions – acoustic properties

1a. Are all 36 Saterland Frisian vowels still distinguished? Do we ﬁnd mergers? • Closed tense vowels are especially likely to merge due to a small functional load (see Tr¨oster-Mutz2002)

6 / 45 Research questions – acoustic properties

1b. Which supplementary acoustic variables (e.g. vowel inherent spectral change (VISC), Nearey & Assmann 1986) are employed to keep the vowels distinct? • See Clements & Ridouane 2006, distinctive feature enhancement • F0-dynamics and a further centralization in F1/F2 were found to aid the distinction among closed vowels in Saterland Frisian (see Heeringa et al. 2014) • As within other large inventories duration (see Bohn 2004 on Fering) and VISC (see Fox & Jacewicz 2009, 2012 on American English and Strange & Bohn 1998 on German) may also be used to aid vowel distinction.

7 / 45 Research questions – regional variation

2a. Do Saterland Frisian vowels exhibit regional variation in spectral features? • Regional varieties may show a base-of-articulation-eﬀect (Bradlow 1995, Jacewicz, Fox & Salmons 2007) 2b. Do Saterland Frisian vowels exhibit regional variation in duration? • Saterland Frisian speakers in Scharrel are generally believed to have the highest speech rate and in Ramsloh the lowest speech rate (cf. Siebs 1893).

8 / 45 Subjects

• Male speakers between 50 and 75 years old. • All born and raised in Saterland; lived there the larger part of their life. • 13 speakers from Ramsloh, 11 speakers from Scharrel, 11 speakers from Str¨ucklingen.

9 / 45 Method

• Goal: to obtain all vowels in /hVt/ context. • /hVt/ syllables were cued by reading aloud real rhyming monosyllabic Saterland Frisian words immediately preceding the production of the /hVt/ syllable (cf. Bohn 2004). • Each such sequence was presented twice, thus two /hVt/ samples were obtained per speaker and per vowel. • Sequences were presented in random order.

10 / 45 Sweet? ‘Schweiß’

H t.

11 / 45 Sweet? ‘Schweiß’

H t.

12 / 45 Strait? ‘(er)streut’

H t.

13 / 45 Strait? ‘(er)streut’

H t.

14 / 45 Data analysis

• For each /hVt/ we measured: • Vowel duration (milliseconds) • F1 and F2 at 20%, 50% and 80% (semitones) • VISC :

q 2 2 (F 150 − F 120) + (F 250 − F 220) +

q 2 2 (F 180 − F 150) + (F 280 − F 250)

(cf. Fox & Jacewicz 2009 and Jin & Liu 2013)

15 / 45 Research question 1a

Are all 36 Saterland Frisian vowels still distinguished? Do we ﬁnd mergers? • For each Saterland Frisian location the 36 vowels were compared to each other. • 36 vowels were pronounced twice by 11-13 speakers. • We used mixed models with vowel as a ﬁxed factor and random intercepts for speaker. • The dependent variable is duration, a spectral feature or VISC.

16 / 45 Mergers

Two vowels were considered mergers, when no signiﬁcant diﬀerences were found for: • duration • F1: 20%, 50%, 80% • F2: 20%, 50%, 80% • VISC

17 / 45 Mergers of monophthongs in Scharrel

18 / 45 Mergers of monophthongs in Str¨ucklingen

19 / 45 Mergers of monophthongs in Ramsloh

20 / 45 Mergers of diphthongs in Scharrel

21 / 45 Mergers of diphthongs in Str¨ucklingen

22 / 45 Mergers of diphthongs in Ramsloh

23 / 45 Summary

vowels monoph- diph- vowels not thongs thongs still used merged merged used Fort 1980 36 Scharrel 3 2 1 30 Str¨ucklingen 2 3 2 29 Ramsloh 2 3 4 27

24 / 45 Research question 1b

Which supplementary acoustic variables (e.g. VISC) are employed to keep the vowels distinct? • Linear discriminant analysis was used to obtain the percentage of correctly predicted vowels per location

25 / 45 Monophthongs

Ramsloh Scharrel Str¨uckl. Dur + F150 + F250 64.3 68.5 68.6 Dur + F120 + F220 + F180 + F280 64.1 70.6 71.9 Dur + F150 + F250 + F120 + F220 + F180 + F280 69.2 73.6 73.6

Dur 16.4 17.0 16.0 B Dur + F150 + F250 74.1 79.5 81.1 Dur + F120 + F220 + F180 + F280 74.4 81.6 82.2 Dur + F150 + F250 + F120 + F220 + F180 + F280 78.0 82.1 82.9

26 / 45 Diphthongs

Ramsloh Scharrel Str¨uckl. Dur + F150 + F250 53.7 58.7 56.6 B Dur + F120 + F220 + F180 + F280 70.5 74.1 69.0 Dur + F150 + F250 + F120 + F220 + F180 + F280 72.0 78.0 70.9

Dur 16.1 16.8 20.6 Dur + F150 + F250 57.5 65.4 58.2 Dur + F120 + F220 + F180 + F280 73.0 73.8 70.3 Dur + F150 + F250 + F120 + F220 + F180 + F280 75.2 78.3 71.5

27 / 45 Research question 2a

Do Saterland Frisian vowels exhibit regional variation in spectral features? • Is there regional variation across the three villages in terms of location in the F1-F2-plane and the amount of VISC?

28 / 45 F1/F2 monophthongs F1 location (20%, 50%, 80%) F2 location (20%, 50%, 80%)

i i iiːii yː iiːii yː iːiiiːːi iː yyyːyː uuːu uː iːiiiːːi iː yyyːyː uuːu uː -300 iiːː i y yyyːyːːː uuuːːː -300 iiːː i y yyyːyːːː uuuːːː iːi y y yː uː uːuuuuːː iːi y y yː uː uːuuuuːː eːeː eːeː eeːːː eː øːː oː eeːːː eː øːː oː eeː øːøøːːː ː ː eeː øːøøːːː ː ː eː øøøːːː o ooooːːː eː øøøːːː o ooooːːː ooːː ooːː ɪ ʏ oː ɪ ʏ oː -400 ʏ ʊ -400 ʏ ʊ ɛː ɛːɪ œʏː ɛː ɛːɪ œʏː ɛːɛːɪ ɪ ɪɪ œœʏː ːœːː ʊ ɛːɛːɪ ɪ ɪɪ œœʏː ːœːː ʊ ɛːɪ ɪ œʏː ʏ ʊ ɔː ɛːɪ ɪ œʏː ʏ ʊ ɔː ɛː ɪ ʏʏʏ ʊ ʊ ɛː ɪ ʏʏʏ ʊ ʊ ɛː œœːœː ː ʊʊ ɛː œœːœː ː ʊʊ ɛː ɛ œ ɔː ɔːɔːɔː ɛː ɛ œ ɔː ɔːɔːɔː -500 ɔː -500 ɔː œœœ ɔː œœœ ɔː average F1 average ɛ F1 average ɛ ɛ ɛ œœ ɛ ɛ œœ ɛ œ ɛ œ ɛɛ ɔɔ ɔɔ ɛɛ ɔɔ ɔɔ ɛɛ ɔ ɔɔ ɛɛ ɔ ɔɔ -600 ɔɔ -600 ɔɔ aa aa aaa aaa a a -700 a -700 a aa aa

-2000 -1500 -1000 -2000 -1500 -1000 average F2 average F2

red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

29 / 45 F1/F2 diphthongs ending on [i] or [y] F1 location (20%, 50%, 80%) F2 location (20%, 50%, 80%)

ɛːɛːi i œoːiːi ɛːɛːi i œoːiːi oːi oːi ɔoœyːi ːi ː oːi ɔoœyːi ːi ː oːi ɛːi ɔːi o i ɛːi ɔːi o i œːi oːi œːi oːi -400 ɛːi -400 ɛːi ɔːi oːi ɔːi oːi ɛːi ɔyɔyœːœi ːi oːi ɛːi ɔyɔyœːœi ːi oːi aːi ɔːi aːi ɔːi ɛːi oːi ɛːi oːi aːi aːi œːi œːi aːi aːi œːi œːi ɛːi œːi ɔːi ɛːi œːi ɔːi ɔy ɔːɔːi i ɔɔːy i ɔy ɔːɔːi i ɔɔːy i -500 ɔːi -500 ɔːi ɔy ɔy ɔy ɔy ɛːi œːi ɔy ɔːi ɛːi œːi ɔy ɔːi ɔ ɔ average F1 average y F1 average y ɛːi ɛːi

-600 -600 aːi aːi aːi aːi aːi aːi aːi aːi aːi aːi -700 aːi -700 aːi

-2000 -1500 -1000 -2000 -1500 -1000 average F2 average F2

red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

30 / 45 F1/F2 diphthongs ending on [u] or [uw] F1 location (20%, 50%, 80%) F2 location (20%, 50%, 80%)

iːuw iːuw iiuwːɪuwiuwuw iiuwːɪuwiuwuw -300 iːuwiːiuwuw ɪuwɪuwɛuw -300 iːuwiːiuwuw ɪuwɪuwɛuw ɪuwiːuwiuw iːuwiuw eːuw ɪuwiːuwiuw iːuwiuw eːuw iːuwɪiuwuw ː iuw iːuwɪiuwuw ː iuw iuw ɪuw i uw iuw ɪuw i uw ɪuwɪuw eːuwoːu ɪuwɪuw eːuwoːu eːuw iːuw ɪuwiuw oːu eːuw iːuw ɪuwiuw oːu eːuweːuw eːuweːuw eːuw eːuw eːeuwːuw ɛuw eːeuwːuw ɛuw aoːuːu aoːuːu -400 ɛɛːuwuw aːu -400 ɛɛːuwuw aːu ɛuw eːuw ɛuw ɛːuw aːu ɛuw eːuw ɛuw ɛːuw aːu ɛuw ɛːuw oːu ɛuw ɛːuw oːu oːu oːu ɛuw ɛuw average F1 average ɛːuw ɛːuw oːu F1 average ɛːuw ɛːuw oːu ɛːuw ɛːuw oːu ɛːuw ɛːuw oːu -500 ɛːuw ɛuw -500 ɛːuw ɛuw ɛːuw aːu ɛːuw aːu oːu oːu ɛuw aaːːuu ɛuw aaːːuu oaːːuu oaːːuu -600 aːu -600 aːu aːu aːu

-2000 -1600 -1200 -800 -2000 -1600 -1200 -800 average F2 average F2

red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

31 / 45 VISC monophthongs

ːi iiiː y -300 iː i yyyːːyː uuːː y uːuu eː eeːː øøːː ooːː øː oː -400 ʏ ɛː ɪ œː ɪ œː ʊ ɪ ʏʏ ɛː œː ʊ ʊ ɔː -500 ɔː

average F1 average œ ɛ œ ɛ œ ɛ ɔɔ -600 ɔ

a a -700 a

-2000 -1500 -1000 average F2 red: Ramsloh, green: Scharrel, blue: Str¨ucklingenA → B: A larger than B

32 / 45 VISC diphthongs

-300 ː iːuwi ɪuwuwiuwiuw ɪuw iːuw ɪuwiuw oːi eːeuwːuw -400 ɛːi eːuw ɛuw oːi ɛː œːœi ːi i ɛuw oːu oːu ɛːi oːi œːi ɛːɔyuwɔːɔːi i ɛːuw oːu -500 ɛuw ɔy average F1 average ɛː uwɔy ɔːi aːu aaːːuu

-600 aːi aːi aːi

-1750 -1500 -1250 -1000 -750 average F2 red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

33 / 45 Analysis per category

Notation of monophthongs according to Fort (1980):

34 / 45 Analysis per category

• We distinguish the following categories: • monophthongs: • front/back • closed/near-close/close-mid/open-mid/open • tense/lax • diphthongs • We used a linear mixed model for each acoustic variable and per category. Location is a ﬁxed factor, random intercepts are included for speaker and vowel. Location is a random slope of vowel only when this improves the model (i.e. decreases the Akaike information criterion).

35 / 45 Analysis per category

• Scharrel compared to Ramsloh: • front monophthongs have higher F1 (20%) • close-mid monophthongs have higher F1 (20%, 50%) • diphthongs have higher F1 (80%) and less VISC • Scharrel compared to Strücklingen: • open monophthong [a] has lower F1 (20%, 50%) • Ramsloh compared versus Strücklingen: • no systematic (categorical) differences found

36 / 45 Research question 2b

Do Saterland Frisian vowels exhibit regional variation in duration? • Is there regional variation across the three villages in terms of duration?

37 / 45 Duration monophthongs

ːi iiiː y -300 iː i yyyːːyː uuːː y uːuu eː eeːː øøːː ooːː øː oː -400 ʏ ɛː ɪ œː ɪ œː ʊ ɪ ʏʏ ɛː œː ʊ ʊ ɔː -500 ɔː

average F1 average œ ɛ œ ɛ œ ɛ ɔɔ -600 ɔ

a a -700 a

-2000 -1500 -1000 average F2 red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

38 / 45 Duration diphthongs

-600 aːi aːi aːi

-1750 -1500 -1250 -1000 -750 average F2 red: Ramsloh, green: Scharrel, blue: Str¨ucklingen,A → B: A larger than B

39 / 45 Conclusions – acoustic properties

1a. Are all 36 Saterland Frisian vowels still distinguished? Do we ﬁnd mergers?

• Some qualities have merged, some have been dropped. 27 to 30 vowels can still be detected. • High short tense vowels have merged with high long tense vowels; only in Scharrel short tense [i] is still shorter than long tense [i:].

40 / 45 Conclusions – acoustic properties

1b. Which supplementory acoustic variables (like VISC) may be employed to keep the vowels distinct?

• Monophthongs are truly monophthongal, they are well discriminated (74-82%) by means of duration and the vowel target (50%-Point). Adding further spectral information (VISC) increases the discrimination score by 1-4% only. • Diphthongs are truly diphthongal, they are well discriminated (69-74%) by spectral information from the onset and the oﬀset. Adding duration increases the discrimination score by 0.3-2.5% only.

41 / 45 Conclusions – regional variation

2a. Do Saterland Frisian vowels exhibit regional variation in spectral features?

• Especially Scharrel has systematic differences compared to the other two locations. • Compared to Ramsloh F1 is higher in several vowel categories and dipthongs have less VISC. Compared to Strücklingenopen vowel [a] has a lower F1. • Ramsloh versus Strücklingen:no systematic differences were found.

42 / 45 Conclusions – regional variation

2b. Do Saterland Frisian vowels exhibit regional variation in duration?

• No signiﬁcant regional diﬀerences were found.

43 / 45 Thanks!

44 / 45 front back close i: y: u: iy u near-close e: ø: o: close-mid I Y U E: œ: O: open-mid E œ O open a a:

45 / 45