Prepositional phrases in German in Austria – identifying patterns of variation


 German in Austria is claimed to be shaped by wide-spread dialect use and historical language contact. In this context, variation in prepositional phrases (PPs) is frequently cited, but still underresearched. Three linguistic variables are particularly interesting: (1) preposition choice (p-choice), (2) case marking in PPs and (3) preposition–determiner contractions. The present study aims at identifying linguistic and sociolinguistic – including regional – patterns of variation in the realization of PPs with two-way prepositions in German in Austria on the basis of natural production data including formal and informal registers from urban and rural adults of different age groups and different socioeconomic backgrounds living in Bavarian regions of Austria. The data were compared against the German standard variety from Germany to identify all constructions (possibly) specific for German in Austria. Results indicate that p-choice (particularly in directed motion constructions) shows mostly regional effects: South Bavarian varieties are characterized by preposition drop, whereas in the other regions, the preposition auf ‘on(to)’ seems overrepresented. However, case marking is more dependent on sociodemographic variables and accusative–dative syncretism appears more frequently in plural than in singular contexts. Finally, specific preposition–determiner contractions are widespread across all regions and groups investigated, even in (close-to-)standard registers.


Introduction
This study aims at identifying linguistic and sociolinguistic patterns of variation in the realization of prepositional phrases (PPs) with two-way prepositions in spoken German in Austria. It describes linguistic constructions that are typical for German in Austria in preposition choice (p-choice), aspects of case marking in PPs and/or preposition-determiner contractions and thus accounts for phenomena on several levels of linguistic description. By hypothesizing about their distribution in socially or regionally defined (Bavarian) varieties of German in Austria, this study lays the foundations for future research on the phenomena in focus.
The remaining first section discusses terminological aspects of and sociolinguistic perspectives on German in Austria and identifies varieties of special interest in getting a first impression of the variation of not yet sufficiently investigated phenomena. It then recapitulates how variation in PPsincluding phonological, morphological as well as phenomena on the syntax-lexicon interfacehas so far been described for German in Austria before it systematically revisits the grammatical conditions of variation As can be seen, this list includes phenomena on various linguistic levels from the interface of phonology and morphology (preposition-determiner contraction, [a]) to the interface of syntax and lexicon (case government, [b]; p-choice in prepositional or directive arguments [c and d], or phrases with character of phraseolexemes, [e and f]).
Most of the mentioned examples, however, include the preposition auf 'on' governed by either a mental verb or a verb of motion or position. Interestingly, publications on (historical) language contact of German in Austria with Slavic languages (e.g. Newerkla 2007, Schleicher 1851, Schuchardt 1884) particularly highlight p-choice and also list the above-given constructions except for SAG in die Schule gehen 'to go to school' as results of this contact situation. Schuchardt (1884, 115) even argues that prepositions in general are especially prone to language contact-induced change and identifies auf 'on' as the "favourite" preposition of speakers of German in the Habsburg monarchy with Slavic L1 (see also Kim 2020, 115-26).
Another phenomenon pertaining to p-choice, namely preposition drop (p-drop), is prominently discussed in the context of present-day language contact and ethnolects of young urban migrants in Germany (Auer 2003, Siegel 2014, Wiese 2012. However, in directed motion constructions in the form intransitive motion verb + ARG DIR , it is also typical for Southern Bavarian dialects in Austria (Lenzhofer 2017, Pohl 1989, Kerschhofer-Puhalo 2019. In addition, Kerschhofer-Puhalo (2019) describes several examples of directed motion as well as locative and temporal constructions in informal speech of autochthonous (non-migrant) people of various ages living in Vienna. Kerschhofer-Puhalo (2019, 132) highlights that these "compact constructions" follow several structural and lexical as well as pragmatic restrictions: Locative constructions only appear with the copula sein 'be', whereas a very small number of motion verbs (particularly gehen 'go', fahren 'drive', rarely kommen 'come') are involved in the directed motion constructions, whereas other more specific motion verbs (e.g. laufen 'run', wandern 'hike') are not found in these constructions that involve p-drop (Kerschhofer-Puhalo 2019, 132). Following Bühler's (1934) two field theory, which distinguishes between the deictic and the symbolic field, Kerschhofer-Puhalo (2019, 132) classifies these compact constructions as deictic, as they are part of the "here and now" and unambiguous in the context of the communication.
Some Upper German varieties also use constructions including two-way prepositions that are fundamentally different from those in standard varieties of German like so-called prepositional dative marking described by Seiler (2002). In these constructions, PPs with either in 'in' or an 'at' replace the dative (e.g. er hats an/in der Mutter gesagt 'the told the mother', see Seiler 2002, 243). Generally, however, data-based research on the use of prepositions or PPs in spoken non-standard varieties of German in Austria is scarce, except for a detailed analysis of preposition frequencies in spoken Alemannic (including Vorarlberg in Austria) and Franconian varieties (Mikosch 1987). In contrast to Mikosch (1987), the present study does not focus prepositions as such, but PPs as a whole. As already indicated above and elaborated in the next section, p-choice is not the only potential position of variation in these phrases.

Potential positions of variation in PPs
Prepositions are functors (or: operative terms) that form PPs by expanding nominal (and other) phrases (NPs). In doing so, they govern the NPs' case. Semantically, they relate an object, property or fact to another object or fact in a specified way (e.g. local, directional, causal, temporal, etc.) (Zifonun et al. 1997, 44-5). Their semantic meaning thus comprises two aspects: First, prepositions constitute and denote a certain region with regard to their "inner" argument (e.g. the noun in the governed NP, the shelf in [1]; for a closer semantic description of some German prepositions cf. Eroms 1981, Wunderlich and Herweg 1991, Zifonun et al. 1997, 2116. Second, they express a relation of its "outer" argument (e.g. the subject of the sentence, the cat in [1] or the head of the NP in [2]) to the "inner" argument (Zifonun et al. 1997(Zifonun et al. , 2099.
(1) 'The cat on the shelf preens itself.' Due to their ability to denote relations between entities, prepositions are not mere function words without or with only marginal lexical meaning. They rather resemble other content word classes in bearing meaning as well as in forming an open word class, i.e. open to new additions such as anstelle 'instead', an univerbation of the original PP an Stelle 'in stead' (see Zifonun et al. 1997Zifonun et al. , 2076. Against the background of these syntactic and semantic properties of prepositions, PPs allow for variation with regard to at least the following three aspects: (a) p-choice, i.e. the semantically motivated choice of a certain preposition at the interface of syntax and lexicon, (b) aspects of case marking or even case government, i.e. the morphological marking of syntactic serialization and (c) preposition-determiner contraction (or: fusion, enclisis), i.e. the morphophonological aspect of serialization.
Hereafter, these aspects will be dealt with more closely with a focus on German and the peculiarities of German in Austria.

P-choice
P-choice interacts with the syntactic function of the PP. Depending on the function, semantic aspects of the local preposition are highlighted differently: In cases of PPs as sentence adverbials (1) or attributes of an NP (2) (adv), both aspects are retained (Zifonun et al. 1997(Zifonun et al. , 2100. If the PP has the status of a local or directional argument (3) (arg-1), e.g., of position words like sitzen 'sit', stehen 'stand', lie 'liegen', wohnen 'live' or motion verbs like gehen 'go', setzen 'cause to sit', stellen 'cause to stand', legen 'cause to lie', the relation is first and foremost expressed by the verb. The preposition only adds the specification of the region (Zifonun et al. 1997(Zifonun et al. , 2102(Zifonun et al. , 2155 In cases such as (4), when the PP functions as a prepositional, non-local or directional argument (arg-2), the preposition is semantically completely or almost completely empty (Zifonun et al. 1997(Zifonun et al. , 2155; for aspects of grammaticalization in German see Hundt 2001). In these cases, p-choice is determined by the predicate rather than by the expressed content and the preposition is consequently reduced to its ability to constitute phrases. (4)

Ich
wart-e auf besser-es Wetter. I wait-PRS.1SG on better-N.SG.ACC weather 'I am waiting for better weather.' Most prepositional constructions regarded as typical for SAG (see Section 1.2) involve p-choice in arg-1 and arg-2 PPs.

Case government and case marking
Some prepositions only govern one case, others more than one. In standard varieties of German, a subgroup of highly grammaticalized spatial prepositions, the so-called two-way prepositions (i.e. an 'at ', auf 'on', in 'in', hinter 'behind', neben 'beside', über 'above', unter 'under', vor 'in front of', zwischen 'between') either govern a dative or an accusative case depending on the exact meaning (Wöllstein and Dudenredaktion, 2016, 620, Zifonun et al. 1997, 2105. In these instances, the dative expresses a static location, i.e. an existing relation (5a), the accusative indicates a dynamic and directional change of location, i.e. an emerging relation (5b) (Leys 1989

Regal.
shelf 'The cat jumps into the shelf.' In opposition to standard varieties of German the formal distinction between accusative and dative case in Bavarian varieties is not as regular. First, they do not formally distinguish case in plural paradigms. Second, phonetic reduction processes ([m] > [n]) lead to increasing case syncretism in the masculine singular paradigms that hence develops a unified object case (in contrast to the nominative subject case) in the singular (cf. Zehetner 1978, 320, Weiß 1998. This development mostly pertains to demonstratives (and adjectives in strong declension). Note that in Bavarianas in standard varieties of Germandemonstratives generally function as the main position of case marking in the NP (cf. Weiß 1998, 44-6).
Under these circumstances, the semantic distinction of case with two-way prepositions cannot be maintained anymore. Due to the fact that case syncretism does not affect all paradigms, we hereinafter speak of variation in surface case marking rather than case government and include both syncretism and cases of (insufficient) morphological case marking due to phonological processes.

Preposition-determiner contractions
In German, certain prepositions fuse with certain forms of the following definite determiner der, die, das 'the' if it is semantically weak, e.g., an dem > am 'at the.M/N.SG.DAT', in dem > im 'in the.M/N.SG.DAT'. The Duden (Wöllstein and Dudenredaktion 2016, 627-31) codifies two layers of contraction forms for German ( Table A1 in the Appendix), none of which include feminine singular or plural determiners: • contraction forms that are obligatory in certain semantic contexts (e.g. an dem > am 'at the.M.SG.DAT', in dem > im 'in the.M.SG.DAT') and • contraction forms that are common in colloquial language and increasingly found in written texts (e.g. hinter dem > hinterm 'behind the.M.SG.DAT', hinter das > hinters 'behind the.N.SG.ACC').
The contraction of prepositions and determiners in standard varieties of German is described as a phenomenon in change (Wöllstein and Dudenredaktion 2016, 630) or a '''construction site' of grammaticalization" (as Nübling 2005 titles her paper) because only some determiners have developed clitics that either facultatively or even obligatorily contract with the preposition under a range of circumstances (for a diachronic account see Christiansen 2016). Generally, Nübling (2005) and Schiering (2005) apply a grammaticalization scale to describe the status of the single contraction forms: It starts from a wide range of allegro forms typical for colloquial registers and regional varieties of German (e.g. *and 'at the.F.SG.ACC', *ind 'into the.F.SG.ACC'). The second step are so-called general clitics in terms of Zwicky (1977), which lead to non-obligatory but increasingly common contraction forms in colloquial (written) communication. The third step are obligatory fusion forms as so-called special clitics (Zwicky 1977), which are the preliminary stage to inflectional morphemes (see Schiering 2005 for Ruhr German). For German in Austria, prepositiondeterminer contraction has not yet been investigated even though the contraction auf dem > am 'on the.M/N.SG.DAT' is amongst the most frequently mentioned typical Austrian constructions and considered part of SAG (see Section 1.2).

Method
This study sheds light on both linguistic and sociolinguistic variation of PPs in various varieties of German in Austria. The linguistic perspective explores language internal factors and restrictions for all types of variations documented in our corpus. It utilizes qualitative and descriptive methods. The sociolinguistic perspective takes social and regional variation into account and is analyzed with statistical generalized linear mixed-effects models. For this purpose, variation was operationalized binarily as the (non-)conformity with the exocentric reference variety SGG, which enabled the identification and inclusion of all (possibly) typical Austrian constructions in PPs (including case marking, which of course applies to NPs in general). By combining the two approaches in the analysis, the study gives a first insight into variation of (subtypes of) p-choice, case marking and preposition-determiner contractions on the dialect-standard axis of German in Austria.

Data
The analyzed data were recorded in two different research projects in the period 2012-2018. The rural data are part of a corpus of informal and formal varieties of German spoken in rural Austria built by Project Part (PP) 03 "Between dialects and standard varieties: Speech repertoires and varietal spectra" and PP 08 "Standard varieties from the perspective of perceptual variationist linguistics" of the Special Research Programme (SFB) "German in Austria: Variation -Contact -Perception" (Budin et al. 2019, Koppensteiner and Lenz 2017. The current investigation focuses on recordings in informal situations. The corpus data will be available online at the SFB's online platform (https://dioe.at/en/) after the completion of the projects in 2024.
The urban data were collected within the project "Investigating Parental and Other Caretakers' Utterances to Kindergarten Children" (INPUT) (Korecky-Kröll et al. 2016) that recorded parents and other caretakers when communicating with (their) children as well as in interview situations. In this context, we only include speech of parents with German as L1 into the analysis even though the corpus comprises speakers of German as L2 and interaction in other languages than German, too. Due to aspects of data privacy, the corpus is not publicly available.
The data types were selected in order to capture spoken varieties of Bavarian German in Austria that presumptively represent the extreme poles on the dialect-standard axis: formal adult-directed speech (ADS) from Vienna as (close-to-)standard and informal peer conversations as dialectal varieties. The CDS from Vienna was added in order to capture an informal, presumptively (close-to-)standard variety.
The participants, whose speech is investigated in the current publication, were 40 younger (18-45 years) and older (60+ years) adult L1 speakers of German (31 women, 9 men) living either in Vienna or in one of three different rural regions distributed over the main Bavarian dialect areas in Austria (Central and South Bavarian as well as the transition area, see Figure 1).
For determining the participants' socioeconomic status (SES) we use the educational level as the main criterion (see Hoff 2003). According to this approach, 21 participants are from higher socioeconomic status (HSES), as they have educational levels of at least high school diploma, whereas 19 participants are from lower socioeconomic status (LSES), with educational levels below high school diploma. Table 1 gives an overview of participants' locations and sociodemographic variables.
However, it must be noted that in rural participants, SES and age had to be treated as one composite variable because the rural sample comprised only younger HSES and older LSES participants. This implies TA … transition area  that we can investigate the SES variable alone only when comparing rural and urban younger adults or among the urban adults who belong exclusively to the younger age group. The analyzed corpus consists of spontaneous speech recorded in different settings. The rural participants were asked to have a 1-h peer conversation with a friend on various, given topics (e.g., language attitudes, hobbies; see Koppensteiner and Lenz 2017, 60-1).
The urban participants, who were all parental main caretakers of young children aged 3-4, were recorded in natural parent-child interaction with their children. These situations showed considerable variation: Some parents engaged in spontaneous conversation with their children, others decided to read storybooks or play games. Overall, there were four 30-min recordings of parent-child interaction per parent over a period of 1 ½ years. Furthermore, the same parents were asked to give a so-called parent interview (total: 30 min, audio-recorded at the third or fourth point of the survey in most cases). In this interview, the researcher asked the parents about their daily routines and preferred activities with their children, their children's development and their expectations for the future.
To sum up, this study is based on recordings of two main settings, namely ADS and CDS. The ADS data furthermore comprise more formal interview data from the urban and less formal peer conversation data from the rural corpus. All data were transcribed by skilled student assistants. Transcription and basic tagging of parts-of-speech and morphology of the urban data was done in CHILDES (MacWhinney 2000), whereas EXMARaLDA (Schmidt and Wörner 2014) was used for transcription and basic tagging of the rural data. Additional tags for PPs were added in MS Excel (using a modified tagset of the one developed in Korecky-Kröll and Buchegger 2018).

Annotation
In the annotation process, special attention was paid to the three aspects of variation in PPs described in Section 1.3: p-choice, case marking and preposition-determiner contraction. Cases which would require a preposition in SGG were included even if the preposition was omitted in the actual utterance. Generally, the annotation system was designed to capture deviation from a normatively defined SGG as codified by the Duden (2019) and the Dictionary of contemporary German (DWDS 2019) with regard to the three aspects of PP-internal variation. To enable the identification of phenomena in several varieties of German in Austria including SAG, we opted for SGG as the reference variety. Therefore, Austrian codifying publications such as the Austrian dictionary (ÖWB 2018) or publications that focus variation in the standard varieties of German (Ammon et al. 2016;Dürscheid et al. 2018) were not consulted in the annotation process. However, they are considered in the interpretation and discussion of results (see Section 4).
If the Duden and the DWDS allowed for more than one realization of the PP regarding p-choice, we chose the one more frequently represented in the DWDS corpora, which was identified with the DWDS word profile function. Concerning preposition-determiner contractions, all contraction forms mentioned by Wöllstein and Dudenredaktion (2016, 627-31; see Table A1), i.e. obligatory and highly common forms, were marked as compliant with SGG.
After excluding two unclear cases, our final dataset consisted of 4,761 concordances including PPs with two-way prepositions as their heads.

Statistical analyses
In the statistical analyses, we used the lme4 package (Bates et al. 2015) of R (R Core Team 2018) to conduct generalized linear mixed-effects models (glmer) of the relationship between participants' production of PPs (the dependent variable) and various independent variables related to the participants, the method and the linguistic properties of the PPs. Fixed participant variables were age group, SES, urban vs rural as well as the location. A fixed variable related to the method was the setting (ADS vs CDS) and the formality of the speech (parent-child conversations as well as conversations with friends were classified as informal, whereas interviews were classified as formal). Participant ID and the preposition were entered as random variables. The dependent variable was always binomial: Two-way PPs that were morphosyntactically identical to the preferred SGG forms were coded as 1, whereas non-SGG-conform forms were coded as 0. We analyzed three different dependent variables: accordance of (1) p-choice, (2) PP-internal surface case marking and (3) preposition-determiner contraction.
In the statistical analyses, different independent variables were tested in separate models and on variouseither regionally or socially restrictedsubcorpora to estimate their impact on the dependent variable, namely deviation from SGG regarding one aspect of inner-PP variation. Potential interaction effects were examined, too, and are reported if significant. For reasons of space, most of the models without significant interaction effects are not reported.

Results
This section presents the results for each dependent variablep-choice, case marking and prepositiondeterminer contractionsseparately. First, the non-SGG-conform constructions are described in order to identify subtypes and linguistic restrictions. Then, statistical analyses uncover the sociolinguistic distribution of deviations from SGG, before both approaches are integrated.

Descriptive results
In the full corpus, 187 of 4,763 (i.e. 3.93%) examples deviate from SGG regarding p-choice. It is thus the least common type of deviation. Three instances are excluded from the further analysis due to their character of spontaneous self-corrections or mistakes (A in the Appendix). The remaining 184 examples were categorized according to their syntactic status (adv, arg-1, arg-2) and semantic roles expressed by them. The largest number of examples (59.24%) are PPs with arg-1 status, whereas the share of PPs with adv and arg-2 status is comparably low (cf. Table A2 in the Appendix). The more detailed description groups the non-SGGconform PPs according to their semantic role and the realized preposition. Note that only four of these semantic role/preposition pairs are documented by more than ten examples. The corpus is hence too small to allow for judgements regarding the regularity of most constructions. It is, however, striking that 66% of deviations include the preposition auf 'on, onto' instead of another, SGG-conform preposition: Three of these four semantic role/preposition pairs that are documented by more than ten instances also include auf, namely PPs with temporal adv status (6),¹ and PPs with arg-1 status expressing either LOCATION (7) or GOAL (8). The fourth well documented type also pertains to arg-1 PPs expressing GOAL. In this case, however, the preposition is dropped (9).
 1 Note that (6a) includes an informal Austrian lexical equivalent of standard German am Abend 'in the evening', namely auf die (>auf d) Nacht. Nine out of 11 instances of temporal adv PPs involve this collocation and only 2 other NPs (6b).

Sociolinguistic analyses
Regarding p-choice, the statistical analyses mainly yield significant effects of areal independent variables on the degree of deviation, whereas social or setting-specific variables play a minor role (Tables A3 and A4 in the Appendix). The strongest effect can be seen when comparing urban and rural participants: Urban participants produce significantly more SGG-conform prepositions than rural participants (model A.4: z = 6.131 * * * , AIC = 861.7741). This result is reflected in the models for diastratically defined subcorpora and in most models calculated with place as the independent variable. A closer look at the data reveals that participants from VIE show a very high rate of SGG-conform prepositions (98.19%). However, participants from different rural locations behave quite differently, too. In comparison to NEC, participants from WEI produce significantly fewer SGG-conform prepositions (z = −2.635 * * ), whereas there is only a nonsignificant trend for NEU (z = −1.876 · ) and a significant positive effect for Vienna (z = 2.746 * * ), as shown in model (A.3) (AIC = 859.4079), the best model for the complete sample. Similarly, in model (A.8), the best model for the rural subcorpus, WEI also shows a negative effect of medium significance (z = −2.945 * * ), whereas NEU shows only a slightly significant effect (z = −2.176 * ) in comparison to NEC (AIC = 509.1945). Interestingly, the older participants in WEI seem to be decisive in this effect: Model (B.4) calculated for the rural data reveals an interaction trend of SES/age and place (z = 1.897 · ), meaning that older participants (with LSES) from WEI produce more non-SGG prepositions than others. Model (A.14), the best model for the LSES subcorpus, supports these results by showing a significant negative place-effect on SGG conformity for WEI (z = −3.228 * * , AIC = 371.2599).
The only significant effects of social or setting-related independent variables were obtained when examining the full corpus: In the complete sample, younger participants produce significantly more SGG-conform prepositions (see model (A.2), young: z = 2.776 * * , AIC = 883.165) as do participants when engaged in CDS (see model (A.5), CDS: z = 2.829 * * , AIC = 882.4053).
To sum up ( Table 2), we can identify a group of participants prone to use SGG-conform prepositions particularly frequently, namely young urban speakers in conversations with their children (1.62% non-SGGconform prepositions). On the other hand, elderly participants from WEI tend to produce the highest percentage of non-SGG-conform prepositions (13.85%). Interestingly, WEI is the only rural place, in which a difference between old and young participants regarding their p-choice can be found, even though it is too small to be statistically significant. Figure 2 gives an insight into the areal and social distribution of the above described types of p-choice that are quite well documented in the corpus. We find that except for p-drop in arg-1 PPs expressing GOAL, all types can be found in the close to standard registers documented in VIE as well. However, many of these examples pertain to the use of phraseolexemes with the nouns Besuch 'visit' or Urlaub 'holiday' such as (12) (arg-1 LOC; 10 of 14, i.e. 71.43%) and (13) (arg-1 GOAL; 13 of 24, i.e. 54.17%) that are typical for SAG (see Section 4). The most obvious regional differences can be seen regarding the use of either auf 'on, onto' or p-drop with arg-1 PPs expressing GOAL. Whereas in South Bavarian WEI p-drop prevails, participants from the places in the Central Bavarian or transition area prefer PPs with auf. Focusing on all directed motion constructions of the form intransitive motion verb + ARG GOAL and their realization in the three rural locations (41.5% of all deviations in p-choice in the rural data and 41.4% in the complete data set),² we find that between 34% and 44% of all directed motion constructions deviate from SGG. As expected, in the Central Bavarian (NEU) and the transition area (NEC), PPs with the preposition auf 'on' are chosen in these cases. In WEI (South Bavarian) p-drop prevails (Figure 3)   2 VIE is excluded due to the generally low frequency of deviations from the standard and their tendency to be restricted to certain phraseolexemes.
3 One reviewer remarked that p-drop obviously only occurs with allative directional phrases, but never with elative ( * I kumm Schui 'I come from school'). We have to admit that our data cannot account for elative directional phrases, because they are usually not realized with two-way prepositions but with others like von (+DAT) 'from', aus (+DAT) 'out of'. However, in the complete corpus data from WEI we can observe the assumed restriction. It remains to be analyzed.

Descriptive results
In the full corpus, we find 292 non-SGG-conform PPs due to aspects of case government or case marking. It is thus the most common type of divergence in the corpus and pertains to 6.13% of the total number of examples. As Table A5 (in the Appendix) shows, four main subtypes can be identified in this category, the most frequent of which is determiner elision illustrated by examples (14) and (15) In (14), the NP is expected to be in dative case, whereas in (15) it is expected in accusative case. However, in both instances, the determiner is elided and the position, in which case could be marked, hence disappears. The data do not allow us to reliably reconstruct the underlying case forms. Interestingly, however, 150 out of 154 instances are found with the preposition in 'in, inside', which suggests that a contraction form is required as an intermediate step in the reduction. Therefore, this subtype of complete determiner elision is probably a further development of preposition-determiner contraction (see Section 3.3) in spoken language, which leads to surface forms indistinct regarding case.
The second and third subtypes include instances of case syncretism in either plural or singular paradigms. In plural paradigms, it occurs almost twice as often than in singular paradigms. In all these 82 cases, dative marking is expected, but accusative forms of determiners, adjectives and/or nouns are realized (16,17). Such examples occur regularly with nouns of all grammatical genders as well as with pluralia tantum, which suggests that in the Bavarian varieties in Austria, dative and accusative plural tend to syncretize.  The examples of case syncretism in singular paradigms are rarer and less regular than those in plural paradigms. More precisely, three patterns that correspond to the position of deviation in the NP listed in Table A6 can be identified (18,19). In (18), an expected dative determiner is in fact realized in accusative case. Examples such as (19) illustrate that the dative case of possessive pronouns may be identical with the nominative/accusative neuter or nominative masculine form.  Both examples (18, 19) possibly involve phonetic reduction of final [-m] > [-n], a process that was already observed in the context of determiner elisions. Examples with feminine determiners, in which this type of reduction cannot take effect, occur only twice (B in the Appendix) and have a similar ungrammatical character to the last subtype of deviations from SGG regarding case government or case marking, namely the choice of non-canonical case forms (C and D in the Appendix). Interestingly, out of the nine cases of ungrammatical form choice, six are realized by the two speakers of (C) and (D) in informal settings. Therefore, these deviations from the standard, which mostly include the use of a dative instead of an accusative form can be judged errors or idiolectal features.⁴ 

Sociolinguistic analyses
The statistical analysis (Tables A7 and A8 in the Appendix) shows that in comparison to p-choice, variation concerning case government is structured rather socially. Regarding this aspect, LSES participants produce significantly fewer SGG-conform PPs than HSES participants (model C.1: z = −3.528 * * * , AIC = 1923.052). Additionally, we found that participants from urban VIE generally use more PPs with SGG-conform case (model C.3: VIE: z = 3.767 * * * , AIC = 1916.31; model C.4: urban: z = 5.239 * * * , AIC = 1913.523). Model (D.4), the best model for the full corpus, also shows a strong interaction effect between SES and urbanity: LSES participants from the urban region use comparably fewer SGG-conform case forms within PPs than the urban HSES participants, whereas LSES participants from rural regions do not differ so much from the HSES rural participants (SES: n.s., urban: z = 8.155 * * * , LSES:urban: z = -4.036 * * * , AIC = 1884.063; see Figure 4). Accordingly, the best model for the urban subcorpus shows a strong SES effect (model C.7: z = −6.014 * * * , AIC = 860.2739).⁵ In a nutshell (Table 3), deviations from the SGG-conform case in PPs seem to be comparably likely throughout various social groups in Bavarian speaking, rural Austria. Generally, speakers from VIE tend to realize SGG-conform case forms more often than speakers in the rural locations (χ 2 (1) = 143.056 * * * , V = 0.173). Within VIE (χ 2 (1) = 74.925 * * * , V = 0.147) and NEC (χ 2 (1) = 4,847 * , V = 0.100), the social background seems to be decisive insofar as LSES speakers deviate from SGG more often than HSES speakers. In the other rural locations, the difference is not significant.
A glimpse into the distribution of subtypes of non-SGG-conform case marking ( Figure 5) allows for the observation that both determiner elision and syncretism in singular paradigmsboth of which can possibly be traced to phonetic reduction processes rather than to morphologyare spread diatopically and diastratically throughout Bavarian speaking Austria. Syncretism in plural paradigms is, however, most common in South Bavarian WEI and least common both in Central Bavarian NEU and VIE.  5 Note that, for the rural subcorpus, the sociodemographic variables of SES and age had to be treated as one composite variable as our rural participants were either younger participants of HSES or older people of LSES backgrounds. Eventually, the social structure of our participants also results in the SES effect triggering an age effect as found for the complete data sample (model C.2: young: z = 3.059 * * , AIC = 1926.217). Effects of setting were not identified regarding case government.

Preposition-determiner contractions 3.3.1 Descriptive results
Throughout the full corpus, 278 out of 4,763 examples (i.e. 5.84%) deviate from SGG in the contraction forms of preposition and determiners used. Out of these, one instance co-occurs with non-canonical p-choice and therefore does not allow us to reconstruct the full determiner form reliably (E in the Appendix). Therefore, it will be left aside in the analysis, which hence pertains to 277 instances. In these, we find no evidence for non-SGG-conform contraction forms with plural determiners or singular determiners in genitive case.
Table A8 (in the Appendix) gives an overview of non-SGG-conform preposition-determiner contractions found. In the case of feminine determiners (25 examples), the vowel is fully reduced and the initial consonant merges with the last syllable of the preposition (e.g., in de schui > ind schui 'to school'). A similar process can be observed with the masculine accusative determiner following the preposition in 'in, into' (two examples), where both the initial and the final consonant are retained in the preposition's last syllable offset (e.g., in den gortn > indn gortn 'into the garden').⁶   6 One reviewer correctly remarked that though the surface output of these two processes resembles each other, they might be different. We suggest that this difference is probably due to a different degree of grammaticalization: While the contraction of semantically weak feminine determiners with the preposition in 'in' is obligatory in Bavarian dialects (similar to that for neuter The vast amount of non-SGG-conform contractions (249 examples, i.e. 89.89%) occurs with masculine accusative or dative and neuter dative determiners following the preposition auf 'on, onto' (20-2). All these cases result in the contraction form am, which is homonymous with the standard conform contraction of the preposition an 'at' and the masculine or neuter dative determiner dem

Sociolinguistic analyses
The dependent variable of deviations from the standard regarding contraction did not yield any significant effects for the full corpus. Apparently, such deviations from SGG are widespread all over (Bavarian speaking) Austria without showing any regionally or sociodemographically determined variation in frequency. We were only able to identify weak effects and nonsignificant trends within some subcorpora: • First, young HSES participants from NEU apparently use more non-SGG-conform contractions than those from NEC (z = −1.835 · , AIC = 776.2769; young HSES subcorpus), whereas older LSES participants from WEI use fewer (z = 1.852 · , AIC = 646.0164; LSES subcorpus) than those from NEC. • Second, when examining the rural subcorpus, slight SES/age effects as well as interaction effects between SES/age and place were found (LSES: z = −2.281 * , place: n.s., LSES:WEI: z = 2.182 * , LSES:NEU: z = 2.044 * , AIC = 394.0726). This model has a slightly worse fit than a less complex model without interaction that does not yield any effect of SES or age (AIC = 392.9336).
 determiners in standard variants of German), the contraction form of the masculine determiner with the same preposition is a phonologically conditioned allegro form. Therefore, in de schui 'to the school' probably has a slightly different meaning than ind schui 'to school', while in dən goatn and indn goatn 'to the garden' do not. However, this article does not give sufficient evidence to empirically substantiate this interpretation and further research should focus on the aspect of grammaticalization degree of contraction forms in Bavarian dialects of German in Austria.
• Third, we find a weak SES effect as well as a trend of an interaction between SES and setting in the urban subcorpus (LSES z = −2.325 * , setting: n.s., LSES:CDS: 1.772 · , AIC = 1004.858). Apparently, urban LSES adults use slightly fewer SGG-conform contractions but slightly more in CDS than in ADS, whereas their HSES peers show the opposite tendency ( Figure 6).
The weak effects and non-significant trends detected in the statistical analyses are depicted in the data as shown in Table 4. Taking the complete place-specific subcorpora into account, we interestingly observe that in places from the Central Bavarian (VIE, NEU) and the transition area (NEC), non-SGG-conform contractions are slightly more common than in Southern Bavarian WEIa tendency probably not reflected in the statistical models due to the very low effect size of this difference (χ 2 (1) = 7.060 * , V = 0.039).
In almost all places, SES does not make any difference regarding the probability of an informant to produce such contractions. Only in NEC, the fact that HSES participants produce 5.55% fewer non-SGG-conform contractions than LSES speakers yields a significant difference (χ 2 (1) = 7.316 * , V = 0.123). Similarly, the above described tendency of LSES parents from VIE to use more non-SGG-conform contractions but fewer with their children, is very weak. The two groups virtually produce the same percentage of non-SGG-conform contractions except for a significant but very small difference in ADS (χ 2 (1) = 7.938 * * , V = 0.076), which then results in the greater difference in SGG-conform vs non-SGG-conform contractions between their ADS and CDS (χ 2 (1) = 17,886 * * * , V = 0.091) than it is the case for LSES parents, where the difference is not significant.⁷  To sum up, informants from all Bavarian speaking areas in Austria use certain non-SGG-conform contractions. Additionally, they occur both in rural dialects and in urban, (close-to-)standard speech styles.
The question arises whether the above described contraction types are distributed evenly throughout the corpus as well. Figure 7 allows for preliminary assumptions even though the number of examples especially for the rural places is limited. Obviously, the contractions of feminine determiners in accusative are more common in the rural dialects (from the Central Bavarian and the transition area) and hardly occur in urban more formal varieties or CDS. Contractions of the preposition auf 'on' and masculine or neuter determiners, however, are common throughout all varieties in the corpus.

Discussion
Overall, the three linguistic main variables, p-choice, case government and preposition-determiner contractions, yield quite different results: While we find mostly regional effects for p-choice (particularly in directed motion constructions), case marking shows more social variation (especially with respect to SES). The non-SGG-conform use of preposition-determiner contractions turns out to be widespread across all regions and social groups investigated and appears even in near-standard registers.

P-choice
Regarding p-choice, we do not only find areal variation between rural regions and the urban region of VIE but also variation between different rural regions: WEI has a significantly lower rate of SGG-conform p-choice than the other rural locations. Particularly, older participants from WEI show a particularly high rate of non-SGG-conform p-choice, which can be attributed to frequent p-drop.
As indicated in Section 1.2, despite being frequently discussed in the context of ethnolects in German and Austrian cities, this very salient phenomenon is also found in native German in Austria. On the one hand, Kerschhofer-Puhalo (2019) reports p-drop in directed motion, locative and temporal constructions in informal speech of autochthonous persons of various ages living in Vienna. On the other hand, p-drop is regarded typical for South Bavarian dialects (see Oberdorfer and Weiß 2016;Lenzhofer 2017). Pohl (1989Pohl ( , 2009    with Slovene typical for Carinthian dialects. Our analysis gives evidence for frequent p-drop South Bavarian dialects (in our case in WEI), particularly in directed motion constructions such as (23). The few instances of p-drop from the other rural locations (2 from NEC and 4 from NEU vs 17 from WEI) attest the existence of p-drop outside of the South Bavarian area. However, it is less frequent and pertains to different constructions: while 10 out of 17 examples from WEI (i.e. 58.82%) are directed motion constructions, temporal and locative contexts prevail in the other locations. The data from VIE do not include any instance of p-drop, which substantiates the claim by Kerschhofer-Puhalo (2019, 141) that p-drop is sensitive to stylistic variation and obviously not adequate in (close-to-)standard contexts. Whether this is true for the South Bavarian area requires follow-up investigation.
In our sample, auf 'on, onto' is the most frequently selected non-SGG-conform preposition, which indicates an overrepresentation of auf 'on' in German in Austria. As far as the dialects are concerned, our results from four Bavarian regions of Austria are consistent with Mikosch's (1987, 46) data on Alemannic varieties, who also finds an overrepresentation of auf 'on' in the Austrian region of Vorarlberg in comparison to different Alemannic regions of Germany, particularly in the younger generation (which was born in approximately the same time as the older generation of our current study). In contrast to p-drop, the use of auf 'on' in non-SGG-conform PPs is particularly relevant in VIE, NEU (Central Bavarian) and NEC (transition area), where it most frequently appears in directed motion constructions such as in (24a). Contemporary written SGG prefers constructions such as (24b). Generally, directed motion constructions are the context with most non-SGG-conform instances of p-choice. In the rural locations, 34-44% of all cases deviate from SGG by either selecting auf 'on' or p-drop. This suggests the need for a generic directed motion construction in spoken German in Austria, i.e. the need to express the goal of a motion event without specifying its exact location relative to the PP inner object (e.g. inside university or next to university). Whether such a generic construction exists requires further research. Two aspects require consideration in this context. First, some of these constructions (auf Besuch gehen 'to visit somebody', auf Urlaub gehen 'to go on holiday') and their local equivalents (auf Besuch sein 'to visit somebody', auf Urlaub sein 'to be on holiday') can be characterized as phraseolexemes. These are common in and typical for SAG (see maps in Dürscheid et al. 2018: auf/zu Besuch, auf/in (den) Urlaub gehen/fahren, auf/im/in Urlaub sein). As shown in Section 3.1, these phraseolexemes account for over half of all non-SGG-conform p-choices in (close-to-)standard registers from VIE, which corresponds to the expected pattern.
Second, these phraseolexemes as well as several other constructions including directed motion constructions in general as well as p-choice in PATIENT arguments of mental verbs (denken 'think', erinnern 'remember' and vergessen 'forget') can be analyzed in the context of language contact with Slavic languages or the Central-European area of linguistic convergence, respectively (see Newerkla 2007). The constructions mental verb + ARG auf+ACC have structural equivalents in neighbouring Slavic languages such as Czech (myslet na něco 'think about something', vzpomínat se na něco 'remember something' zapomínat/zapomenout na něco 'forget about something'); however, it is not yet clear whether all of them correspond to the respective constructions in German in Austria regarding their semantic and/or syntactic distribution as shown for vergessen and zapomínat/zapomenout 'forget' in .
Regarding directed motion constructions, already Schuchardt (1884, 115) lists several examples with non-SGG-conform choice of auf 'on' in the "jargon" of Slavic and especially Czech or Polish L2-learners of German (e.g. auf die Schule/Predigt/etc. gehen 'to go to school/the sermon/etc.'), suggesting a language contact explanation (through pattern replication). However, in Czech linguistics the generic use of the equivalent Czech preposition na 'on, at' in directed motion constructions is only being discussed as an innovation in spoken and written standard registers from the 1960s onwards (e.g. Mára 1968;Čechová 1981;Hrdlička 2015). Mára (1968, 134) assumes this change to have begun in the 1940s. Still, earlier descriptions of Eastern Moravian dialects suggest that it was common at least in some parts of the Czech language area in the nineteenth century (Bartoš 1886, 186). Generally, much research remains to be done regarding the areal distribution and diachronic development of (generic) directed motion constructions in German and Czech. It is therefore not possible to unequivocally speak of the use of the preposition auf 'on' in such contexts in German in Austria as a contact-induced phenomenon. Since the patterns used in German in Austria and Czech seem to be structurally and semantically equivalent, we may however treat the (generic) directed motion construction with Germ. auf 'on' or Czech na 'on, at' as an areal, Central European phenomenon (see Januška 2020 for a literature review on the Central European linguistic area) andin terms of Muysken (2010Muysken ( , 2013 suggest convergent development under prolonged stable bilingualism. Probably, other languages in Central Europe developed similar patterns (see Mára 1968;Newerkla 2007).
Interestingly, Pohl (1989, 64) assumes that p-drop in directed motion constructions has spread from the Slovenian-German bilingual area in Southern Carinthia. In later publications, however, he expresses uncertainty whether this phenomenon can be traced to language contact or not (Pohl 2009, 127). He presumes the Slovenian two-way prepositions v 'in' and na 'on, at', which reflects the older Slavic system of spatial prepositions (see Herrity 2000, 262), to be incompatible with the South Bavarian system that according to his opinion prefers auf 'on' in any case, similar to what our data have revealed for other Bavarian varieties in Austria. According to Pohl's early assumptions, German-Slovene bilinguals in Southern Carinthia might have chosen to drop the semantically almost emptied preposition in generic constructionsa strategy that can be referred to as the recourse on universal principles governing improvised language behaviour (Muysken 2013, 716). However, like the situation for Czech and German, the distribution and use of v 'in' and na 'on' in directed motion constructions in Slovene dialects requires further investigations to substantiate this assumption.
To sum up, our analyses show that p-choice varies areally throughout Austria. Partly, this variation, especially regarding directed motion constructions, may possibly be explained by areal convergence in Central Europe and language contact with neighbouring languages.

Case marking
Most concordances in the complete sample deviate from SGG due to case marking within the PP. These can mainly be traced to widespread case syncretism in Bavarian dialects. In the statistical analyses, deviations concerning case government were strongly associated with sociodemographic variables. Both SES and age displayed highly significant effects, but the strongest effects were shown by a model that investigated an interaction between SES and region (urban vs rural): whereas LSES participants from the urban region used comparably fewer SGG-conform case forms within PPs than their urban HSES peers, the LSES and HSES participants from the rural regions showed only slight differences. These findings are underpinned by the separate analyses of data sub-samples: in the urban data, the most significant SES effect concerns case marking, which indicates that LSES people use fewer SGG-conform case forms within PPs than HSES people, while in the rural data, we do not find any significant effects for case marking. The examination of only data from younger participants yields the urban/rural contrast, too: Young people from VIE use considerably more SGG-conform case forms than young people from the rural regions. An additional SES effect can also be found only for the urban data. If the urban data are examined separately, generally only SES effects (regarding several aspects of variation in PPs taken account of here) are found. Our results are thus consistent with those of language attitudinal studies (e.g. Moosmüller 1991;Steinegger 1998), according to which dialect use is associated with 'lower social classes' in urban areas, whereas dialect is widespread regardless of people's social status in rural regions.
Concerning case syncretism, we find evidence for a unified case (germ. Einheitskasus) in noun plurals as well as indications for such a common case in masculine and neuter singular nouns. As outlined in Section 1.3, the unified case in plural paradigms as well as the tendency for case syncretism in the masculine singular has been described earlier (see Zehetner 1978, Weiß 1998, whereas accounts focusing on syncretism in neuter paradigms are lacking to our knowledge. Regarding the areal distribution, our data indicate that syncretism in plural paradigms is most common in South Bavarian WEI (see also Lipold 1976 on the East Tyrolean location Kals) and least common both in Central Bavarian NEU and VIE. In the latter, the SES effect also applies to the frequency of syncretism in plural paradigms.

Preposition-determiner contractions
Preposition-determiner contractions are widespread across all Bavarian regions of Austria and show little regional and sociodemographic variation. Accordingly, the statistical analysis of the complete dataset does not show any significant effect, neither for regional nor for sociodemographic or setting variables. The separate analyses of sub-samples of our data however suggest that for certain groups of informants the use non-SGG-conform contractions varies between settings or regarding frequency. In the rural data, we find a slight effect of SES/age on contractions, indicating that older LSES participants use fewer SGG-conform contractions than younger HSES participants. In the urban data, a slight trend is found for an interaction between SES and setting: whereas HSES adults show a slightly higher SGG-conform use of contractions in ADS as opposed to CDS, LSES adults use slightly more SGG-conform contractions in CDS. These results suggest that HSES parents are more likely to adapt their utterances to the formal interview-setting with the researcher, while LSES parents might focus on their children's educational prospects, which corroborates earlier attitudinal findings: according to Wiesinger (2008), urban parents believe these prospects to increase if children grow up with (close-to)-standard varieties, and therefore prefer (close-to)-standard varieties over dialect. Wiesinger (2008) describes these preferences as specific for urban areas.
Even though the proportion of non-SGG-conform contractions does not significantly vary within the analyzed corpus, the investigation of the most common contraction types suggests that these are subject to areal as well as register variation. For instance, non-SGG-conform contractions of prepositions with feminine singular determiners in accusative case (auf die > aufd 'on(to) the.F.SG.ACC', in die > ind 'in(to) the.F.SG.ACC') are most common in rural dialects of the Central Bavarian and the transition area. According to our data, besides the standardized contractions of an 'at' with masculine or neuter determiners in dative case, the contraction form am has three possible sources involving the preposition auf in German in Austria: (1) auf dem 'on the.M.SG.DAT', (2) auf dem 'on the.N.SG.DAT' and (3) auf den 'on(to) the.M.SG.ACC'. Despite not being codified by the Duden (Wöllstein et al. 2016), (1) and (2) are regarded SAG even in written texts (see Dürscheid et al. 2018: Gebrauch von Präpositionenam/auf dem 'Usage of prepositionsam/auf dem'). This status is also evident in our data: in all investigated locations at least 50% of all combinations of auf 'on' and the masculine singular dative determiner (1) are contracted. They are least common in the South Bavarian dialect of WEI. Contractions of type (2) are not attested in WEI at all but are quite common in the other locations and varieties (at least 35% of all possible instances in the corpus). The form am as a result of the contraction of auf with the masculine singular accusative determiner (3) is explicitly marked 'incorrect', i.e. as not part of the standard variety by Ebner (2008: 46). However, according to our data it is equally distributed throughout different (Bavarian) varieties of German in Austria. Approximately 25% of all possible instances are contracted.
Notably, like in standard German, there are no contractions involving plural determiners. Thus, contractions may be regarded as singular markers, particularly in such dialects where contractions are found in all genders.
Altogether, contraction forms appear more common in Central Bavarian dialects and such from the transition area than in South Bavarian varieties. Contractions leading to the form am are frequently used in urban (close-to-)standard registers as well.

Conclusion and future research perspectives
This study investigated recordings of informal and formal speech from the three main Bavarian dialect areas in Austria in order to inductively describe the (sociolinguistic) variation of and within PPs with twoway prepositions. As point of comparison, we used a Germany-centred and normative notion of standard German. In our context, variation is thus operationalized as the number of non-SGG-conform PPs. Generally, PPs may vary regarding three structural aspects, namely (1) preposition choice (p-choice), (2) case marking and (3) preposition-determiner contractions. To make a long story short, (1) p-choice considerably varies across dialect regions, particularly in directed motion constructions. The South Bavarian data stand out due to the high frequency of p-drop, whereas in other regions, non-SGGconform usage of auf 'on(to)' is particularly frequent. As elaborated in Section 4, these phenomena can be investigated in the context of linguistic convergence in Central Europe.
(2) Non-SGG-conform PP-internal case marking can be explained by dative-accusative case syncretism in Bavarian dialects, which is particularly frequent in plural contexts and appears to be most prominent in South Bavarian varieties. In the analyzed corpus, non-SGG-conform case marking is mainly associated with the informants' SES: In urban areas, LSES informants tend to use significantly more non-SGGconform case forms than HSES informants. No SES-related difference can be found in rural locations. This provides evidence for the fact that in the urban areas, dialectal features are related to the lower classes and are thus social markers, whereas in rural Austria this is not the case.
(3) Non-SGG-conform preposition-determiner contractions are distributed relatively equally across all varieties, groups and registers in focus, but are limited to singular contexts. However, the fact that LSES parents use fewer contractions when speaking to their children hints at a high degree of salience. This probably leads parents with lower standard competence to avoid certain contraction forms if raising their children in (close-to-)standard varieties. Moreover, the results indicate that amresulting from the contraction of auf 'on(to)' and either the masculine singular determiner in dative or accusative or the neuter singular determiner in dativeis quite frequent in rural dialects as well as urban (closeto-)standard varieties despite the fact that it is not regarded part of SAG with masculine singular accusative determiners. Generally, contractions seem to be more frequent in the Central Bavarian and the transition area.
This study has of course some limitations: The participant sample is rather small and the regional distribution of the locations does not allow final conclusions about the entire Austrian area. Thus, future studies should focus on single aspects based on more data from more participants (including bilingual speakers), more locations and more different settings (particularly formal interviews in addition to the informal conversations of the rural participants). Regarding p-choice, investigations into the diachronic development with a contrastive character promise to yield insights into aspects of language contact and areal convergence in Central Europe. For such analyses, newspaper corpora may complement contemporary spoken data, as many of these specific Austrian variants are used in (close-to-)standard registers. Concerning case syncretism in plural or singular paradigms in German in Austria, further state-of-the-art sociolinguistic research is needed in order to determine its regional distribution and 'vertical' variation. Furthermore, preposition-determiner contractions should also be investigated in their diachronic development to be able to trace back grammaticalization processes and relate them to results on Alemannic varieties or Ruhr German.
This study tried to lay the foundations for this type of future research by providing data-based hypotheses. Such research will provide insight into the rich regional, social and situational variation of PP twoway prepositions in German in Austria. Funding information: This research was funded in whole, or in part, by the Austrian Science Fund (FWF) within the Special Research Programme "German in Austria: Variation -Contact -Perception" (FWF F60-G23), Project Parts "Between dialects and standard varieties: Speech repertoires and varietal spectra" (FWF F06003) and "German and the Slavic languages in Austria: Aspects of language contact" (FWF F06006). For the purpose of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.

Abbreviations
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.