Children’s Linguistic Repertoires Across Dialect and Standard Speech: Mirroring Input or Co-constructing Sociolinguistic Identities?

ABSTRACT

The present study analyses 3- to 6-year-old children’s dialect-standard repertoires in an Austrian-Bavarian sociolinguistic setting and investigates how far individual repertoires can be explained by input and sociodemographic factors. Adults’ linguistic repertoires in the area typically comprise a certain spectrum on the dialect-standard continuum but individual acquisition processes have hardly been studied yet. We collected language data from 49 children in five different communicative interactions each and analyzed the repertoire each child exhibits. The majority of children could be shown to have a bi-varietal repertoire at their disposal, but there were substantial numbers of children who exhibited either standard-only or dialect-only repertoires. We then examined the relationships between a child’s repertoire and potentially relevant input and sociodemographic variables. While language variety use in the home and maternal education did not prove significant predictors of children’s repertoires, gender, age, location, bilingualism and frequency of being read to did.

Language acquisition not only implies the learning of sounds, words and sentence structures of one’s L1(s) but also entails the learning of different sociolinguistic variants of the language(s) and using them in adequate ways in appropriate situations. Depending on the sociolinguistic setting, children are presented with more or less numerous and more or less different varieties (i.e. styles, registers, dialects) of the language(s) surrounding them and they have to master the different form-meaning mappings, including their sociopragmatic functions and socio-indexical meanings. The main aim of this article is to lay out the dynamics of sociolinguistic and input constraints which influence the development of children’s dialect/standard repertoires. The area of investigation is located in the Bavarian-German speaking part of Austria.

Bi- and multidialectal language acquisition in children

As is widely acknowledged in linguistics, the boundary between dialects and languages is an arbitrary one determined by ideology and language policy as much as linguistic criteria (cf. the famous adage attributed to Max Weinreich: ‘A language is a dialect with an army and a navy’). Here, we will use the words ‘bilingual’ and “multilingual” to refer to children growing up learning a variety of German and some other non-German language(s), and “bidialectal” or “multidialectal” to refer to children growing up learning two or more varieties of German, while acknowledging that from a psycholinguistic point of view bi- or multidialectalism may not be fundamentally different from bi- or multilingualism (Chevrot & Ghimenton, 2018; Kirk et al., 2018; but cf. Schmidt, 2016; Melinger, 2018).

The issue of children’s abilities in handling different sociolinguistic varieties of the same language has received considerable attention internationally (see the volumes/issues by Chevrot & Foulkes, 2013; Gunther & Katerbow, 2017; Smith & Durham, 2019; for example). The dearth of current research in German-speaking areas, in contrast, seems astonishing given that German is regarded as a language with a particularly high extent and range of variation (Barbour & Stevenson, 1998). The acquisition of different varieties of German was researched in German-speaking Switzerland in the 1990s (cf. Häcki Buhofer & Burger, 1998; Häcki Buhofer et al., 1994; Schneider, 1998) and in Germany in the 1980s (cf. Mattheier, 1980; Scholten, 1988) and has recently been taken up again by Katerbow (2013). Regarding the Austrian context, until recently there was only one small case study by Moosmüller and Vollmann (1994) which specifically addresses this question and a study by Penzinger (1985) which focuses on parental input and mother-child interaction.

So far, the findings from international research on children’s sociolinguistic abilities – though extremely heterogeneous in setup and results – can be summarized as follows:

• (i)The acquisition of sociolinguistic abilities starts at an early age and in accordance with children’s communicative needs. It is an integral part of language acquisition.

From a usage-based and social-interactional perspective on language acquisition (e.g., Tomasello, 2003), acquiring different language varieties is inextricably linked to acquiring the social and pragmatic usage constraints attached to them. Adaptation of sociolinguistic variables has been shown to occur far earlier than previously envisaged (e.g., by Labov in his acquisition model of standard English, Labov, 1964), although the complexity of the conditioning of the variable in question seems to affect exactly when and how acquisition sets in (Smith & Durham, 2019, pp. 17–23, 190–193). Research suggests that children are capable of selecting some sociolinguistic variants according to the type of interaction and/or according to their interlocutor from as early as 3 or 4 years (cf. Katerbow, 2013; Patterson, 1992; Roberts, 1994; Smith et al., 2007). Artificial language learning experiments have demonstrated that children tend to regularize input that varies unsystematically (Hudson Kam & Newport, 2005) but that they are able to learn the probabilistic regularities of socially conditioned language variation (Samara et al., 2017). Thus, acquiring patterns of variation is an integral and manageable part of language acquisition, which starts early and occurs (at least in part) simultaneously with other aspects of language acquisition (Johnson & White, 2020). At the same time, while linguistic and socio-indexical knowledge are likely to be acquired in parallel, the development of sociolinguistic competence is a prolonged process that takes years (Dossey et al., 2020).

• (ii)The acquisition of sociolinguistic variation is subject to a child’s (changing) social network and the (concomitantly changing) input the child receives. Parental input is the major source in the beginning, but the peer group seems to gain influence earlier than previously thought.

Evidence regarding the development of children’s sociolinguistic usage patterns as they grow older suggests the following non-linear trajectory: For pre-school children, where age-graded changes have at all been observed, there seem to be tendencies toward more frequent nonstandard (i.e. vernacular) production (Nardy, 2008; Smith et al., 2007) during development. During the first years of school, several studies note an increase in the production of standard variants (Chevrot et al., 2000; Janneke & Wolfram, 2010; Patterson, 1992; Romaine, 1984), whereas the frequency of vernacular variants seems to peak in the early to mid-teens (Janneke & Wolfram, 2010; Kerswill & Williams, 2000; Portz, 1982). It has to be added, however, that there are hardly any longitudinal panel studies to verify individual real-time trajectories or to follow individuals’ development beyond adolescence.

Most authors do not posit a direct causal relationship between biological age and a child’s changing sociolinguistic repertoire but they usually suggest an indirect link via new linguistic experiences, changing social networks and, concomitantly, the changing input the child receives. The first few months and years of primary socialization and thus of first language acquisition are typically characterized by the parents and their linguistic input. For example, the increase in nonstandard features in young children’s speech as they grow older can be explained by a change in maternal input since mothers typically reduce the frequency of nonstandard variants when speaking to very young children, at least of the more salient variants (Smith & Durham, 2019), and they gradually increase frequency of vernacular variants as the children grow older (de Houwer, 2003; Foulkes et al., 2005; Smith et al., 2007). While parents may thus strive to scaffold their very young children’s language acquisition by using more standard-near variants, they gradually approach the patterns of inter-adult conversations, which are typically more deviant from standard speech, with their older children. Sooner or later, however, a child is confronted with a larger variety of caregivers and other interlocutors and thus with a wider range of linguistic varieties. Children increasingly extend their radius of communication to their peers, the most pervasive change presumably occurring with the entrance to (pre-)school. Younger children accordingly exhibit linguistic variants very close to those of their parents (Kerswill & Williams, 2000; Smith et al., 2007), whereas older children increasingly produce variants comparable to those of their peer group (cf. e.g., Foulkes & Docherty, 2006; Kerswill & Williams, 2000; Labov, 1964, 2001; Macha, 1993). In fact, Cornips (2020) shows that the entrance to preschool may even lead to a (perhaps temporary) shift from the regional language variety to the majority language as children’s home language, even when both parents speak the regional variety. But media also play a part in children’s expansion of their L1 repertoires (Häcki Buhofer et al., 1994), although their exact impact on language acquisition depends on child attributes, content, stimulus characteristics and environmental context (Linebarger & Vaala, 2010; Richert et al., 2011). However, children are not strictly “monostylistic” speakers (changing from the parental system to the peer group system) but the systems provided by parents, peer-group and perhaps other influential input sources may co-exist in the child’s mind (cf. Nardy et al., 2013).

As we know from first and second language acquisition research, however, language development is not directly governed by the input the learner receives but is mediated by learner-internal factors that determine which aspects of the input become intake (Wijnen, 2000), which effectively shape the learner system. Factors such as the perceptual salience of features and the learner’s attention on the input are known to affect the integration of new language structures into the language learner’s system (e.g., Divjak, 2019; Truscott & Sharwood Smith, 2011). Moreover, the acquisition of sociolinguistic variants also depends on the social understanding children bring along, including knowledge about the evaluative patterns with regard to language variation in the speech community in which they are growing up. Children are sensitive to asymmetries in the power dynamics of the languages and language varieties surrounding them, which impact linguistic practices, e.g., in preschool (Cornips, 2020). In fact, Sumner et al. (2013) suggest that the strength of (exemplar-based) linguistic representations is not only determined by the frequency or typicality of variants in the input but also by their social weight, i.e. depending on whether the input stems from reliable, socially prestigious, idealized sources or not, it receives more or less attention by the listener and thus leads to more or less weighted encodings.

Several studies have provided evidence of an early evaluative preference for members of children’s own native language group (Kinzler et al., 2007; Paquette-Smith et al., 2019), but the findings on young children’s understanding of the status and prestige of the language varieties in bi-dialectal communities are mixed. It has been shown that even children as young as three years have developed certain attitudes and preferences toward language varieties which correspond to those of their surroundings (Rosenthal, 1974, 1977). Some studies, however, have not found significant and consistent preferences or evaluations in preschool children (Häcki Buhofer et al., 1994; Kinzler & DeJesus, 2013). McCullough et al. (2019) suggest that adult-like status ratings may be in place by ages 4–5 years while the development of solidarity ratings may be protracted because they are more variable and hinge more on interpersonal experience. Consistent with findings from international studies (Cremona & Bates, 1977; Day, 1980; Giles et al., 1983; Kinzler & DeJesus, 2013), Kaiser & Kasberger (forthc.) show the development of an attitudinal preference pattern in Austrian children in the first two years of primary school with a clear tendency in favor of the standard variety between ages 7 and 9. The authors ‒ as others before ‒ propose that this is to a large extent due to the influence of schooling.

One might argue that the truly stylistic use of language varieties, e.g., code-switching as a means of contextualization (Blom & Gumperz, 1972; Gumperz, 1982) or “speaker design” (Schilling-Estes, 2002), typically capitalizes on the socio-indexical value, i.e. the common associations of language varieties with certain speakers/situations as well as their evaluation in society, and thus presupposes some (not necessarily conscious or explicit) understanding of them. Having said that, mechanisms of linguistic mimesis and convergence to one’s addressee(s) may work differently (i.e., to some extent automatically, Markham, 1997; Gambi & Pickering, 2013; Foulkes & Hay, 2015, p. 298) and may be possible at an earlier stage in development than the truly stylistic use of varieties (cf. Smith & Durham, 2019, pp. 15–16). The modeling of patterns of variation in the input and interpersonal adaptation in communication (which is addressed theoretically by a range of different models and approaches such as Communication Accommodation Theory, see e.g., Giles & Ogay, 2013, or Interactive Alignment, see Pickering & Garrod, 2013), may thus pave the way to acquiring and mastering linguistic variation and its socio-indexical meaning, which can then be employed more and more independently and strategically. Moreover, these processes may play a crucial role in language acquisition – with regard to language varieties and in general – since they involve both comprehension and production on the child’s part and thus possibly enhance depth of processing and the integration of input into the learner system.

iii. Individual variation in the acquisition of sociolinguistic variation is immense, indicating a complex interplay of individual, linguistic and social variables.

Several studies (e.g., Van Hofwegen & Wolfram, 2010; Katerbow, 2013) observed that the acquisition of variation is a highly individual process and seems to be dependent on a wide range of factors. Among the few individual variables which have been studied systematically are the children’s socioeconomic background (socioeconomic status = SES) and gender.

Family SES seems to correlate with children’s production of sociolinguistic variants increasingly from age 3 onwards, with children from high-SES backgrounds producing more standard variants as compared to sociolectal variants (e.g., Chevrot et al., 2011; Díaz-Campos, 2005; Van Hofwegen & Wolfram, 2010; Nardy, 2008). Notably, high-SES children also seem to show an earlier preference for standard forms than low-SES children (Kaiser & Kasberger, forthc.; Barbu et al., 2013).

Findings on gender differences among children regarding the production of sociolinguistic variants are mixed, with some studies yielding no gender differences and others pointing in contradictory directions (for a brief review see Nardy et al., 2013). Some of the findings of the studies which do show a gender effect in the acquisition of sociolinguistic variation parallel the general tendencies observed in adults (Labov’s Principle I in Labov, 1990; Cheshire, 2002, pp. 425–26; Romaine, 2003; Brouwer, 2011), at least for children from age 3;6 onwards (Foulkes & Docherty, 2006, pp. 423–24), in that girls used fewer nonstandard features than boys. Barbu et al.’s (2014) findings suggest that girls and boys may be equally sensitive to the social prestige of speech varieties but that males and females may orient toward different social groups and therefore different dimensions of prestige (cf. “overt” and “covert” prestige, Trudgill 1972): They found that boys tend to use regional variants in adapting to local interlocutors more than girls; masculinity thus seems to be associated with the adherence to local/regional speech norms.

A different line of argument may consider the differences in the input girls and boys receive and in the role models they are offered. If children orient to same-sex adults in their linguistic behaviors, existing gender-differences will be transmitted (Ladegaard & Bleses, 2003). In addition, variation in standard vs. nonstandard input with respect to a child’s gender has been observed by Foulkes and Docherty (2006, p. 421), Lenz (2003) and by Kasberger and Gaisbauer (2020). In these studies, adults used more standard variants with girls than with boys. Thus, both input and social learning may collude in the emergence of gender differences in language acquisition.

Background to the study: dialect-standard variation in Austria

Austria is a German-speaking country whose pattern of dialect-standard usage has usually been conceptualized as a “dialect-standard continuum”. The term implies that there is no clear-cut distinction between two varieties in use but that a range of speech forms “in-between” is composed of variable proportions of standard and dialect¹ forms (which are subject to complex co-occurrence restrictions, see Felix & Kühl, 1982; Auer, 1986 for Bavarian speakers in Germany; Scheutz, 1985 for Bavarian speakers in Austria),² as opposed to the “diglossic” situation in German-speaking Switzerland. Moreover, there are “intermediate” phonological and lexical forms which are neither part of a local base dialect nor of the Austrian standard variety. This intermediate range between the poles of “standard language” and “dialect” is typically called “Umgangssprache” (or less frequently “Regiolekt”).

The following examples should serve to illustrate possible realizations within the dialect-standard continuum in the Central-Bavarian area:

Heute Nacht hat es viel geschneit. (“Last night it snowed a lot.”) (Austrian) Standard German (ASG): ‘hɔɪ̯tə naxt hat ɛs fi:l gə’ʃnaɪ̯t Umgangssprache (several possibilities): ‘hɔɪ̯t nɔxt hɔts fi:l kʃnaɪ̯t ‘haɪ̯t nɔxt hɔts fy: kʃnaɪ̯t Bavarian base dialect variety (BD): ‘haɪ̯̃d nɔxd hɔ:ds fi: kʃni:m

These illustrations should not be taken to mean that only these realizations are possible nor that there are hard and fast demarcations between the varieties. For example, the omission of final schwas (as in hɔɪ̯t) may be considered acceptable in a “usage standard” (“Gebrauchsstandard”) and/or in allegro standard speech. In the “middle” range of the “Umgangssprache” in particular, further intermediate forms or different combinations of features are conceivable. But the examples serve well to illustrate the considerable differences between Standard German and the local base dialects in the area and possible “Umgangssprache” realizations. Typical features of Central-Bavarian dialects in the examples above include A-raising (ASG hat → BD hɔ:d̥), L-vocalization (ASG fi:l → BD fi: or fy:) and diphthongs different from the standard language (ASG hɔɪ̯tə → BD haɪ̯̃d̥). These are typical and highly frequent phonological features of Austrian-Bavarian dialect; there are further salient differences in the realm of diphthongs (/ai/-diphthongs and diphthongization, e.g., ASG bʀu:dɐ → BD bʀuɐ̯dɐ) and monophthongs (different E-qualities) and less salient differences in consonants (lenition of voiceless plosives). Furthermore, there are typical morphological features in Bavarian German dialect, most notably the personal pronoun for 2^nd person plural: /e:s/(nom.), /ɛŋ/ or /ɛŋk/(dat./acc.) (instead of /i:ɐ/ <ihr> and /ɔɪ̯ç/ <euch> in ASG) and the resulting inflection morpheme for 2^nd person plural {-s} or {-ts} (e.g., /e:s hɔpts/ as opposed to /i:ɐ ha:pt/ in ASG). Morphological differences also include differences in participle forms (strong forms vs. weak forms) as in BD /kʃni:m/ vs. /gəˈʃnaɪ̯t/  (weak form) in the example above. Also, lexical differences are possible in everyday language.³

The standard variety is Austrian Standard German, but a majority of the population (according to self-report) also speaks a local dialect (Steinegger, 1998), which for all but the westernmost part of the country is a Bavarian dialect variety. In a survey by Ender et al. (2009), 72% of Austrians from the Bavarian region reported having good or fairly good proficiency in both dialect and Standard German. Situational language choice typically depends on – but is by no means fully determined by – the formality of the situation and on the (variety spoken by the) interlocutor (see also Ender & Kaiser, 2014; Wiesinger, 1992) while also being subject to inter-individual variation. In interaction, processes of accommodation (of upward and/or downward convergence or divergence) can be observed (Kaiser & Ender, 2013), e.g., a shift toward “Umgangssprache” and “Hochdeutsch” is to be expected when speaking with people from other (German- or non-German-speaking) countries (Ender et al., 2009, p. 286).

Language variety acquisition and use by children in Austria has not yet received much research attention. Only recently did Kaiser and Kasberger (2018) examine the development of discrimination abilities in Austrian children. They found that 5-year-old children are able to match speakers of the same variety on a perceptual level, i.e. when the same sentences are spoken. But it is only considerably later (from the age of eight years onwards) that children are able to resort to more abstract categories of language variety, i.e. to match speakers of the same variety when they say different sentences. Furthermore, children seem to acquire explicit metalinguistic terminology and knowledge about dialect and standard language primarily in tandem with the acquisition of literacy (Kasberger & Kaiser, 2019). Kaiser and Kasberger (2018) suggest that the acquisition of discrimination abilities in Austrian children might be somewhat protracted because of the complex sociolinguistic setting, which does not provide clear-cut distinctions between dialect and standard language in everyday communication, neither in linguistic terms nor in terms of the sociolinguistic situations linked to the varieties in question.

The socio-indexical attributions which have frequently been described in many different constellations of co-existing standard vs. nonstandard or dominant vs. non-dominant varieties around the world have also been observed in Austria: Dialects are associated with preferred characteristics on the “social attractiveness” dimension (likability, humor, sociability, honesty …) while also being linked to less favorable associations in regard to aspects of “competence” (intelligence, education, politeness, seriousness) (Soukup, 2009, p. 169) and, interestingly, standard speakers are judged as being more attractive and better dressed (Bellamy, 2012, p. 224). These typical “indexical fields” (Eckert, 2008) are invoked by competent Austrian speakers in practices of code-switching and code-shifting as described by Kaiser (2006) and Soukup (2009, 2013). Kaiser & Kasberger (forthcoming) studied Austrian children’s sociolinguistic preferences. In an adapted, child-friendly “matched-guise” task, children were asked to choose between two male and two female doctors after they had heard a pair of each speaking in either Standard German or the local dialect. Preschool children (ages 3 to 6) did not show a distinct preference toward either of the speakers overall. However, from age 7 or 8, a clear preference for the standard-speaking doctors develops, which however seems to be dampened at age 10. In this study, a closer look at the SES-subgroups and at the gender variable proved relevant (see above). Girls showed a slightly but consistently stronger preference for the standard speaker than boys.

Several decades ago, the discussion of so-called “language barriers”, which were believed to preclude academic and professional success for (child) dialect-speakers, led to a new trend among dialect-speaking parents toward raising one’s children in a variety closer to standard speech, i.e. typically in some intermediate “Umgangssprache”, in order to supposedly open up better academic and professional opportunities (cf. Mattheier, 1980; Penzinger, 1985, p. 14; Macha, 1993). In a non-representative survey by Ender et al. (2009), the majority (75%) of the Austro-Bavarian participants “(almost) always” or “frequently” use dialect with their Austrian colleagues from work but considerably fewer opt for dialect use with their children (49%). 45% of the participants report using “Umgangssprache” with their children (if they had any). In this survey, only a tiny 2% reported using Standard German with their children. Still, the amount of Standard German input must not be underestimated. Practically all children’s books, for example, are written (and read out loud) in Standard German⁴ and almost all children’s TV programmes are in the standard language, too. So it can safely be said that all Austrian children come into contact with the standard language receptively even before entering kindergarten or school.

It is one of the tasks of the educational system to guide children to the standard language (Bundesministerium für Bildung, Wissenschaft und Forschung. Bundesrecht konsolidiert, 2018; Charlotte-Bühler-Institut, 2009), particularly in their own linguistic productions. Furthermore, the proportion of bilingual children (with migrant background) is growing and is typically higher in urban areas, which leads to a tendency for teachers to use more Standard German in order to promote these children’s acquisition of the standard language (for preschool teachers, see Kroisenbrunner, 2014, pp. 97–104). The linguistic reality at Austrian schools regarding language variety use seems to be quite diverse between and within regions, institutions, school subjects and individuals and it includes many different forms of speech (cf. de Cillia/Ransmayr, 2019; Vergeiner et al., 2019). Unfortunately, however, there are hardly any empirical data on actual language use in Austrian homes to date.

At present, there are no comprehensive studies on the sociolinguistic repertoires emerging in Austrian children. The present study sets out to tap into Austrian preschool children’s expressive repertoires by observing the children’s verbal behavior in different settings. By comparing children of different ages, i.e. using a cross-sectional apparent-time design, we can cautiously infer hypotheses about what the real-time development of children in this area may look like. Furthermore, we are interested in the variables which shape the individual emerging repertoires. We thus examine the relationship between a selection of sociodemographic and input variables and children’s repertoires in order to further our understanding of how social and cognitive-psychological mechanisms may intertwine in the acquisition of sociolinguistic competence.

The current study

Participants

We collected data from 35 monolingual and 20 multilingual children growing up in two different locations (one rural; one urban in a town in the greater Salzburg metropolitan area), both in the province of Salzburg, Austria, which belongs to the Central-Bavarian dialect region described above. The children’s ages ranged from 3;4 (40 months) to 6;4 (76 months) (mean 60 months, standard deviation 10.43). 49 children (33 monolinguals and 16 bilinguals) provided sufficient data for analysis (see below). The majority of the bilingual children were simultaneous bilinguals (n = 11), i.e. with one parent being an L1 speaker of German. Children’s (additional) first languages besides German were Bosnian/Croatian/Serbian (n = 10), Turkish (n = 2), Spanish (n = 1), Greek (n = 1) and Austrian Sign Language (n = 1).⁵ All of these children had substantial German input at home, from parents and/or siblings. For a more detailed analysis of the multilingual children’s use of varieties see Kaiser (2019).

The most important sociodemographic and input data are given in Table 1.

Table 1. Participants

		Location		Gender		Main variety of German at home^a			Mono-/Bilingualism		Maternal education (SES)^b
		Rural	Urban	Male	Female	Dialect	“Umgangssprache” or Standard German	Various	Mono-lingual	Bilingual	Low	High
Age group	3 yrs. (n = 7)	5	2	2	5	2	4	1	5	2	5	2
	4 yrs. (n = 16)	6	10	8	8	6	4	6	11	5	8	7
	5 yrs. (n = 19)	7	12	12	7	5	8	5	13	6	15	4
	6 yrs. (n = 7)	1	6	4	3	3	3	1	4	3	5	2
Location	Rural			8	11	5	7	7	17	2	10	9
	Urban			18	12	11	12	6	16	14	23	6
Gender	Male					11	10	4	16	10	19	7
	Female					5	9	9	17	6	14	8
Main variety ofGerman at home	Dialect								13	3	12	4
	“Umgangssprache” or Standard German								12	7	13	6
	Various								8	5	7	5
Mono-/Bilingualism	Monolingual										20	13
	Bilingual										13	2
Total (n = 49)		19 (38.8%)	30 (61.2%)	26 (53.1%)	23 (46.9%)	16 (32.7%)	19 (38.8%)	13 (26.5%)	33 (67.3%)	16 (32.7%)	33 (67.3%)	15 (30.6%)

^aData were missing from one child.

^bData were missing from one child.

Data collection and coding

Data were collected in two different preschools. We videotaped the children in five different semi-experimental settings so as to provide the best circumstances for eliciting children’s full repertoires:

• Two to three children played two rounds of Memory, each with a different experimenter. One of the experimenters exclusively spoke Standard German, the other experimenter spoke exclusively in the local dialect. The experimenters animated conversation by asking questions related to the pictures on the Memory cards.

• Each child was asked twice to re-tell a story they had previously heard in the other variety: The experimenter narrated a story in the local dialect (accompanied by pictures to verify children’s comprehension). After having heard the story in dialect, the child was asked by a standard-speaking hand-puppet, who “hadn’t understood”, to re-tell the story and show her the pictures. The hand puppet (played by the experimenter) asked many questions about the details of the story. In the second round, the child first heard a story in Standard German and was then asked to re-tell it to a dialect-speaking puppet (played by a different experimenter).

• In groups of three to four, the children played grocer. This pretend-play was prompted by a play shop arranged in the middle of the room with numerous items that could be bought and a till, and it was initiated by the experimenter’s request to buy a couple of items for her which were shown on a picture shopping list. The children were also given some play money to “pay” for their purchases. The children decided for themselves which roles they wanted to take on and organized the play themselves with only occasional help by the experimenter if needed.

The conversations were transcribed in CLAN with basic discourse-analytic annotation, and each child utterance was additionally transcribed in wide phonetic transcription. Each word within a child utterance was then coded as either “standard language”, “dialect”, “mixture” or “Umgangssprache”, if it was unambiguously possible to do so; the rest was coded as “ambiguous”. This coding was based on the presence/absence of features of Central Bavarian dialects described in the literature (see Wiesinger, 1983, pp. 836–42; Zehetner, 1985; Mauser, 2009, pp. 62–66; Scheutz, 2009, pp. 21–23, 41–44).

Children’s utterances were then segmented into TCUs (“turn-constructional units”, Selting, 2000). A TCU is “a potentially complete turn” (Selting, 2000, p. 480) in a conversation and is as such an interactionally relevant unit. It is a unit confined by possible points of turn-taking, which are mainly signaled by syntactic and prosodic cues. We decided to work with TCUs rather than words or other units on several grounds: First, turns are the “natural” units of (spontaneous) face-to-face conversation and they can most unambiguously be determined in this kind of communication. But more importantly, since we largely exploit mechanisms of linguistic accommodation (cf. “Speech Accommodation Theory”, e.g., in Giles et al., 1991, later “Communication Accommodation Theory”, e.g., Giles & Ogay, 2013), focusing on turns seems adequate from a psycholinguistic point of view, as each turn by the conversation partner could potentially trigger (new) accommodation processes. In our case, TCUs are frequently identical to turns because the children’s turns are typically very short and only consist of one turn-constructional unit. In the – by far – rarer cases of longer turns, e.g., comprising compounded sentences, we aimed to determine whether these more complex constructions involved additional psycholinguistic planning processes. We therefore decided to segment the turns into TCUs by analyzing the interplay between syntax and prosody, i.e. only when syntax and prosody (i.e. pauses, hesitations) converged, the turn was segmented into several TCUs.

On the basis of the word-level codings, each child TCU was then coded as either “standard language” (i.e., only including standard (and ambiguous) words), “dialect” (i.e., only dialect (and ambiguous) words) or “mixed/intermediate” (mixtures of standard, dialect and/or “Umgangssprache” words). “Ambiguous” turns (i.e., realizations which would have been homophonous in dialect and standard language) were not considered in the following calculations. Children were only included in the following analyses if they provided at least 10 TCUs in the respective communicative situation.

Analyses

In order to analyze children’s varietal behavior across conversations, the proportion of turn-constructional units in the different varieties was calculated for each child in each of the five situations (but only if the child contributed at least 10 TCUs to the interaction).

In the next step, i.e. for determining children’s varietal repertoires, speech levels were calculated for each conversation based on the proportions of dialect, standard and “intermediate” TCUs. A “dialectal” speech level comprises at least 50% of dialectal TCUs and fewer than 30% of standard TCUs. A “standard-near” speech level would be the inverse. When neither dialectal nor standard TCUs reached the 50% level, the TCU was designated as “intermediate”. Types of repertoires were then gauged for the 49 children who provided sufficient data (i.e., at least 10 TCUs) in at least two conversations with different interlocutors. Even though our methodology aimed at capturing as wide a dialect-standard range as possible, it has to be kept in mind that only observed behaviors are being analyzed here and that we cannot be certain of having captured children’s varietal abilities in their entirety.

Results

Variation across conversations and children’s repertoires

The comparison of mean percentages of TCUs per variety (see Table 2) confirms that children overall do vary their speech between the conversational situations. For example, on average 58% of children’s TCUs were in standard German when they were re-telling a story to the standard-speaking hand-puppet while only 32% of children’s TCUs were in standard language when they were re-telling a story to the dialect-speaking hand-puppet. Overall then, children adapt the degree of “dialectality” to the speech of the interlocutor but at the same time they generally use more dialectal speech in play than in story-telling. Furthermore, children employ standard speech in pretend-play to index a role they are taking on. Approximately two thirds of children’s TCUs were in standard German when they were playing grocer, even though the majority of children does not come from standard-speaking homes.⁶ At the same time, it is clear that many children do not fully switch from one variety to the other between conversations. The degree of linguistic convergence differs between children, but almost all children show some accommodation to the conversational partner. While for some this means only inserting certain dialectal or standard particles (such as “yes” and “no”), others almost fully converge to their interlocutor.

Table 2. Mean proportions of TCUs per variety in each situation

Situation	Mean percentage of dialect TCUs (standard deviation)^a	Mean percentage of standard TCUs (SD)	Mean percentage of mixed/intermediate TCUs (SD)	Total TCUs (including ambiguous TCUs)
‘Memory‘ with standard German speaker	48.2% (26.4)	38.4% (26.2)	13.4% (11.1)	587
‘Memory‘ with dialect speaker	60.1% (26.4)	27.5% (26.1)	12.4% (7.2)	925
Story-retelling to standard German speaker	28.2% (24.3)	57.9% (29.2)	14.0% (12.1)	697
Story-retelling to dialect speaker	46.8% (23.8)	32.2% (28.8)	20.9% (15.0)	381
Pretend play	22.1% (22.4)	66.3% (28.8)	11.7% (9.1)	1357

^aOnly includes children with at least 10 TCUs in the respective conversation. Proportions were calculated for each child individually (without ambivalent TCUs) and then averaged across all children. In this way, differences in the number of individual TCUs per conversation do not skew the overall pattern.

Table 3 gives a summary of all the children’s observable repertoire types, i.e. of the dominant speech levels employed by them in the conversations. Only one child produced neither a clear “dialectal” nor a “standard-near” speech level in any of the conversations and was thus classified as having only an “intermediate” repertoire.⁷ About half of the children have both end poles of the dialect/standard continuum at their disposal: They are able to produce consistently standard-near speech in at least one conversation and dialectal speech in at least one other conversation and might thus be labeled as “bidialectals”. In contrast, about a quarter of the children only employ standard-near speech levels and about a fifth of them seem to master only dialectal speech.

Table 3. Observable dialect-standard repertoires

Observable type of repertoire
Both standard-near and dialectal (bi-varietal)	Dialectal only	Standard-near only	Intermediate only	Total
25 (51.0%)	9 (18.3%)	14 (28.5%)	1 (2.0%)	49 (100%)

Predictors of children’s repertoires

Children obviously develop different dialect-standard repertoires. Whereas about half of them are bidialectal, the other half exhibit monolectal repertoires which comprise only one of the end-poles of the continuum. An exclusively “intermediate” repertoire seems to be the exception, however; children typically master at least one of the varieties of standard or dialect speech. In the following, we aim to get an understanding of the variables that shape the individual repertoires. Background variables considered were: age, maternal education (2 levels), bi-/monolingualism (dichotomous), main variety of German at home (3 categories: dialect, “Umgangssprache”/Standard German and “various”), location (rural/urban), gender and frequency of being read to (2 levels). These variables were chosen on theoretical grounds, as they are known to be the most influential sociodemographic factors in first language acquisition (age, gender, maternal education as a proxy for SES) or as they include relevant information about the child’s language (variety) input (bilingualism, main variety of German at home, rural/urban location and frequency of being read to). Information about these variables was collected in an extensive parental questionnaire. For statistical reasons, maternal education had to be coded as a simple two-level variable, “low” level comprising mothers without a secondary school leaving qualification (i.e. apprenticeship or only obligatory schooling), and “high” level comprising mothers with a secondary school leaving qualification or tertiary education. Similarly, frequency of being read to is coded as a two-level variable, “low” referring to a frequency of 3 times per week at most, “high” referring to daily reading.⁸ This reduction to two levels, again, was due to problems of distribution/sample size.

The main variety of German spoken in the home was determined by aggregating the reported frequencies of use in the parental questionnaire. Parents had been asked to indicate the frequency of use from 0 (never) to 4 (very often) for each of the varieties of dialect, “Umgangssprache” and Standard German for each parent (or other caretaker) and the child’s siblings. The variety that reached the highest overall frequencies was identified as the “main variety”. If two (or three) varieties reached the same frequencies, the fourth category was applicable, i.e. “various”.

Figure 1 gives an overview over bivariate relationships between different sociodemographic or input variables and the outcome (main types of repertoire, i.e. excluding “intermediate-only”).

Figure 1. Bivariate relationships between background variables and repertoire.

To verify the contribution of each predictor, multinomial logistic regression models for the main types of repertoires were calculated with the same background variables (age in months was added as a covariate). Due to our small sample size, no interactions were built into the models. The addition of the predictors to a model that contained only the intercept significantly improved the fit between model and data, χ² (16, N = 46) = 33.252, Nagelkerke R² = .595, p < .01). As shown in Table 4, significant unique contributions were made by location (rural/urban), gender, bi-/monolingualism, frequency of being read to and age, not, however, by maternal education and main variety at home.

Table 4. Predictors´ unique contributions in the multinomial logistic regression (full model)

Predictor	Likelihood-ratio tests
	χ^2	df	p
Constant	.000	0	.
Location (rural/urban)	8.465	2	.015*
Gender	7.573	2	.023*
Bi-/monolingualism	8.521	2	.014*
Maternal education	1.189	2	.552
Frequency of being read to	8.248	2	.016*
Main variety at home	1.923	4	.750
Age	6.038	2	.049*

χ² = Amount by which −2 log likelihood increases when predictor is removed from the full model.

We will now consider each of the predictors emerging as significant in the multinomial logistic regression models and describe the relationship between predictors and outcome variable as it can be observed in our data. In addition, we will analyze descriptively relevant interactions between these variables.

Bivariate relationships

As we can see in Figure 1, the overall tendency is for children to develop a bi-varietal repertoire. This notwithstanding, more children from a rural location have a dialect-only repertoire relative to children from urban areas and more children from the urban location exhibit a standard-only repertoire. Furthermore, female participants are more inclined to develop a standard-only repertoire whereas more males than females show a dialect-only repertoire. Bilinguals are the group with the weakest inclination to bi-varietal language use and a strong tendency to develop a standard-only repertoire. Far fewer monolinguals employ a standard-only repertoire. With increasing age, the probability of employing standard-near speech only seems to increase. This is mainly reflected by a decrease in dialect-only repertoires in the older age groups and an increase in standard-only language use, while the relative numbers of children with bi-varietal competence remain stable. Furthermore, children who are read to only rarely more often exhibit a dialect-only repertoire and less often a bi-varietal repertoire than their peers who are read to on a daily basis. There are only small differences between the two SES-groups (indexed by maternal education as a proxy) and this variable does not come out as a significant predictor in the regression model. Neither does the variable “main variety of German at home”; even though the descriptive figure may suggest a relationship between this variable and the repertoire developed, the multinomial regression model shows that it is not a predictive factor once other variables are considered.

Location and gender

Overall, most children develop a bi-varietal repertoire. Beyond that, however, location and gender proved to be significant predictors of the type of repertoire developed in our regression models. On closer inspection, these two variables seem to partly interact: Figure 2 shows that rural children more often develop a dialect-only repertoire, especially boys, and that urban children, particularly girls, virtually never employ a dialect-only repertoire. In contrast, developing a standard-only repertoire is more likely in urban areas, especially with girls. Considering gender, it seems that in all locations boys generally tend more strongly toward developing a dialect-only repertoire than girls and girls have a greater inclination toward standard-only repertoires.

Figure 2. Repertoire by location and gender.

Frequency of being read to and location

The question of how frequently children are read to also seems to play a part in the development of individual dialect-standard repertoires, but it seems to be particularly relevant in the rural setting (see Figure 3). Whereas for children in the city the relationship with reading frequency is weak, for children in rural areas it can be substantial: Frequently being read to increases the probability of acquiring and using standard speech and therefore decreases the probability of having a dialect-only repertoire.

Figure 3. Repertoire by location and frequency of being read to.

Age, gender and location

Generally speaking, as children get older, their usage of standard-speech increases. On top of that, gender and location seem to have different shaping effects on children’s repertoires. For urban children and for rural girls, the growing attraction to standard speech results in increasing proportions of standard-only repertoires. While the proportion of dialect-only repertoires remains relatively stable among rural children of different age groups, it goes down for urban boys. Notably, no rural boys exhibit a standard-only repertoire and no urban girls show a dialect-only repertoire. Figure 4 gives an overview of the different quasi-longitudinal trajectories for girls and boys in the urban and rural settings, based on the observed data.

Figure 4. Repertoire by age, gender and location.

Discussion and conclusion

Our study of children’s language variation behaviors is obviously limited with respect to the number of participants and the transferability of findings to other sociolinguistic settings. Due to the small sample size, all conclusions must be drawn with caution and should be considered as hypotheses in need of further verification by future research. Nonetheless, the study provides support of the central observations gathered from previous research. First, it confirms that the acquisition of sociolinguistic competence starts at an early age. Three- to six-year-old Austrian children already show considerable intra-individual variation in their choice of language variety in accordance with the communicative situation they are engaged in (see Table 2). However, children differ in the kinds of dialect-standard repertoires they develop. In the present study, about half of the children exhibit a bi-varietal repertoire, which comprises both dialect and standard-near speech varieties. They are able to produce consistently standard-near speech in at least one conversation and dialectal speech in at least one other conversation. About a quarter of the children only employs standard-near speech levels and about a fifth of them seems to master only dialectal speech.

We attempted to get more insight into the variables contributing to the development of children’s individual language variety repertoires. Multinomial regression analysis revealed that the most significant factors are a child’s age, gender, mono- vs. bilingualism and the frequency of being read to as well as the location in which the children live. In the regression models, maternal education and the variety of German mainly used at home do not significantly predict the type of repertoire the child develops. Again, due to the limits of sample size, the findings have to be interpreted with caution, but they can plausibly be integrated into current theory.

The first conclusion which can be drawn seems to be that from the age of three years on, children’s language variety repertoires are not or are no longer a mere reflection of the parental input they receive, thus corroborating observation ii, which posited the reflection of children’s changing sociolinguistic experiences in their own acquisition of sociolinguistic variation. While for babies and toddlers parental input has clearly been demonstrated to be the driving force of language development (Hart & Risley, 1995; Hoff, 2006; Smith et al., 2007 and others), other factors seem to come into play from an early age on. These factors may exert their influence in different but interrelated ways: They may shape and (re-)define the overall input a child receives and – via (socio)cognitive processes which effectively constrain what is integrated into the learner system – they may relate to the social (and therefore sociolinguistic) identity the child starts to co-construct for him-/herself in relation to parents, peer-group, teachers and other important role models.

Children’s communicative radius expands as they grow older: They go to kindergarten, meet peers, and kindergarten teachers and gradually experience a wider range of different social encounters, which entail different patterns of language use with new communicative partners. Standard German input especially from kindergarten teachers and from (a child’s increasing use of) the media can be hypothesized to be decisive in this regard. Additionally, in the given sociolinguistic context, older children may increasingly overhear intra-individual variation, e.g., from their parents in interaction with different interlocutors. Variation in the input therefore not only increases because of growing inter-individual variation due to a wider range of communicative partners a child meets but also because of intra-individual variation in the communicative partners he/she observes and learns from. It has to be kept in mind, however, that results may be different for 3- and 4-year-olds who do not yet attend preschool (preschool attendance is obligatory for 5-year-olds in Austria) and who therefore have not yet experienced such a far-reaching extension of their social-interactional field.

Age and the concomitant changes in linguistic experiences lead to a reduction of dialect-only repertoires in the present context. From a quasi-longitudinal perspective, dialect-speaking children seem to expand their repertoires as they get older. In contrast, the percentage of standard-only speaking children increases with age. Age, then, does not per se seem to be linked to an expansion of children’s repertoires. It is, however, generally associated with an increase in Standard German use. Depending on children’s starting points, this may result in a high incidence of bi-varietal repertoires (6-year-old monolingual boys and rural girls) or even higher probabilities of a standard-only repertoire (6-year-old urban girls). This finding apparently runs counter to the results by Smith et al. (2007) and Nardy (2008), who found an increase in nonstandard features during the preschool years. However, Smith et al. (2007) studied younger children than in our sample (2- and 3-year-olds) and Nardy (2008) studied formal and informal forms of French (variable liaisons) rather than dialect-standard usage.

Apart from age, the variables of “frequency of being read to”, “location” and “mono-/bilingualism” have direct effects on children’s input. The frequency of being read to is one indicator of the exposure to standard language input in our context since reading is inextricably linked to Standard German in our context. Moreover, in cases where parents generally speak dialect with their children, the switch to the standard language when reading out loud provides children with models of bi-dialectalism. Obviously, mono- and bilingual children differ in their linguistic experiences, and this also seems to be true with respect to the amount and nature of German varieties they encounter. In educational settings in particular, children with German as their second language are typically spoken to in standard-near speech in order to facilitate the acquisition of the language variety which is necessary for educational purposes. The fact that the proportion of bilingual children is far greater in urban locations than in rural ones also has implications for monolingual children: When kindergarten teachers need to make more “concessions” to comprehensibility for bilinguals, this increases the amount of Standard German that all children hear. This effect may provide one explanation of why the degree of urbanity of the location is another highly predictive factor in the development of children’s repertoires, regardless of whether they themselves are mono- or bilinguals.

Quite apart from obvious changes in the input, however, children’s repertoires may also reflect progress in the children’s knowledge about the social world and in their awareness of (socio-)linguistic phenomena as they get older. This may result in higher sensitivity toward the social meaning of linguistic variation, perhaps via “social weighting” of linguistic encodings in the mind (Sumner et al., 2013). In other words, children learn about the social prestige and the social-indexical meanings of linguistic variation (cf. Cornips, 2020), even if this knowledge remains largely implicit during the preschool years (Kasberger & Kaiser, 2019; Kaiser & Kasberger forthc.). The obvious importance of the gender variable seems to underscore the growing importance of (different facets of) prestige (cf. Trudgill, 1972). As our questionnaire data shows, the girls and boys in our study experience slightly different input (see the crosstabulation between gender and main variety spoken at home, Table 1) – a gender-related difference others have observed before (Foulkes & Docherty, 2006; Kasberger & Gaisbauer, 2020; Lenz, 2003; Smith et al., 2007). These differences in input do not seem to suffice in explaining the importance of the gender variable in the development of sociolinguistic repertoires in our data, however. It is in the question of the importance of the gender variable that an obvious issue in the study of sociolinguistic enculturation processes becomes manifest: A child’s (linguistic) personality is co-constructed in an iterative interactional process between the child and the environment. Clearly, there are no single decisive factors that determine children’s sociolinguistic development, so we can only give probabilistic statements based on the combination of different factors. And yet, there remain individual differences between children apparently experiencing very similar input and sociolinguistic environments, which we cannot fully explain at this stage. One major question yet to be answered therefore regards the personality factors and biographical experiences which contribute to the “selection process” as parts of the input a child is offered become intake for the development of the child’s own communicative repertoire.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Hugo-Moser-Prize for German Linguistics and by the Research Foundation of the University of Salzburg (Stiftungs- und Förderungsgesellschaft).

Notes

¹ Note that in keeping with German-speaking sociolinguistics, the term “dialect” is used in the sense of “local base dialect” or “local vernacular” rather than synonymous to “any language variety”.

² Whether the existence of co-occurrence restrictions means that there are in fact different distinct varieties within the “Umgangssprache” continuum is still a matter of debate; see, e.g., Durrell (1998), Martin (1996), Reiffenstein (1973), Reiffenstein (1982), Scheuringer (1997), and Scheutz (1999).

³ For a more detailed dialectological description of Bavarian dialects see Zehetner (1985) and for information on German dialect regions in general, see Wiesinger (1983).

⁴ In contrast to German-speaking Switzerland there is no routine of reading aloud in dialect in Austria, nor are there children’s books available in dialect.

⁵ Data were missing from one child.

⁶ For a detailed discussion of children’s variation behavior see Kaiser (2020).

⁷ This was a five-year-old monolingual girl from the urban location.

⁸ This categorization is based on the categories given in the questionnaire.