CROSS-LINGUISTIC VOWEL VARIATION IN SATERLAND: SATERLAND FRISIAN, LOW GERMAN, AND HIGH GERMAN Wilbert Heeringa, Heike Schoormann, Jörg Peters Institute of German Studies, University of Oldenburg, Germany [email protected], [email protected], [email protected] ABSTRACT Figure 1: Location of Saterland in the northwest of Germany (marked in gray in the map on the This study investigates the vowel space of trilin- left) and within the district of Cloppenburg (white gual speakers of Saterland Frisian, Low German, area in the map on the right). and High German. The three vowel systems show differences in the number of distinct categories but share the majority of vowel qualities. The speakers were instructed to read vowels of all three languages in a /hVt/ frame. We examine whether the disper- sion and size of the vowel space as well as inter- language variability of individual vowels correlate with the number of vowel categories. Additionally, systematic cross-linguistic differences were mea- sured regarding duration and mid-vowel F1 and F2. High German monophthongs were found to be produced with longer and more variable duration. Moreover, High German monophthongs were pro- According to Sjölin [13], Fort [5] and Kramer [8] duced with smaller F1 and larger F2 values than Saterland Frisian has a complete set of closed short the respective Saterland Frisian and Low German tense vowels: /i y u/. In our data set these vowels categories. These results suggest that the subjects were found to be merged with the closed long tense may use the same base-of-articulation for Saterland vowels: /i: y: u:/. Frisian and Low German but not for High German. Table 1: Monophthongs used in closed sylla- Keywords: Saterland Frisian, trilingualism, vowel bles in Saterland Frisian, Low German and High German spoken in Scharrel. Notation according to space, adaptive dispersion, base-of-articulation Fort [5]. 1. INTRODUCTION Saterland Frisian Low German High German i: y: u: i: y: u: i: y: u: Saterland Frisian is spoken in three small villages – e: ø: o: e: ø: o: e: ø: o: Strücklingen, Ramsloh and Scharrel – in the north- IYU IYU IYU western corner of the district of Cloppenburg in E: œ:O: E: œ:O: E: Lower Saxony (see Figure 1). It is the only remain- E œ O E œ O E œ O ing living variety of East Frisian. Saterland is be- a: a: lieved to have been colonized by Frisians from the a a a coastal areas in the eleventh century. According to the most recent count, Saterland Frisian is spo- 13 diphthongs were attested for Saterland Frisian, ken by 2250 speakers [14]. Many of these speakers seven for Low German and three for High German in are trilingual. In addition to Saterland Frisian, they closed syllables. The High German diphthongs /ai/, speak Low German and High German. /au/ and /Oy/ are shared by all three languages. ˆ ˆ We compared the vowel systems of the three lan- Adaptiveˆ dispersion theory [9, 11, 12] states that guages as spoken in Scharrel. Table 1 shows for each the distinctive sounds of a language tend to be posi- language those monophthongs which were attested tioned in phonetic space so as to provide sufficient in closed syllables in the data we collected (see Sec- perceptual contrast, and therefore to minimize the tion 2.2 below). 16, 17 and 15 monophthongs were potential for perceptual confusion between the dis- distinguished in the respective languages. tinct categories. The contrast between vowels can be maximized by increasing the vowel space and by relate with the number of vowels in the vowel sys- spreading the vowels over the whole area, i.e. by tems of the three languages? (3) Are there system- moving vowels away from the center. atic differences between the three languages in dura- We are especially interested in the effect of the tion and in mid-vowel F1 and F2, which are relevant inventory size on the acoustic vowel spaces of for a language-specific base-of-articulation effect? the three different languages. Jongman et al. [7] compared the German vowel space (14 monoph- 2. METHOD thongs) with the Greek vowel space (five monoph- thongs), and found that the German vowel space 2.1. Speakers was expanded compared to the Greek vowel space. Bradlow [3] compared English (11 monophthongs) The experiments were carried out with eleven trilin- with Spanish and Greek (both languages have five gual male speakers, aged between 51 and 75 years. monophthongs). For closed-syllable words she All speakers were born and raised in Scharrel and found that the English vowel space was expanded have lived in this village for all or the majority of compared to the other two languages. their lives. Adaptive dispersion theory also states that vari- 2.2. Procedure ability of individual vowels is inversely related to the number of phonemes in the vowel inventory, i.e. We elicited all shared vowels in a /hVt/ context for vowel formant values should vary to a larger extent each of the three languages. The vowels were ob- in smaller than in larger systems [10]. This predic- tained in three separate sessions, with interim peri- tion is based on the idea that since the potential for ods of two to three months. Each session was in- perceptual confusion is lower in less crowded vowel structed by a native speaker of the respective lan- spaces, speakers have the freedom to vary more. guage to ensure the intended language mode. /hVt/ However, Flege [4] was not able to confirm this pre- words were cued by reading aloud real rhyming diction for English and Spanish. monosyllabic words in the respective language im- Furthermore, when comparing the locations of mediately preceding the production of the /hVt/ tar- the common vowels of English and Spanish, Brad- get word [2]. For example, in order to obtain [e:] low [3] found that the English vowels are all signif- in Saterland Frisian, first the Saterland Frisian word icantly higher in the F2 dimension than their Span- leet ‘late’ was shown, together with its High Ger- ish counterparts. According to Bradlow [3] this sug- man translation. The subject read [le:t]. Subse- gests that the English vowels are all articulated with quently, the frame H_t was added below leet and the a fronted tongue position compared to Spanish vow- translation disappeared. Since leet is pronounced as els. Similar results were found for bilingual speakers [le:t], the subject built the rhyming target word and of Spanish and Quichua by Guion [6]. Because of pronounced H_t as [he:t]. Each such sequence was the systematic difference for all of the shared vowels presented twice, thus two /hVt/ samples were ob- of the two languages, Bradlow refers to the notion tained per speaker and per vowel. Sequences were of a language-specific base-of-articulation, which presented in controlled randomized order. The elici- means that vowels belonging to the same phono- tation via rhymes and thus self-built instead of read logical category may have different phonetic real- target words was preferred over a reading task be- izations across different languages. Since different cause both Saterland Frisian and Low German or- shifts were found for the formants, the observed F2 thography were unknown to our speakers and the difference could not be explained by differences in written form may have had a direct influence on the the vocal tract length between the two groups of production data. speakers. We used monosyllabic trigger words ending with We investigated whether the number of monoph- [t] in the three languages. If this was not possible, thongs in Saterland Frisian, Low German, and High an intermediate form was shown between the trigger German correlates with vowel space, vowel disper- and the target word. For example in order to obtain sion, and inter-language variability of vowels. In ad- the Saterland Frisian [œ] from the trigger löskje ‘ex- dition, we looked for systematic differences in dura- tinguish’, the intermediate form lött was added in a tion and spectral properties between the three lan- second step. The intermediate form then led to the guages. The following research questions are ad- production of the rhyming target word [hœt]. dressed: (1) Do the three languages’ vowel systems Both series of words obtained in two steps and differ in vowel space and dispersion? (2) Does the three steps were preceded by practice words, so that inter-language variability of individual vowels cor- informants became familiarized with the test. 2.3. Acoustic variables 3.2. Inter-language variability of vowels For each variable – duration, F1 and F2 – we mea- Acoustic variables were measured with PRAAT [1]. sured the standard deviation of the 11 speakers per For each vowel we measured the vowel duration and vowel and per language. For any pair of languages mid-vowel F1 and F2 (in Hertz).1 the standard deviations of corresponding vowels were compared.5 3. RESULTS Table 2 shows that in several cases High German vowels have a higher standard deviation than Sater- All analyses were based on the 14 shared vowel cat- land Frisian and/or Low German vowels. This con- egories of the three languages. Furthermore, we dis- cerns inter-language variability of vowel duration tinguished between monophthongs and diphthongs. for monophthongs as well as diphthongs. High Ger- Within monophthongs we further distinguished sub- man also shows more F1 variability than Saterland groups of front and back vowels as well as sub- Frisian among close vowels, but less variability than groups of closed and open vowels. The subgroups Low German regarding the diphthongs. In the F2 were determined relative to the gravity center of the dimension High German front vowels vary stronger vowels, which has average F1 and F2 as coordinates.