Intra-individual variation in adults and children: measuring and conceptualizing individual dialect–standard repertoires

1 Introduction: Standard-dialect variation in Bavarian-speaking Austria

Wandruszka (1975, 1979 coined the term ‘internal multilingualism’ (“innere Mehrsprachigkeit”) in order to refer to the co-existence of several varieties of one language within one speech community and within one speaker’s mind. He explicitly referred to the situation in Austria, stating that a human language is a complex, flexible and dynamic polysystem, a conglomerate of languages that merge into each other (Wandruszka 1979: 39). Besides the notion of ‘internal multilingualism’, the Austrian sociolinguistic setting has been described in terms of a ‘polyglossia’ (Wiesinger 1992), a ‘standard-dialect continuum’ (Ammon 2003), an ‘attenuated diglossia’, or a ‘diaglossia’ (Auer 2005). Accordingly, the relation between standard (Austrian) German and (Bavarian) base dialects has been conceptualized as a continuum or as one of more or less gradual transitions between several speech levels or varieties; with the exact number of postulated speech levels being still up for discussion (Kranzmayer 1953; Scheutz 1999; Wiesinger 1992, 2014).

Most of these conceptualizations were derived from informed, yet informal and inevitably limited and partial observations and introspections by linguists. More recent survey data partially confirm the psychological reality of ‘internal multilingualism’ for the majority of speakers in the Bavarian-speaking part of Austria: 72% (n = 82) of Austrians in the Bavarian dialect region report mastering both the regional base dialect and standard German (Ender and Kaiser 2009). For a long time, however, these conceptualizations of language variation in Austria had barely been backed up with evidence from extensive authentic speech data from the relevant speech community.^[1]

Social macro-factors such as age, level of education, and gender are well-known to correlate with language (variety) use, and these correlations have also been observed in the Austrian context (Steinegger 1998). Along these lines, people in rural areas or smaller cities and/or people with a lower educational background and/or people at an older age are generally expected to have better dialect knowledge. Importantly, however, in the history of sociolinguistics, these broad social categories have also been shown to offer only limited explanatory power with regard to language variation (e.g., Eckert 2012). In fact, these categories form relevant intersections, they interact with other factors and develop very different potentials depending on one’s current place of living and the current stage in one’s life. The differences in dialect proficiency and dialect use between rural and urban areas, older and younger speakers, and well- and less well-educated speakers, for example, are less pronounced in certain geographical areas of Austria than in others (Steinegger 1998). Lichtenegger (2015) showed that mothers tend to use less dialect and more standard-near language when speaking to their small children than when communicating in everyday interactions with friends or colleagues. Therefore, the general bias of individuals toward certain forms of speech on the standard-dialect spectrum also depends on their specific social and linguistic experiences across the lifespan as well as the (changing) social networks they belong to.

Moreover, the concrete usage of the individually available dialect-standard repertoire has been described as being influenced by social and interactional factors in specific conversations. Depending on the (expected) linguistic repertoire of the addressee, the closeness of the interlocutors and the formality of the situation, different sections from the individual dialect-standard repertoire are made use of. These ‘shifting’ routines, e.g., decreasing the use of dialectal speech from private situations toward more formal ones and from intimate to non-familiar interlocutors, are deeply rooted in speakers’ language awareness in the Bavarian-speaking part of Austria (Steinegger 1998: 372; Vergeiner 2019). However, these correlations between situational characteristics and linguistic behavior have hardly been studied by collecting speech data in actual communicative situations in Austria. Only fairly recently (with the early exceptions of Scheutz 1985 and Scheuringer 1990) have attempts been made to capture glimpses at authentic samples of intra-individual variation in the Bavarian-speaking part of Austria (Ender and Kaiser 2014; Vergeiner 2019). That language variation can be interpreted not only as a reflection of social identity but as an active stylistic practice for positioning oneself (Eckert 2012) in an interaction has been acknowledged in the so-called ‘third wave’ of sociolinguistics and has further added to the complexity of the issue.

Quantifying and adequately describing individual speakers’ language variation behaviors is particularly challenging in the present context, where reliable reference data and consensual conceptualizations of community repertoires are missing. Methodological decisions at every stage of the research process may predetermine or restrict our findings and their interpretation (see also Milroy and Gordon 2003: 138–168; van Hout 2008). In the current contribution, we bring together data from adults and children which have been collected to capture intra-individual standard-dialect variation in the Bavarian-speaking part of Austria and the acquisition thereof. Some of this data has been analyzed before (adults: Ender and Kaiser 2014; children: Kaiser n.d.; Kaiser 2019). By re-analysing the data we aim to highlight some of the methodological challenges and the consequences of methodological decisions in researching standard-dialect variation in different groups of speakers, which seem to make methodological and analytical triangulation essential. In so doing, we want to also further our understanding of the nature, the dynamics and the acquisition of dialect-standard variation in Austria.

2 Methodology

The data presented here were collected in the Central-Bavarian dialect region near the city of Salzburg. The methods used primarily exploit the mechanisms of ‘speech accommodation’ (Giles et al. 1991) or – without presupposing a social-psychological mechanism driving the effect – of speech convergence, which may also be caused by automatic priming and mimesis (Foulkes and Hay 2015: 298; Gambi and Pickering 2013; Markham 1997). By varying the interlocutor, and thus motivating intra-individual variation across situations, we aimed to capture the spectrum of varieties at speakers’ disposal. The present contribution brings together spoken data from three adults and three children.

3 Analysis I: Measuring d-values

In German variationist linguistics, measuring phonetic dialectality values (Herrgen et al. 2001; application to Austrian data in Kaiser and Ender 2013; Ender and Kaiser 2014) has become a standard method for capturing intra-individual variation across situations (e.g., Lenz 2003; Kehrein 2012). The degree of dialectality (d-value) is a measure of the phonetic distance between a certain stretch of speech and its idealized realization in standard speech, similar to the measurement of Levenshtein distances (Heeringa 2003). Each phonetic segment is assigned a score of deviation, with a maximum of two points for each consonant and a maximum of three points for each vowel. These maximum scores were adopted to reflect the distribution ratio of vowels and consonants in German (Lameli 2004). The phenomena considered include the following: differences in the levels and classes of the IPA vowel chart, substitution of diphthongs for monophthongs and vice versa, differences in the manner or place of articulation of consonants, omission of segments, etc. Phonetic differences are thus weighted according to the articulatory dimensions involved. By adding up the point values of each segment of a word, the phonetic dialectality of individual words can be determined. The ecological validity of the measure has been confirmed in different ways: dialectality scores have been shown to correlate strongly with listeners’ judgments of dialectality (Herrgen and Schmidt 1985; Kehrein 2009; Lameli 2004; and explicitly in the Austrian context Kaiser and Ender 2013), and they reflect the traditional areal structure of German dialects (Herrgen and Schmidt 1989).

In (1) we present the procedure of assigning d-scores in the example of the verb <habe> ‘have’+ 1 p. sg.

This dialectality value (d-value) is based on the following differences: no difference in the first consonant (0 points); one step up on the IPA vowel chart from [a] to [ɔ] and shortening of the vowel (2 points); no difference in the place of articulation and manner of articulation of the second consonant (0 points). The omission of the final first-person singular schwa (which leads to devoicing of the final consonant) is one of the reduction and assimilation phenomena that are typical of spoken language and is therefore not considered as a feature of dialectality. Averaging the scores of about 150 words of free speech yields the ‘dialectality value’ for this stretch of speech.

Clearly, this procedure offers several advantages. It is not necessary for the researcher to preconceive different varieties/dialects/speech levels beyond a solid definition of standard speech. The procedure yields a score which does not have to be assigned any linguistic category or term. Inter-situational variation on the phonetic level can be attested by comparing handy overall values for each situation. Figures 1a, b illustrate the d-values from our adult and child data described above.

Figure 1:

a. Dialectality-values of three different adults speaking about the same topic with four different interlocutors: (a) second language speaker, (b) German speaker, (c) speaker from another dialect region, (d) speaker from the same dialect region.

b. Dialectality-values of three different children in five different communicative situations each: (e) memory standard, (f) memory dialect, (g) story standard, (h) story dialect, (i) pretend-play.

The overview of d-values in different situations clearly attests to intra-individual variation across situations, both in adults and in children, but to varying degrees. Even though the overall tendency of using the most standard-near speech with the non-native interlocutor and the most dialectal speech with the person from the same dialect area is observable in all three adults, the three speakers vary with respect to the range covered between standard and dialect as well as the grading of dialectality in-between. The individual children’s patterns differ more strongly from each other than the adults differ among themselves, but all of them clearly exhibit intra-individual variation.

That said, this method has a number of shortcomings, too. The delineation of standard Austrian German as a base for calculating deviations is more problematic than it may seem at first glance. Whereas some may restrict the term ‘standard German’ only to those realizations of German which are devoid of any salient regional characteristics (Schmidt and Herrgen 2011), others posit a wider ‘standard’ category usually called Gebrauchsstandard (‘usage standard’) which also accounts for highly frequent regional phenomena of spoken German in formal contexts spoken by non-professional but educated speakers (Elspaß and Kleiner 2019). Most linguists today agree that there are several equivalent forms of standard German, but the question of how many there are and/or what their exact characteristics are is far from settled. While some claim the existence of an overall Austrian standard norm (proponents of the ‘pluricentric’ model of German, cf. Clyne 1984; Muhr 2007), others insist that the varieties of standard German do not conform to national borders and, in fact, rather coincide with regions defined by dialect isoglosses and other geolinguistic factors (such as large cities) (i.e., the proponents of the ‘pluriareal’ model of German, cf. Elspaß et al. 2017).

Furthermore, since the procedure of calculating d-values is based on phonetic distances, differences between two language varieties on a lexical or morphological level are not taken into consideration. Finally, one of the greatest weaknesses seems to be that this method does not give any insights into how this average score of ‘dialectality’ comes about. It may be the result of heavy mixing on the clause or phrase level, the result of switching or shifting between the two poles of the (posited) standard-dialect continuum, or it may reflect a rather stable ‘intermediate’ form of speech. It thus seems to capture intra-individual variation only on a coarse level of granularity.

4 Analysis II: Categorizing TCUs

Kaiser n.d.; Kaiser 2019 and Kaiser and Kasberger 2020 developed a different method for capturing intra-individual cross-situational variation, i.e., calculating proportions of speech in each variety. Each word is first classified as belonging to the local dialect, to standard German, to something ‘in-between’ (intermediate Umgangssprache or a mixture of dialect and standard German) or as being ‘ambivalent’ (i.e., homophones in standard German and the local dialect). Even though at this point, the difficulties in defining standard speech remain, it is advantageous that different linguistic levels – notably phonetic, morphological, and lexical levels – can be taken into account in the classification process. In the next step, utterances are segmented into ‘turn-constructional units’ (TCU; Selting 2000). A TCU is “a potentially complete turn” (Selting 2000: 480) and, as such, an interactionally relevant unit.

In the children’s case, TCUs are frequently identical with turns because the children’s turns are typically very short and only consist of one turn-constructional unit. With longer turns, e.g., comprising compounded sentences, we aimed to determine whether these more complex constructions involved additional psycholinguistic planning processes. We therefore segmented turns into TCUs by analysing the interplay between syntax and prosody. In other words, only when syntax and prosody converged, the turn was segmented into several TCUs. A turn which could be analyzed into several units on syntactic grounds was only treated as comprising more than one TCU if prosody (i.e., pauses, hesitations) and syntax both indicated a new planning process on the part of the speaker at the same position in the utterance. Each TCU is then assigned to one of the categories of ‘dialect’ (i.e., containing only dialect and ambivalent words) or ‘standard German’ (only standard German and ambivalent words) or it is designated as a ‘mixed/intermediate’ TCU. Finally, the proportion of TCUs in each of these three categories is calculated for each of the situations.

This procedure yields more detailed insights into the variation behavior of the individual by highlighting the degree and level of code mixing/switching within one situation on top of the differences between situations. Let us compare two conversations by adult speaker B2 that yielded very similar d-values. With the interlocutor from Northern Germany, B2 speaks at a dialectality value of 1.77; with the speaker from the Alemannic dialect region, he uses speech at a d-value of 1.76. In fact, however, an analysis of the proportions of TCUs in different varieties reveals that the proportion of clearly dialectal turns is much higher with the Austrian speaker from the Alemannic area than with the German interlocutor. With the German speaker, the relative number of mixed or intermediate turns is higher than with the Austrian-Alemannic partner (see Figure 2a). Nonetheless, the information which can be derived from both d-values and proportion of TCUs per variety complements each other: the fact that the d-values were similarly high in both situations but the proportion of mixed/intermediate turns is quite different suggests that even with the Northern German speaker, B2 generally makes only small concessions toward standard German within the mixed/intermediate turns.

Figure 2:

a. Proportions of TCUs: adults speaking about the same topic with four different interlocutors: (a) second language speaker, (b) German speaker, (c) speaker from another dialect region, (d) speaker from the same dialect region. b. Proportions of TCUs: children in five different communicative situations each: (e) memory standard, (f) memory dialect, (g) story standard, (h) story dialect, (i) pretend-play.

Similarly, we can gain more information from considering the proportion of TCUs per variety in our child data, e.g., by juxtaposing situation ‘Memory Standard’ for Tobias and ‘pretend play’ for Felix, which yielded very similar d-values (0.71 and 0.78, respectively). Whereas for both, the proportion of standard TCUs is similarly high, Felix uses a considerable amount of dialect TCUs whereas Tobias mixes more on the TCU-level and uses hardly any dialect TCUs. By way of the classification of TCUs we can therefore identify Tobias’s ‘technique’ of converging toward a slightly more dialectal mode of speaking as ‘mixing’ (i.e., TCU-internal) and Felix’s technique as ‘switching’ (i.e., between-TCUs) (see Figure 2b).

If we compare the individual repertoires suggested by the depiction of d-values on the one hand and by the proportion of TCUs calculated from the child data on the other, different conclusions about the children’s repertoires may be drawn. A look at Sascha’s d-values suggests that he does not master standard-near speech. In contrast, scrutinizing what he produces in terms of proportions of TCUs, we can conclude that he can definitely produce standard-near turns or TCUs. Felix’s d-value of approx. 0.5 would not normally be classified as ‘standard-near speech’ (Herrgen et al. 2001) either, but if we consult Felix’s TCUs, he clearly masters standard-near speech within turns. In contrast, Tobias’s monovarietal competence in standard German seems to be corroborated by the TCU data, although he, too, exhibits some proficiency in the local dialect as can be seen in several TCUs.

Another observation based on the classification of TCUs concerns the differences between children and adults in the speech levels employed. At least from this small set of data, it seems that adults make more use of mix/intermediate turns than children. This difference may, however, be an artifact resulting from the fact that children’s TCUs are on average much shorter than adults’ TCUs, and it cannot be determined at this point whether the ontogenetic development of language variety repertoires may in fact proceed from having two distinct varieties (standard German vs. dialect) to increasingly bridging the range between the two poles by utilizing forms ‘in-between’.

Clearly, then, the ‘proportions per variety’ method comes with problems and shortcomings, too, the most prominent issue being the process of categorizing and classifying words and TCUs as belonging to ‘dialect’, ‘standard language’ or a ‘mix/intermediate form’ between the two. From a linguistic point of view, one might consider stipulating more than three speech level categories or set the boundaries at different points in the ‘continuum’. Furthermore, the issues raised above with regard to the definition of ‘Austrian standard German’ are equally relevant with this procedure. Similarly, ‘local dialect’ is a challenging category if one is no longer interested in the (somewhat) clear and well-documented ‘base dialect’ spoken by the NORMs (non-mobile, old, rural males) but in the dialect spoken by the younger generation. These forms of speech typically involve some (frequently lexically constrained) dialect-leveling and have not been systematically documented for the area so far. Linguists have a way of resorting to the descriptions of the general characteristics of ‘Central-Bavarian dialect’, which may lead to disregarding local specificities, newer developments and – even more problematic – specific lexical diffusion patterns of phonological dialect-standard ‘rules’ (see Scheutz 1985: 244–252).

It needs to be mentioned that further methods exist, which would be able to circumvent some of the problems associated with calculating d-values and proportions of TCUs. Variable analyzis and its recent, statistically sophisticated implementations, such as cluster analysis and correspondence analysis, rely on large amounts of data and the selection of linguistic features, which are counted and analyzed for covariance patterns, using advanced multivariate statistical techniques (e.g., Ghyselen and de Vogelaer 2018; Kehrein 2012; Lanwer 2015). This procedure thus does not require the pre-conception of varieties but enables bottom-up detection of systematicity in variation. It also involves, however, methodological decisions regarding the (theory-driven) selection of variables to be considered and crucially, it requires large amounts of data, which, in practice, entails numerous obstacles regarding data collection (especially with regard to children), research setting, data storage, computational power, etc. Typically, this leads either to a reduction of data collected per individual or to a reduction of the number of participants being studied, which again impairs the ecological validity of the results.

5 Conclusion

In this contribution, we have presented results on intra-individual variation in adults and children in the Central-Bavarian dialect region. We have addressed questions as to the nature and development of intra-individual variation, while critically discussing the concepts and methods currently available to linguistic research in this context, as well as the epistemological consequences methodological decisions and presuppositions may have.

The linguistic categorization and classification process with regard to specific varieties is neither straightforward nor fully objective. Moreover, we may seriously question whether it gives us ‘the whole story’ (see also Ghyselen and de Vogelaer 2018). Particularly when we are interested in the social meanings of language and language varieties, linguistic analysis demands to be complemented by an emic perspective i.e., the perceptions and categorizations by the language users themselves. Accordingly, social and contextual factors can and should fruitfully inform discussions about intra-individual variation in specific contexts and about the human ability to vary speech in general. In this vein, we must not disregard the fact that dialect and standard language are imbued with socio-indexical meanings. They can also be utilized to organize communicative situations or to position oneself or others in an interaction by evoking associations with certain groups and attitudes toward them. And finally, evidence of the cognitive representations of these links between linguistic and social information is key: How are dialect-standard variants stored and processed? How is social information shared across or attached to variants in speakers’ minds? After all, it is the individual’s mind where links between knowledge of language and knowledge of the social world emerge right from the beginning of language socialization.

It thus seems necessary in order for linguists to fully capture intra-individual dialect-standard variation in Bavarian-speaking Austria to have and combine (a) more and more representative authentic speech data, with (b) epistemologically-aware, multi-method analyses of the data, and (c) insights into the perceptual and socio-cognitive representations of linguistic entities in speakers’ minds.