Historical */s/ in Preconsonantal Position in the German Minority Language of Sauris/Zahre in North-Eastern Italy

In this paper aspects of the sibilant system in the variety spoken in the Germanic language island Sauris/Zahre in north-eastern Italy are presented. I investigate whether an intermediate, postalveolar “shibilant” (from Protogermanic */s/) in preconsonantal position is preserved from Middle High German as observed in other German minority languages in northern Italy. In Sauris/Zahre, in preconsonantal position a postalveolar realization is sometimes mentioned in literature, while in Modern Standard German and most related varieties in this position historical, postalveolar */s/ is retracted to palatoalveolar /ʃ/.


Introduction
In this contribution, I examine phonological change in the German minority language of Sauris/Zahre located in the Carnic Alps in north-eastern Italy using data elicited from four native speakers. I specifically investigate sibilants occurring in preconsonantal position. In this context, Proto-Germanic (PG) */s/ has undergone a sound change: the intermediate postalveolar articulation that was still present in Middle High German (MHG) has merged with palatoalveolar /ʃ/ in Modern Standard German (MSG) and most related varieties (but perhaps not in Sauris/Zahre, see the timeline in (3) in section 3).
Compare, e.g.: ( This sound change is called 's-retraction' and can be formulated as follows. Benware (1996, 265) gives the following description: (2) */s/ > /ʃ//#____C S-retraction in word-initial preconsonantal position is currently a productive phonological process in MSG (Alber et al. 2021;Benware 1996) and most related varieties (see below). The paper is organized as follows. In section 2, I present the Germanic language island Sauris/Zahre. Section 3 provides a broad overview of the evolution of the sibilant system in MSG starting from Old High German (OHG)/MHG. Section 4 presents a review of the available data from historical phonology and dialect atlases on the variety in question (Sauris/Zahre). Denison and Grassegger's (2007) dictionary and the work of the dialectologist Kranzmayer (1956) in particular are referred to and compared. In section 5, I present Carnic Friulian, the surrounding Romance variety, as this dialect is reported to also feature preconsonantal s-retraction. In section 6, I present my own data collected in Sauris/Zahre, and in section 7 the results of acoustic and auditory analyses. It turns out that in the Sauris/Zahre dialect historical */s/ in preconsonantal position is retracted to voice-This research was supported by the DAAD (German Academic Exchange Service) funding a two-month research stay at the University of Bamberg in 2019.

2
The Germanic Language Island Sauris/Zahre in the Carnic Alps Sauris/Zahre is a German language island in the north-eastern Italian region of Friuli-Venezia Giulia, which shares borders with Austria and Slovenia. Due in part to the presence of a Slovenian-speaking minority, the region is one of five in Italy with autonomous status. The language island is part of the historically and culturally defined area of Friuli, which has more than 600,000 speakers of Friulian, one of the three Western Romance languages spoken in the Alps (besides Ladin and Romansh). According to available documents, the mountain village of Sauris/ Zahre was founded in the 13th century by Southern Bavarian-speaking settlers from Carinthia and East Tyrol. For centuries, the settlers led very isolated lives, but from the 16th century onwards there was increased contact with the surrounding Romance-speaking communities. 2 The community in the language enclave has been bi-or trilingual since the 19th century: besides the local German dialect, the inhabitants of Sauris/Zahre spoke Carnic Friulian (a variety of Friulian) as a lingua franca, while standard Italian was their official language as a consequence of compulsory schooling. 3 The picture changes in the second post-war period: following Angeli (2003), parents then stopped gradually talking to their children in the Zahrar Sproch, so the number of speakers of the local dialect is declining, and among the younger generations only a few people still use it (Geyer 2018 Language islands sometimes preserve features from earlier stages of language development. In some of the Germanic language islands in northern Italy, a contrast in the place of articulation between postalveolar 'shibilants' and alveolar sibilants is still attested (see, e.g., Alber, Rabanus 2018;Alber 2014). These sibilants and the palatoalveolar [ʃ] form the 3-sibilant system based on PoA attested in MHG.
In the following, I use the symbols [ś, ź] to designate the articulatory position intermediate between alveolar and palatoalveolar without specifying further phonetic parameters, such as apical, retroflex etc. (see, e.g., Cercignani 1979;Joos 1952;Penzl 1968 for discussion), instead concentrating on the 3-way PoA distinction. The intermediate articulatory position will be labelled postalveolar, and I will also use the term 'shibilant' (blend of esh+sibilant). This terminology and the symbols have been used in recent publications (e.g., Alber et al. 2021;Alber, Rabanus 2018) and are also found in traditional dialectology.
Historical */s/ was a fortis sound: [ś] (Lessiak 1933;Cercignani 1979, 73-82); during the Althochdeutsche Frikativenschwächung (OHG-FL, around 750 AD), a lenis allophone arose ([ź]). In dialectological and historical linguistics the fortis vs. lenis distinction, referring (broadly speaking) to the realization of the fricatives with relatively greater or lesser energy, is used and applied to the two series of sibilants  Kranzmayer (1956, 89) assumes that in the Central Bavarian dialects the lenis shibilant [ź] became a voiced alveolar [z] in word-initial prevocalic position first (before 1200), while the change to a palatoalveolar articulation in preconsonantal position began later. According to Moser (1951, 222), s-retraction began in the second half of the 13th century throughout the High German region. The Southern Bavarian speakers who settled in Sauris/Zahre during the 13th century may have left the inland variety before or during this change and therefore preserved */s/ in preconsonantal position. The following timeline can be drawn: Timeline of important changes in the sibilant system before/during settlement in Sauris/Zahre: In MSG and most related varieties, the sound change /s/ > /ʃ/ before consonants is today a synchronically active phonological process as also evidenced by adaption of recent loanwords like e.g. spaghetti [ʃpaˈɡɛti]. It occurs word initially, but is also attested in word-medial or word final coda position in some dialects of German, as e.  (König 2011, 150;Stevens et al. 2015). As mentioned above, historical */s/ was a voiceless fortis sound (Lessiak 1933), while lenis [ź] entered the system during the OHG-FL. According to descriptions in historical grammars and in diachronic phonology, in MHG the shibilant [ś] realized in preconsonantal position before a voiceless consonant (/p, t/) was a fortis sound. Before the sonorants /m, n, l, w/ a lenis [ź] is sometimes assumed (Szulc 2002, 117;Paul [1881] 2007).

4
Preconsonantal Sibilants in the Variety of Sauris/Zahre -Descriptions in the Literature and in (Acoustic) Atlases The literature does not, unfortunately, contain any clear, consistent descriptions of the phonemic and phonetic properties of the sibilants in preconsonantal position in the system of the Sauris/Zahre dialect. Kranzmayer (1956, 89), who gives detailed descriptions of the In the dictionary compiled by Denison and Grassegger (2007), however, in word-initial or word-medial preconsonantal position before stops we mostly find the voiceless palatoalveolar fricative, the same sibilant that is found in MSG word initially: [ż]naidar 'tailor' (this last in Cimbrian in preconsonantal position before voiced stops and sonorants). As mentioned in section 3, a fortis fricative is assumed in context b. in MHG by, e.g., Paul [1881. 7 Kranzmayer (1956 and Denison and Grassegger (2007) describe the diphthong differently. In preconsonantal position before voiced stops and sonorants, Kranzmayer (1956) transcribes a voiced palatoalveolar [ʒ]: [ʒ]naider (MSG: [ʃ]neider 'tailor'); the same voiced sibilant is also reported in the neighboring Sappada/Plodn and Timau/Tischelbong language islands (see § 3). Denison and Grassegger's (2007) dictionary reports the voiced fricative [ʒ] in only a few items in the contemporary Sauris/Zahre system. In most contexts, the voiceless palatoalveolar sibilant is transcribed, even in word-initial position before sonorants, e.g.

Historical */s/ in Preconsonantal Position in the German Minority Language of Sauris/Zahre
[ʃ]naidar 'tailor'. The dictionary only contains six entries with initial preconsonantal [ʒ], mainly proper names borrowed from other languages and before /w/, e.g.

Hans, Johannes, Ital. Giovanni).
A further source for the spoken language of Sauris/Zahre -and other Germanic minority languages, like Cimbrian -is the VIVALDI multi-media atlas of Italian dialects and minority languages in Italy. 8 The atlas contains lists of stimuli grouped according to linguistic criteria along with IPA transcriptions and sound recordings. The focus is on dialects of Italian, which is reflected in the choice of stimuli, although individual items from the Germanic minority languages are also featured. Here, too, a voiceless (prepalatal/palatoalveolar) fricative in preconsonantal position, including before sonorants, is transcribed, e.g., To summarize, when describing the current sibilant system in the Sauris/Zahre dialect, the literature and dictionaries are not wholly in agreement with regard to the PoA of preconsonantal /s/. Before voiceless consonants, sometimes a postalveolar articulation is assumed to be preserved (e.g. [ś]tan 'to stand'), and before voiced consonants sometimes a voiced palatoalveolar sound [ʒ] is transcribed. Table 1 provides an overview of the descriptions given in linguistic atlases and dictionaries and in the literature in the contexts under investigation (D&G = Denison and Grassegger, K = Kranzmayer).  (Kranzmayer 1956, 89). In order to verify the hypothetical influence of language contact, in the following section I present a short description of the sibilant system in Carnic Friulian based on current information.
The following table presents examples from the Forni Avoltri variety for the relevant contexts taken from the VIVALDI multi-media atlas. 11 The data show that the Germanic dialect spoken in Sauris/Zahre is in contact with a variety of Friulian (Carnic Friulian) that has a 2-sibilant PoA inventory (alveolar and palatoalveolar). The fricative in preconsonantal position is the outcome of Lat. /s/, while the palatoalveolar /ʃ/ is the outcome of a previous sound change. In preconsonantal position, s-retraction applies, while in intersonorant contexts a voiced palatoalveolar [ʒ] is found. 12 In the following production study, I will address the following questions: A. Does the system in the Sauris/Zahre dialect still preserve a postalveolar articulation ([ś], [ź]) in the preconsonantal position, as hypothesized by Kranzmayer (1956) and as attested for Cimbrian and Mòcheno, for instance (see Alber et al. 2021;Alber, Rabanus 2018), thus representing an intermediate stage in the development from a 3-sibilant system in MHG to a 2-sibilant system in MSG? B. Do we find a voiced allophone of /ʃ/ in the pre-sonorant/prevoiced stop contexts, as described by Kranzmayer (1956) and as attested in other language islands in the Carnic Alps (Sappada/Plodn and Timau/Tischelbong)?
6 Production Study

Consultants and Data Collection
As mentioned in section 3, I specifically target the preconsonantal context in this study. Acoustic and auditory analyses were carried out to investigate whether the system in the Sauris/Zahre dialect still preserves a postalveolar articulation ([ś], [ź]) in preconsonantal position, and whether there is a voiced palatoalveolar articulation in the preconsonantal context. Two or three items for each context were taken from Denison and Grassegger's (2007) dictionary. Four native speakers of the Sauris/Zahre language were recorded producing 187 items. The recordings were annotated manually by the author using Praat. The Sauris/Zahre data were elicited through the following procedure: the author presented the items orally in Italian and asked the informants to translate them into their German dialect. If they were unable to recall the relevant item, it was first 12 I found no examples in the VIVALDI corpus of [ʒ] in word-initial prevocalic position in the Forni Avoltri variety, although the phoneme /ʒ/ is assumed in this position in northern Friulian (see, e.g., Carrozzo 2000, 47; cf. [ʒ]ave <'sjave> 'toad'. https:// www2.hu-berlin.de/vivaldi/index.php).

Historical */s/ in Preconsonantal Position in the German Minority Language of Sauris/Zahre
Annali di Ca' Foscari. Serie occidentale e-ISSN 2499-1562 56, 2022, 19-42 29 paraphrased in Italian, and then if necessary the German equivalent was suggested. This was (in the following analysis) the case with the items bešpl (two speakers), burštle and šlös (one speaker each). Table 3 shows the tested items. The recordings were made on an Olympus LS-14 (sampling frequency 48 kHz, 16 bit) with the informant about 25 cm from the microphone, and were conducted in February 2019, either in the homes of the participants or at the "Haus van der Zahre" ethnographic center, so the sound quality is not optimal.
The Sauris/Zahre informants were two women and two men, aged 66-77. They were native speakers of the dialect which was the only language they used to speak until they went to school at the age of six and started learning Italian. They also had active and passive knowledge of Friulian as the colloquial oral language they mainly used at work with people who did not speak the German dialect. One of the two women was a housewife and had only a passive knowledge of Friulian, since she had no need to speak it for work purposes.

Auditory Analysis
The data were analyzed auditorily and acoustically. For the auditory analysis, four linguists classified the data as voiced vs. voiceless. Evaluation sheets, one per speaker, each containing a list of 29 items were given to the linguists to rate. Only those items with sibilants in preconsonantal contexts and two items with final, post-vocalic [ʃ] are analyzed here, not the postvocalic realizations (except in word-final contexts for control) as they are not under investigation in this study. The relevant fricatives were indicated on the evaluation sheets with the neutral symbol <S> (see Appendix) in order to avoid a writing bias. Otherwise, the items were transcribed following Denison and Grassegger (2007). Along with the evaluation sheets, the linguists were given the audio recordings of the items (each pronounced twice) listed in Table 4.

Acoustic Analysis
For the acoustic analysis, the author labeled the audio files using Praat. 13 The visual representation of the sound and the audio signal were both referred to in annotating the fricatives. Visually, the aperiodicity of the waveform in the oscillogram clearly shows the beginning and end of the fricative; in the spectrogram, the beginning and end of the fricative are shown by a dark pattern signaling the frication noise. Spectral moments are used to classify the PoA of fricatives. Fricative noise moves towards lower frequencies as the place of articulation moves further back (as the length of the cavity in front of the main constriction becomes longer). Center of gravity (CoG), which will be used in the following analysis, is a measure of the average height of the frequencies in the spectrum over the frequency domain weighted by the amplitude. The CoG value will be higher for sounds with greater high frequency energy (see Kiss 2013). Current research gives a value of about 5000/6000 Hz for alveolar fricatives [s/z], and only 2500/3000 Hz for palatoalveolar fricatives [ʃ] (see Jongman 2000Jongman , 1253. Zygis (2010) investigated the 3-sibilant inventory in Polish (/s/ʃ/ɕ/) and reported the following values: alveolar /s/: 6000 Hz, alveolar palatalized /ɕ/: 4000 Hz, postalveolar /ʃ/ (transcribed as retroflex): 3000 Hz.
A word of caution is needed here. Since the recordings were not carried out in soundproof spaces, the analysis may well be affected by background noise and vowel echoes. Moreover, measurements 13 Praat: doing phonetics by computer (version 5.3.56). http://www.praat.org/.

Historical */s/ in Preconsonantal Position in the German Minority Language of Sauris/Zahre
Annali di Ca' Foscari. Serie occidentale e-ISSN 2499-1562 56, 2022, 19-42 31 can be influenced by casual acoustic variability within the frication noise. Lastly, CoG values vary depending on the speaker and the context (for example, the preceding and following vowels).

Auditory Analysis
The auditory analysis shows that the sibilant in preconsonantal position was classified as voiceless in all but one case (word-medially), and this was confirmed by the author's auditory judgement and by visual analysis (no clearly visible voice bar, no pitch curve, no visible formant structures in the spectrogram). This would not be surprising before a voiceless consonant, but is less expected before a voiced consonant. The auditory analysis therefore shows that in preconsonantal, word-initial contexts the system in the Sauris/Zahre variety has the same distribution as MSG, i.e. all sibilants are neutralized to voiceless, which is also confirmed by the transcriptions of the items in Denison and Grassegger (2007) and in the VIVALDI multi-media atlas. * One rater made no choice between voiced and voiceless in this context on two evaluation sheets. ** Three evaluation sheets did not contain the relevant items (hence only 8 evaluations); the item raStn was not rated on one evaluation sheet.
14 There were 11 evaluation sheets; where there were 3 items to rate in a given group (e.g., glos, haus, hols) a total of 33 evaluations were made, where the group contained only two items, 22 evaluations were made.

Acoustic Analysis
The acoustic analysis was carried out with a script (Elvira-García 2014) 15 that generates an FFT spectrum over the whole fricative (minus 1ms at the edges); frequencies below 1000 Hz were filtered out. 16 * In this context, one female speaker (G., aged 66) was replaced by another (R., aged 49) because the first speaker failed to produce the items requested. Table 6 shows that the sibilants in question have very similar values. Focusing specifically on CoG, there was a small difference between /s/+C VD with a value of about 2900 Hz, and word-medial and word-initial /s/+C vl with values around 2500 Hz. Preservation of an intermediate postalveolar 'shibilant' in word-initial preconsonantal position, which has been suggested in the literature, could not be confirmed. In all contexts, the CoG was within the 2500-3000 Hz range previously reported for palatoalveolar articulation (Jongman 2000(Jongman , 1253Zygis 2010).
The differences of about 400-500 Hz found between, for example, the values for [ʃ] (< OHG /sk/) in word-final position and the values for /s/+C vl in word-initial position, were too small to confirm a distinct, contrastive PoA. If these CoGs are taken as relational values, 15 http://stel.ub.edu/labfon/sites/default/files/zero-crossing-and-spectral-moments13.praat. The script calculates the following parameters: duration of the interval, zero crossing in the first 30 ms of the interval, number of crosses in the whole interval, zero crosses in the interval*10/duration of the interval, intensity, center of gravity, skewness, kurtosis, standard deviation, central moment.
16 In order to reduce the background noise in the spectrum settings the dynamic range was changed to 60 dB. This contribution presents an acoustic and auditory analysis of one aspect of the sibilant system of the dialect of the Germanic language island Sauris/Zahre. Specifically, sibilants in preconsonantal position were investigated and compared with the sibilant systems in MSG and the surrounding Romance variety, Carnic Friulian. The following questions were raised (see § 5) and are reiterated here: A. Does the system in the Sauris/Zahre dialect still preserve a postalveolar articulation in this context, which represents an intermediate stage in the development from a 3-sibilant inventory in MHG to a 2-sibilant system in MSG? B. Do we find a voiced allophone of /ʃ/ in the pre-sonorant/prevoiced stop contexts, as described e.g. by Kranzmayer (1956), Lessiak (1933;1959) and as attested in other language islands in the Carnic Alps (Sappada/Plodn and Timau/Tischelbong)?

Question A
Acoustic analysis shows that in the Sauris/Zahre variety a postalveolar articulation ([ś, ź] < PG */s/) is no longer evident in word-initial preconsonantal position. That is, the reorganization from a 3-sibilant inventory to a 2-sibilant inventory based on PoA is complete, and historical */s/, the only sibilant realized in preconsonantal position, is realized consistently as a voiceless palatoalveolar [ʃ] in word-initial position (and frequently word medially before /p/, /t/). Therefore, as far as the preconsonantal position is concerned, the Sauris/Zahre dialect follows the same development as most varieties on which MSG is based as well as MSG itself, 17 while in other Germanic varieties in northern Italy, e.g., Cimbrian spoken in Lusérn (Alber, Rabanus 2018) a postalveolar realization is still detectable in some contexts. Recent studies have drawn attention to the process of s-retraction observed in Germanic and Romance varieties in northern Italy and have raised the question if language contact could have played a role (Alber et al. 2021;Schmid 1956). The language islands in this area provide important insights into the discussion about contact related change and therefore I will briefly analyze the results of the production study in Sauris/Zahre against the background of contact linguistics.
As claimed by e.g. Schmid (1956), contact may have played a role in the implementation of s-retraction in the German based minority lan-guages in the northern Alps, i.e. also in the Sauris/Zahre variety: between the 11th and 13th centuries, s-retraction was an active process in northern Italy (see the Lombard language in the Alps: [ʃ]tela 'star', [ʃ]piga 'spike' [ʃ]kur 'dark' (Pfister 1984)), and might therefore have triggered s-retraction also in the Germanic minority languages in Italy. Schmid (1956, 50) concludes that in the areas of Romance-Germanic language contact (Tyrol, Vorarlberg, Eastern Switzerland), the Germanic dialects borrowed s-retraction, and the process then spread to the north. One of Schmid's arguments is that the two centers of s-retraction, Romansh and Ladin -in close contact, respectively, with Tyrol and Western Upper Kärnten -pattern with Swabian and Alemannic (s-retraction also word-internally). Sauris/Zahre would therefore represent the eastern edge of the area reached by this Germanic-Romance language contact phenomenon.
On the other hand, also contact independent motivations for s-retraction are discussed in literature. From a phonetic point of view, there is experimental evidence for a general lowering effect of /s/ in preconsonantal position, that is, /s/ is observed to have a lower centroid frequency and therefore to be more /ʃ/-like when followed by /p, t, k/ (Phillips 2020;Stuart-Smith et al. 2019;Stevens, Harrington 2016;Stevens et al. 2015). Besides this phonetic bias, there is evidence that a three-way articulatory contrast is difficult to maintain. Typologically, such systems are rare, reflecting the fact that the intermediate category presents articulatory and acoustic instability. Polish is one of the comparatively few languages with a 3-way articulatory contrast (dental, retroflex and alveopalatal sibilants). Regarding the intermediate sibilant, retroflex [ʂ], Bukmaier & Harrington (2016) provide evidence for the relative weakness of acoustic cues for PoA, and for low perceptual salience. They also point to data that suggest that retroflex [ʂ] is acquired relatively late by children. Reorganization of the sibilant system through removal of the intermediate category may therefore be interpreted as a kind of optimization of the sibilant system. The motivation for the 'direction' of the intermediate category towards retraction rather than towards fronting in the varieties with s-retraction could be the above-mentioned phonetic bias (/s/ more /ʃ/-like before voiceless stops). The extension to all consonants in MHG is accounted for by Benware (1996) as phonetic analogy induced by acoustic-auditory similarity: /ʃr/ triggered the following diachronic scale: [l > n > m > w]. 18 Realignment of the system after change is recognized by Alber et al. (2021) as the driving force behind preconsonantal s-retraction. The authors analyze Germanic and Romance varieties and identify the ine-ISSN 2499-1562 56, 2022, 19-42 36 troduction of a new phoneme, namely palatoalveolar [ʃ], into the sibilant system as the common denominator in all the language groups involved in the process of preconsonantal s-retraction -with subsequent reorganization from a 3-sibilant inventory (in Germanic varieties) or 1-sibilant inventory (in Romance varieties) to a 2-sibilant one. Both types of reorganization share the characteristic that only the historical sibilant is found in preconsonantal position, 19 so it could merge with the 'new' palatoalveolar sibilant /ʃ/ without a loss of distinction in this position. Both dialects discussed here, the Germanic Sauris/Zahre variety and the Romance Forni Avoltri variety, fit this pattern and can therefore count as further examples illustrating this language contact independent trajectory. To conclude, the latter arguments provides an internal motivation for the change in Sauris/Zahre. Lessiak (1933;1959) and Kranzmayer (1956, 89) claim that in the pre-sonorant/pre-voiced stop contexts, the voiced palatoalveolar sibilant [ʒ] is realized in the Sauris language. In their view, lenited [ź] evolved into voiced [ʒ] in the Sauris/Zahre dialect by adapting to the inventory of the surrounding varieties Carnic Friulian and Slovenian (Neweklowski 1990) (see § 4). Thus, the existence of the voiced palatoalveolar [ʒ] in the German minority language is ascribed to language contact, too. However, auditory analysis shows that (more than 50 years later) we do not find a voiced realization of the palatoalveolar sibilant, which is also confirmed by the transcriptions of the items in Denison and Grassegger (2007) and in the VIVALDI multi-media atlas. By consequence, in this context and regarding the feature [+/-voice] no language contact influence is observable. A voiced palatoalveolar articulation [ʒl-], [ʒn-], [ʒm-], [ʒb-] is reported in the varieties of the neighboring language islands Sappada/Plodn and Timau/Tischelbong (Geyer 2018). This might reflect the lenis character hypothesized in MHG for */s/ in this context, but in Sauris/Zahre a voiced allophone of /ʃ/ as remnant of OHG-FL is not observable in preconsonantal position.

Question B
Taking the different pieces of evidence together, we can say that preconsonantal s-retraction in the Sauris/Zahre dialect is no different from the development observed in MSG and related varieties: it is part of the reorganization of the sibilant systems (elimination of the intermediate articulatory category), is grounded in phonetics (has a perceptual and articulatory trigger). Hypothesized language contact influence in Sauris/Zahre is explainable by internal motiva-