Adolescent Heritage Speakers: Morphosyntactic Divergence in Estonian Youth Language Usage in Sweden

Abstract

Heritage language (HL) research has investigated adults and children, while adolescents have garnered far less attention, despite adolescence being a crucial time in the development of idiolects and identities, and, hence, also for language maintenance. This study describes HL usage among Estonian-Swedish bilingual teenagers. Data were collected from 21 first- and second-generation Estonian heritage speakers (aged 12–17). Non-standard usage occurred in our corpus at low rates but was found across most speakers in certain areas of morphosyntax. We describe which factors drive such non-standard usage based on the example of two structures with more frequent non-standard occurrence: (1) object marking and (2) experiencer constructions with the verb meeldima ‘to like/please’. Around 6% of objects were marked in divergent ways. Speakers employed two strategies for marking non-standard objects: case omission (i.e., using nominative and/or unmarked forms) and substitution (i.e., using non-target-like marking). Non-standard forms occurred in 11% of experiencer constructions. Speakers diverged more with marking the nominative Stimulus than the dative-like Experiencer, although both occurred in standard and non-standard forms. The reported usage patterns can be explained through the combined effect of cross-linguistic influence, simplification, and input frequency. We also tested the relation between non-standard usage and the speaker’s sociolinguistic background, with mixed results. Considering the limited size and scope of the corpus, overall, the reported divergent usage patterns evidence the role of linguistic input and cross-linguistic effects.

1. Introduction

While recent decades have seen a growing interest in heritage languages (HLs) and their users, studies on heritage speakers have primarily focused on adults (e.g., Karaca et al. 2024) or children (e.g., Meir and Janssen 2021), with adolescents remaining under-researched, despite adolescence being a crucial period in heritage language maintenance (see Pułaczewska 2021). Studies observing teenage heritage speakers have looked into factors that influence their HL competence and usage. For instance, Minkov et al. (2019) compared Russian heritage speakers aged 13 to 17 in the USA, Finland, Israel, and Germany to describe what factors influence the continuum of HL proficiency. Their findings indicated that home language use and bilingual education significantly influence noun morphology acquisition. Comparing error rates involving inflectional noun morphology among four heritage speaker groups, they found that Russian-German and Russian-Finnish bilinguals produced significantly fewer case and gender marking errors than Russian-English and Russian-Hebrew bilinguals. The latter two groups did not report receiving any formal bilingual education. Moreover, the speakers who produced the fewest morphological errors also indicated using only Russian at home, meaning they typically had more HL input. However, these differences may also indicate the influence of L2 in language contact: namely, German and Finnish have more complex noun morphology than Hebrew and English. Pułaczewska (2021) investigated adolescent Polish speakers in Germany, tracing the influence of various factors on the frequency of Polish language use in their daily interactions. Findings indicated that the factors driving teens’ language usage differ from those that guide adults or children. Social networks, particularly interactions with age peers, were found to have a major effect on the speaker’s HL use. Furthermore, the local youth culture as well as digital media were deemed to influence the teenagers’ language use more than their familial and ethnic ties. Thus, it is important to further investigate adolescent speakers in other contexts.

In this paper, we focus on adolescent Estonian heritage speakers in Sweden. An estimated (nearly) 150,000–200,000 Estonian residents reside outside of Estonia, most of whom have relocated to neighboring countries such as Finland and Sweden, as well as English-speaking countries like the USA and Canada (Kaldur et al. 2022; Statistics Estonia 2023). Migration has occurred in several waves, with a notable efflux following the Second World War. Since the 1990s, the primary motivation for relocation is often linked to work or education. One of the largest and most active Estonian diaspora communities is situated in Sweden, where various Estonian organizations, including a kindergarten and school, continue to operate. While determining the precise size of this community is difficult, census data since 2000 indicate that at least 10,000 permanent residents of Sweden hold Estonian citizenship and/or were born in Estonia. (Raag 2010; Korkus 2021) Henceforth, we will refer to members of this community as Swedish Estonians.

There have been at least five generations of Swedish Estonians, although Estonian-born speakers and their offspring (i.e., the first two generations) can typically be viewed as active users and maintainers of Estonian as a heritage language (cf. Fishman 1989; Raag 2010). According to Raag (2010, p. 390), the first Swedish Estonian generation involves Estonian-born individuals who relocated to Sweden as children or adults; thus, the second generation consists of the first Swedish-born generation. Earlier research has shown that language contact effects are most prevalent among younger speakers (Raag 1983). In terms of language proficiency, Raag (2010, p. 402) notes a general trend: those born and educated in Estonia speak Estonian more fluently than Swedish, and those born and educated in Sweden are more proficient in Swedish than Estonian; likewise, younger speakers have a more limited grasp of Estonian, and older speakers are less proficient in Swedish.

Young people, particularly teenagers, are open to linguistic innovations (see Eckert 1997; Tagliamonte 2016), and, hence, differ in language use from their parents. Yet adolescent heritage speakers are less likely to diverge from their parental language model. This phenomenon stems from their parents serving as the primary source for HL acquisition, due to limited contact with L1 peers in the origin country. Consequently, we can expect young heritage speakers to be less prone to linguistic innovations. The language contact environment of contemporary youth is very different from earlier generations, as they inhabit a mediatized, globalized, multilingual culture (Androutsopoulos 2015; Koreinik et al. 2023). While various studies of Swedish Estonians have explored language contact, very few have included the effect of English, which has become more prominent over recent decades. However, alongside the prevalence of English-language input, Swedish Estonians in the present day also have easy access to Estonia, where Estonian is spoken as a majority language.1 Swedish Estonian teenagers, then, navigate between Estonian, Swedish, and English and may provide valuable insights into the variable effects of HL usage.

This paper explores the language use of 21 bilingual Swedish Estonian adolescents, 12 to 17 years of age. All the speakers are either first- or second-generation Swedish residents. The contrast between Estonian and Swedish morphosyntax makes this language contact a fruitful context for investigating HL divergence from standard Estonian. This study focuses on the participants’ non-standard use of Estonian morphosyntax. We focus on this domain since earlier studies (e.g., Benmamoun et al. 2013; Polinsky and Scontras 2020) have indicated that it is particularly vulnerable to change when morphologically different languages (here: Swedish, Estonian, and also English) are present in the speakers’ active repertoire. We investigate non-standard usage in two linguistic constructions—object marking and experiencer constructions with the verb meeldima ‘to like/please’. These structures were selected for several reasons: previous studies (Torn 2003; Pool 2007; Teral 2007; Ehala 2012) have shown object marking in Estonian to be challenging for L2 learners and vulnerable in HL usage, often undergoing simplification (see Section 2.2. for more on simplification). The use of experiencer constructions, on the other hand, has not yet been described in the context of Estonian heritage speakers; however, its morphosyntactic complexity compared to Swedish and English allows for a study of divergences within a more narrowly defined linguistic context. We selected the verb meeldima ‘to like/please’, due to its prominence in the corpus, since participants often spoke of their interests during the recordings. Overall, these two structures stood out in the utterances coded for non-standard usage in our corpus: object marking was the most frequent structure within the extracted non-standard nominal morphology, being a highly frequent structure overall, and the ‘like’ construction was notable in its high proportion of non-standard usage. We analyze the extent to which adolescent HL speakers’ usage shows effects of language contact, simplification, and frequency along with sociolinguistic factors. Thus, this paper addresses the following research questions:

• How do speakers make use of non-standard inflection in object marking and experiencer constructions?
• Does the participants’ use of object marking and experiencer constructions indicate cross-linguistic influence, simplification, or effects of input frequency?
• How do sociolinguistic factors, namely, the speakers’ home language environment and generational belonging, explain the use of divergent forms?

In Section 2, we outline the concept of heritage languages, summarize findings from previous studies involving the use of morphosyntax in heritage speakers, and describe how the structures under investigation (object marking and experiencer constructions) are formed in Estonian, Swedish and English. Section 3 describes the corpus data and method, providing an overview of the participants’ linguistic background and the data coding process. We present our results in Section 4, where the observed structures are analyzed in two subsections, followed by a short discussion of the factors influencing divergent usage. We discuss the implications and conclude in Section 5.

2. Background

2.1. Heritage Language Research

As the term heritage language (HL) is a relatively recent addition to linguistic research, it is important to define it here. Certain key points differentiate an HL from first (L1), second (L2), and foreign languages. A heritage language is typically an L1, i.e., the first language (or one of the first) a heritage speaker acquires. HL acquisition occurs in a naturalistic environment, through family members (as opposed to a foreign language, which is acquired in a formal setting), within the broader context of a region where the dominant language differs from the heritage language (Kupisch and Polinsky 2022). HL acquisition differs from typical L1 acquisition in that speakers have limited linguistic input, thus potentially leading to incomplete HL acquisition. Polinsky and Kagan (2007) argue that heritage speakers also differ from L2 speakers thanks to the cultural links a heritage speaker has to the language. Heritage speakers also often differ from L2 learners in having high proficiency in the spoken language while being less accustomed to writing in their HL (e.g., see Rothman 2007; Montrul 2010; Montrul and Bateman 2020).

Heritage speakers’ degree of HL fluency is greatly variable and shaped by HL input and output. Polinsky and Scontras (2020, p. 4) highlight that both the quality and quantity of heritage speakers’ linguistic input is different from that of a monolingual learner. This results in divergent attainment, in which the linguistic system acquired by heritage speakers is different from the baseline that served as the input for HL acquisition. Benmamoun et al. (2013, p. 134) suggest that proficiency in the HL can be predicted by examining the manner of acquisition of the language and the length of exposure a learner has to their HL. Thus, heritage speakers born in the L1 environment (i.e., first generation immigrants) have a period of exposure to richer linguistic input than speakers who are raised entirely in a society with a different dominant language. There is also a difference in the input of first- and second-generation speakers born abroad, as the latter receive even more limited input, with other heritage speakers serving as their primary input model. Montrul and Bowles (2009) found that limited linguistic input during childhood can lead to the incomplete acquisition of the heritage language (as reflected in a preference for simplified grammatical forms among Spanish heritage speakers). Au et al. (2008) also emphasize the impact of early experiences, highlighting how the exposure to and use of the HL in childhood shape the speakers’ knowledge of grammar and phonology. With the example of Korean-American college students, Kang and Kim (2012) find a positive correlation between a speaker’s perceived and actual HL competence and between their ethnic identification (i.e., how strongly they identify as Korean) and HL competence. By the same token, Cho’s (2000) findings indicate that higher HL fluency increases a speaker’s likelihood of social connection with their heritage community. Hence, there seems to be a bidirectional effect: increased HL input results in higher HL proficiency, while greater confidence in using the HL strengthens the speaker’s ability to connect with their (ethno-)linguistic community, thereby increasing input.

The linguistic competence of heritage speakers can be described by a continuum (Polinsky 2018b; Polinsky et al. 2023). Certain HL structures acquired in childhood may remain weak in adult grammar (Polinsky 2018a, p. 559), since the HL may gradually become secondary compared to the societal language after the child begins kindergarten or school (Polinsky et al. 2023). Rothman (2007, p. 360) points out that, while all heritage speakers can be viewed as bilinguals, not all bilinguals can be described as heritage speakers. Based on how actively they use their heritage language, HL speakers can be productive or receptive bilinguals. Productive, or active, HL speakers vary in terms of how much they diverge from the input language (Polinsky 2018b). A baseline serves as an important point of comparison, as studies on HLs are comparative in nature. However, it is not always clear what should be treated as the baseline. Some studies (e.g., Montrul 2010) have compared multilingual heritage speakers with monolingual native speakers from the country of origin, with the latter’s language use being the baseline. Rothman et al. (2023) problematize using monolinguals’ language use as the baseline, suggesting that, by contrasting bilinguals to monolinguals, the full extent of bilingualism remains overlooked. Cabo and Rothman (2012) argue that HL speakers are exposed to different input than monolinguals, and their linguistic competence should be viewed as different, rather than incomplete. Thus, they suggest taking first-generation immigrants to represent the baseline. In longitudinal studies, heritage speakers can serve as their own baseline for comparing their language use at different ages (Montrul 2015). In this study, we will investigate the divergences of two constructions, using the language of monolingual Estonians as a baseline. Though we recognize the limitations of this approach, we lack data on the speakers’ parents’ language, which would provide a more appropriate point of comparison.

2.2. Morphosyntax in Heritage Language Usage

The use of morphosyntax has been a major focus of studies involving heritage language speakers. Benmamoun et al. (2013) suggest that certain areas of grammar are particularly vulnerable when a language is acquired in a bilingual environment. In the case of heritage languages, “nominal morphology is more vulnerable than verbal morphology, and within verbal morphology, agreement is the most vulnerable” (Benmamoun et al. 2013, p. 143). Polinsky and Scontras (2020, p. 6) suggest a number of reasons as to why morphology is such a vulnerable domain, most prominently noting that heritage speakers struggle with understanding the dependency relations of constituents. Rinke et al. (2024) suggest that child heritage speakers struggle most with acquiring structures that also develop late in monolingual language acquisition; however, the acquisition of some morphosyntactic structures takes more time for heritage speakers than monolinguals due to the differences in input. The effects of language contact depend partly on the morphological complexity of the languages involved: when the HL has rich morphology and the dominant societal language (e.g., English) does not, the HL typically starts losing overt morphology. However, complexity can be affected in different ways. A study by Varatharaj et al. (2024) analyzed heritage speakers of Cantonese, Faetar, Italian, Russian, Ukrainian, and Korean across three generations in Canada, observing that complexity increased and decreased in different areas. The Ukrainian sample, which had the most data, showed that, as the generations progressed, word structure became less complex, while word order became more so.

A prominent feature of HL use is claimed to be simplification, or structural shrinkage,2 especially in inflectional morphology (Kusters 2000). The notion of simplification is somewhat ambiguous, since what is ‘morphologically complex’ depends on the purpose of the definition; thus, certain constructions deemed as complex by the linguist may not necessarily be difficult to acquire for the speaker (Anderson 2015). Moreover, languages that undergo simplification in some areas may maintain complexity in other domains (Kantarovich et al. 2021; Lohndal and Putnam 2024). On a morphosyntactic level, simplification manifests itself in numerous ways, such as changes in word order, the reduction in inflection (i.e., analytic forms are used in place of synthetic ones), or the overuse of regular forms (see examples in Polinsky 2018b). Several studies have explored patterns of simplification. For instance, Saad et al. (2019) investigated Abui-Malay bilinguals and suggested that possessive marking is subject to simplification, mostly driven by adolescent male speakers. While the construction is variable, younger speakers tend to overgeneralize the non-reflexive possessive prefix. This change is attributed to two main factors: cross-linguistic influence (see also Lucas 2012) and limited linguistic input. Both of these factors can lead to incomplete HL acquisition (see also Polinsky 2006), exhibited in various constructions, such as nominal and verbal agreement and case-marking. Bolonyai’s (2007) study on the production of possessive and verbal inflections of Hungarian-English bilingual children showed signs of reduction induced by the speakers’ dominant L2. Specifically, the inflections were often omitted in possessive be-clauses. Bolonyai argues that instability is found at the syntax-semantics interface, along with other contributing factors, such as structural complexity and ambiguity in L1 and L2 transfer. Håkansson (1995), observing a group of expatriate Swedish students, noted that noun phrase morphology had undergone simplification, whereas Swedish word order was less subject to change; this finding again shows that certain aspects of grammar are more vulnerable than others. Lexical frequency is another important factor, as shown by Nagy and Petrosov’s (2024) study on heritage Russian speakers from Canada. Their findings revealed that lexical frequency significantly influences case marking, with more frequently used words using more canonical case marking. Their study also explored the effects of the speaker’s gender and generation, finding that both play a role but on a smaller scale.

Object marking has been of particular interest in contexts of contact between languages in which object marking differs. Estonian exhibits differential object marking (DOM; see also Argus 2008, 2016; Ogren 2015; Vihman et al. 2020), a phenomenon that has been of interest to HL researchers (such as Montrul et al. 2015; Di Salvo and Nagy 2023; Thane 2024). Montrul (2014), for example, investigated DOM among Spanish heritage speakers (see also Montrul 2004, 2012; Montrul and Bowles 2009). Incomplete acquisition was attested in the simplification of DOM in marked contexts. This indicates that greater divergence may result from the speakers’ limited HL exposure as well the influence of English. In comparison, data from Romanian heritage speakers showed that Romanian DOM is less vulnerable in contact with English than when the contact occurs between Spanish and English (Montrul and Bateman 2020). Hence, the language pairs matter, in addition to the complexity of the observed structure. The age of acquisition also plays a significant role since simultaneous bilinguals were more prone to divergences than sequential bilinguals. Coskun Kunduz and Montrul (2022) conducted a study of Turkish heritage speakers in the US, mapping out factors that lead to variability in DOM acquisition. The study involved second-generation heritage speakers (both children and adults) as well as first-generation immigrants, who were then compared to monolingual speakers. While the first-generation adults were not subject to L1 attrition, second-generation speakers displayed more variable and divergent use of the DOM construction; this phenomenon was found to be related to the quantity of childhood input, rather than the quality of input. The contrasts involved in the Estonian DOM system (between bounded and unbounded events, paired with wholly and partially affected objects), the marking options (overt genitive and partitive case, unmarked nominative), and the factors governing the choice of object marking (including lexical, semantic, and syntactic levels, see Section 2.3.1 and Ogren 2015) are unlike those at play in the more widely researched Spanish or Romanian.

The DOM system of Estonian has also been of interest in earlier research on heritage Estonian speakers. For example, Torn (2003) reported that, as a result of limited exposure as well as English dominance, Estonian-English bilingual children in Britain displayed systematic overuse of the partitive case for marking objects. Maandi (1989) conducted a cross-generational study involving Estonian speakers in Sweden to describe object case-marking trends. Since all participants spoke Estonian fluently, they displayed awareness of the object marking system in Estonian; however, Maandi noted that “age is the most important variable with respect to the use of objective case markers” (Maandi 1989, p. 239). While age is an important factor, other studies observing Swedish Estonians (e.g., Raag 1983; 2010; Laagus et al. 2004) have reported that non-standard object marking is the result of Swedish influence, most evident in the frequent use of the nominative form in marking total and partial objects (see also Section 2.3.1). This pattern was also noted in the language use of Estonians living in Denmark, who also sometimes showed a tendency to mark total objects with the non-standard partitive case, which the study attributed to limited input (Teral 2007). In fact, in these contexts, it is not always possible to make a difference between cross-linguistic influence and simplification.

Experiencer constructions, on the other hand, have not been as extensively scrutinized as direct object marking; within heritage Estonian studies, this structure has not been previously investigated. Irizarri van Suchtelen (2014) examined the use of dative constructions (including experiencer constructions similar to the Estonian construction) among heritage Spanish speakers in the Netherlands. A comparison between first generation bilinguals and heritage speakers revealed that the former resembled monolingual L1 speakers in their use of datives. Heritage speakers, on the other hand, diverged from the L1 usage. Furthermore, Irizarri van Suchtelen reports that the construction is subject to innovation among bilinguals due to cross-linguistic influence from Dutch. Innovations include the restructuring of canonical datives and the avoidance of optional dative constructions. Another study observing five dative constructions (involving give-events) on the example of Ambon Malay heritage speakers in the Netherlands (Moro and Klamer 2015, p. 263) found that “a change in frequency of certain constructions” can lead to higher rates of innovative give-constructions. In turn, this frequency change is intertwined with other factors—cross-linguistic influence and the path of HL acquisition—that drive the use of divergent forms. For example, speakers who grew up with limited Ambon Malay input tended to substitute ‘two predicate’ constructions with a single verbal predicate and display a much higher incidence of double object constructions than homeland speakers. Teasing apart the influence of various influential factors is difficult, particularly when the heritage language has more complex morphology than the societal language.

2.3. Differences Between Estonian, Swedish, and English

Typologically, the Swedish-Estonian contact constitutes a case of morphosyntactically contrasting languages. English, the other contact language in the participants’ active usage, contrasts also with Estonian, and is structurally more similar to Swedish.

Estonian inflectional morphology is richer: the Estonian nominal system has fourteen cases and two numbers, with nominative, genitive, and partitive being the most frequently used cases (Kaalep 2010; Miljan and Vihman 2023). The partitive case is the most challenging to acquire for L2 learners (Pool 2007), due to both allomorphy (three general patterns used in partitive singular, five patterns in partitive plural formation) and lexically specific information (theme vowels and stem changes) (Blevins 2008). Among the three most frequent cases, partitive is the most morphologically marked, while nominative is unmarked (Viht and Habicht 2019). The nominative, genitive, and partitive each serve as stems for the formation of other cases, with the genitive form serving as a stem, or base, for the singular form of the remaining 11 cases in the paradigm. Inflectional affixes also depend on the number, with the plural marker -d added to the genitive stem to form the plural nominative, and other plural forms, most of which use the genitive plural as a stem. The partitive plural is exceptional, with a number of possible formatives (see Viht and Habicht 2019, p. 92).

Within noun phrases, both the modifier and head decline and agree in number and case for most cases, except for the final four cases, as shown in

Table 1. Noun declension in Estonian, showing paradigms of two examples in singular and plural: ilus keel ‘beautiful language’ and õnnelik teismeline ‘happy teenager’.

Case	Singular	Plural
nominative subject/citation form	ilus keel ‘beautiful language’ õnnelik teismeline ‘happy teenager’	ilusad keeled ‘beautiful languages’ õnnelikud teismelised ‘happy teenagers’
genitive possessive, affected object	ilusa keele õnneliku teismelise	ilusate keelte õnnelike teismeliste
partitive partially affected object	ilusat keelt õnnelikku teismelist	ilusaid keeli õnnelikke teismelisi
illative3 ‘into’	ilusasse keelesse/keelde õnnelikusse/õnnelikku teismelisse/teismelisesse	ilusatesse keeltesse õnnelikesse teismelistesse
inessive ‘in’	ilusas keeles õnnelikus teismelises	ilusates keeltes õnnelikes teismelistes
elative ‘out of’	ilusast keelest õnnelikust teismelisest	ilusatest keeltest õnnelikest teismelistest
allative ‘to’	ilusale keelele õnnelikule teismelisele	ilusatele keeltele õnnelikele teismelistele
adessive ‘on’	ilusal keelel õnnelikul teismelisel	ilusatel keeltel õnnelikel teismelistel
ablative ‘off of/from’	ilusalt keelelt õnnelikult teismeliselt	ilusatelt keeltelt õnnelikelt teismelistelt
translative ‘becoming’	ilusaks keeleks õnnelikuks teismeliseks	ilusateks keelteks õnnelikeks teismelisteks
terminative ‘up to’	ilusa keeleni õnneliku teismeliseni	ilusate keelteni õnnelike teismelisteni
essive ‘as’	ilusa keelena õnneliku teismelisena	ilusate keeltena õnnelike teismelistena
abessive ‘without’	ilusa keeleta õnneliku teismeliseta	ilusate keelteta õnnelike teismelisteta
comitative ‘with’	ilusa keelega õnneliku teismelisega	ilusate keeltega õnnelike teismelistega

. As shown in the examples, noun phrases exhibit adjective concord, with the adjective usually taking the same inflectional ending as the noun; in terminative, essive, abessive, and comitative cases, the adjective is in (unmarked) genitive. As the inflectional endings contain considerable grammatical information, indicating the syntactic role of arguments, Estonian word order is much less rigid than in Swedish and English.

Swedish, in contrast, expresses more nominal categories but makes use of only three cases: nominative, accusative, and genitive, with the accusative expressed only on pronouns (e.g., du ‘you’ → dig). The genitive case is marked with the suffix -s. In addition to marking case, Swedish also uses inflectional endings to mark the gender, definiteness, and number of nouns. Swedish differentiates between two grammatical genders (common and neuter), which are marked on articles in (singular) indefinite form; in definite contexts, gender is marked with a suffix on the noun and the appropriate form of the article (den or det).

In noun phrases, the modifier and head agree in gender, number, and definiteness but not in case, as shown in

Table 2. Noun declension in Swedish, showing paradigms of two examples in singular and plural: en glad tonåring ‘a happy teenager’ (common gender) and ett vackert språk ‘a beautiful language’ (neuter gender).

Case	Singular		Plural
	Indefinite	Definite	Indefinite	Definite
nominative	en glad tonåring ett vackert språk	(den) glada tonåringen (det) vackra språket	glada tonåringar vackra språk	(de) glada tonåringarna (de) vackra språken
genitive	en glad tonårings ett vackert språks	(den) glada tonåringens (det) vackraspråkets	glada tonåringars vackra språks	(de) glada tonåringarnas (de) vackra språkens

. Information about other grammatical functions can be conveyed either through prepositions (e.g., en vän till mig ‘a friend to me’, han är med oss ‘he is with us’) or word order, which is more rigid in Swedish than in Estonian, though both languages follow an unmarked SVO order (see Teleman et al. 1999).

Morphosyntactically, English is similar to Swedish, with three main cases: nominative, genitive, and accusative. Nouns and most pronouns make use of the latter two. Only personal pronouns (and the interrogative/relative who) have a distinct accusative form, which functions as an object case, in addition to the nominative subject case. Nouns also take the inflectional ending -s for plural marking. English displays no modifier-head agreement in noun phrases (compare the big apple and the big apples). Like Swedish, English has articles, but they do not show inflection. Another similarity with Swedish is the role of prepositions, which help determine the grammatical functions of syntactic units; like Swedish, argument structure is determined primarily by word order, which is relatively rigid (Quirk et al. 1985).

In this paper, we focus on how two syntactic constructions, direct object marking, and experiencer constructions with the verb meeldima ‘to like/please’, are used by the participants. While object marking has been the focus of earlier heritage Estonian studies, thus providing a point of comparison for exploring HL usage, we also include experiencer constructions, because their structure in Estonian differs from that of Swedish and English in form and complexity. We selected the verb meeldima ‘to like/please’, for two main reasons: (1) utterances using this verb are prominent in our corpus due to the content of the recordings (speakers talked about everyday topics, including their likes and dislikes), and (2) its non-standard use stood out in the speech of several speakers in the corpus. In order to probe whether non-standard use is the result of cross-linguistic influence, we first outline how these constructions compare in Estonian, Swedish, and English.

2.3.1. Object Marking

Object marking is complex in Estonian, because of both the nominal inflection paradigms and because of the differential marking system involved (see Ogren 2018). It is one of the more complicated aspects of grammar for L2 speakers to learn and is reported to be challenging for heritage speakers (Teral 2007, pp. 72–74; Pool 2010). Children acquiring L1 Estonian use object marking accurately very early (Argus 2009; Vihman et al. 2020), although they continue to make errors with less frequent parts of the nominal system for much longer. Granlund et al. (2019) found that genitive objects were replaced by partitive as a possible escape strategy in an elicitation experiment with 3- to 5-year-olds. This only happened in 3% of trials, but most of these involved complex, stem-changing target noun forms. Objects are most frequently marked with the partitive case, but this alternates with both the genitive and nominative (Miljan and Vihman 2023; see also Figure 1), depending primarily on predicate semantics, as well as sentence type, polarity, and object number.

Estonian differentiates between total and partial objects, a distinction based primarily on aspectual properties of the predicate. Partial objects take partitive case and are used together with both verbs like keetma ‘to boil (transitive)’ (known as aspectual verbs, which take both total and partial objects, see ex. (1a), direct object in bold and verb is underlined), and partitive verbs like armastama ‘to love’, which are used only with partitive objects (as in (1b), direct object in bold):

1.	a.	ma	keet-sin	suppi	/	supi
		1sg.nom	boil-pst.1sg	soup.par.sg		soup.gen.sg
		‘I boiled some soup/the soup.’
	b.	ma	armasta-n	sin-d
		1sg.nom	love-prs.1sg	you-par.sg
		‘I love you.’

Total objects are marked with genitive or, in particular, syntactic contexts, nominative case. They occur in perfective, telic predicates, expressing an action that has been completed, as in (2a)–(2b). Nominative case is used with plural total objects (2b) as well as total objects in subjectless clauses like imperatives and impersonals (for a more detailed explanation, see Ogren 2018; Vihman et al. 2020, pp. 27–30; Miljan and Vihman 2023).

2.	a.	ma	leid-sin		uue	raamatu
		1sg.nom	find-pst.1sg		new.gen.sg	book.gen.sg
		‘I found a new book.’
	b.	ma	ost-sin		uhke-d	kinga-d
		1sg.nom	buy-pst.1sg		fancy-nom.pl	shoe-nom.pl
		‘I bought fancy shoes.’
	c.	ma	ei	leid-nud	uu-t	raamatu-t
		1sg.nom	neg	find-app	new-par.sg	book-par.sg
		‘I did not find a new book.’
	d.	ma	ei	ost-nud	uhke-id	king-i
		1sg.nom	neg	buy-app	fancy-par.pl	shoe-par.pl
		‘I did not buy new shoes.’

A negated clause always takes partitive objects, as in (2c)–(2d). Pronouns and numerals behave exceptionally. First- and second-person plural pronominal objects always use the partitive (3a); with singular pronouns, the form varies between the genitive and partitive (3b). Numerals and the pronouns kõik ‘all’ and teine ‘other’ always use the nominative when they are in a total object position (3c) and the partitive when marking a partial object (3d) (Erelt and Metslang 2017).

3.	a.	ma	nägi-n	te-id				eile
		1sg.nom	see-pst.1sg	2pl-par				yesterday
		‘I saw you yesterday.’
	b.	ma	kutsu-sin	sin-d	/		sinu			tantsu-le
		1sg.nom	invite-pst.1sg	2sg-par	/		2sg.gen			dance-all.sg
		‘I invited you to a dance.’
	c.	ma	tegi-n	kõik	ära
		1sg.nom	do-pst.1sg	everything	away
		‘I finished everything.’
	d.	ma	vaata-n	teis-i
		1sg.nom	look-prs.1sg	other-par.pl
		‘I look (at) others.’

In Swedish, objects are generally unmarked (except for personal pronouns, which can also be in accusative) and are denoted by their linear position in the clause (4a). However, if the direct object is a personal pronoun, it is marked with accusative (4b). Additionally, indirect objects are marked with prepositions (4c) (see Teleman et al. 1999). In English, objects are marked similarly to Swedish, as can be seen in the translations of ex. (4a)–(4c) (Quirk et al. 1985).

4.	a.	jag	kör	bil
		1sg.nom	drive.prs	car.sg.indf.nom
		‘I’m driving a car.’
	b.	jag	såg	dem	igår
		1sg.nom	see.pst	3pl.acc	yesterday
		‘I saw them yesterday.’
	c.	jag	gav	en	present	till	henne
		1sg.nom	give.pst	art	present.sg.indf.nom	to	3sg.acc
		‘I gave her a present.’

2.3.2. Experiencer Constructions

In Estonian, experiencer constructions are typically expressed with non-canonical subject Experiencers, with the verb agreeing with the nominative, often post-verbal, Stimulus (Seržant and Kulikov 2013; Seržant 2015; Lindström and Vihman 2017), as shown in Figure 2. L1 children may not acquire this construction before the age of three. We focus on clauses expressing the predicate ‘to like/please’, with the verb meeldima. Meeldima-constructions take an Experiencer argument with many subject-like properties (given, clause-initial, human animate, often pronominal) but marked with an oblique case; these arguments are classified as non-canonical subjects (Metslang 2013). There is some variation in usage, and speakers use both allative (5a)–(5b) and adessive (5c) Experiencers (Lindström and Vihman 2017, p. 10). In Estonian, the verb agrees with the Stimulus subject in number and person, reflected in the verb marking in (5a)–(5b); when there is no nominative subject, the verb takes default third-person singular inflection (5c).

5.		EXPERIENCER	VERB	STIMULUS/THEME
	a.	mu-lle	meeldi-b	sügis
		1sg-all	like-prs.3sg	autumn.nom.sg
		‘I like autumn.’
	b.	ta-lle	meeldi-vad	koera-d
		3sg-all	like-prs.3pl	dog-nom.pl
		‘S/he likes dogs.’
	c.	ta-l	meeldi-b	kõndi-da
		3sg-ade	like-prs.3sg	walk-inf
		‘S/he likes to walk.’

Swedish, like other mainland Scandinavian languages, has undergone several changes, one of which resulted in dative subject-like arguments becoming nominative subjects (Árnadóttir and Sigurðsson 2013). The Swedish Experiencer makes use of the nominative case for both arguments (6a–6c; see also Figure 3).

6.		EXPERIENCER	VERB	STIMULUS/THEME
	a.	jag	gilla-r	vår-en
		1sg.nom	like-prs	spring-def.sg
		‘I like spring.’
	b.	hon	gilla-r	katt-er
		3sg.nom	like-prs	cat-indf.pl
		‘She likes cats.’
	c.	han	gilla-r	att	sova
		3sg.nom	like-prs	to	sleep.inf
		‘He likes to sleep.’

In Estonian, the verb agrees with the Stimulus subject in number and person, reflected in the verb marking in (5a) and (5b); when there is no nominative subject, the verb takes default third-person singular inflection (5c). In Swedish, the verb form depends only on the tense, using the same suffix across all person-number categories (6a–6c). In English, the verb form also remains the same regardless of the Stimulus form (see translations of ex. (5), in which the verb form depends on the person of the subject Experiencer, not the Stimulus). In all three languages, the Stimulus can be a nominative singular or plural noun phrase (5a–5b and 6a–6b), an infinitival verb phrase (such as in 5c and 6c) or a relative clause.

3. Data and Method

3.1. Participants

This study analyzes conversational data collected from 21 participants (aged 12–17 years; M = 14.57; 15 female, 5 male, one unspecified gender). All informants (see also

Table 3. Overview of the participants’ sociolinguistic background (i.e., their age range, place of birth, and home language environment). Each participant is marked with a code. For the sake of anonymity, we do not report their gender, exact age, or, in the case of Estonian-born speakers, their time of relocation to Sweden.

Age Group	Participant	Place of Birth		Home Environment
		Estonian-born	Swedish-born	Monolingual	Multilingual
12–14	T1	x		x
	T2		x		x
	T3	x		x
	T4	x		x
	T5		x		x
	T6		x		x
	T7		x	x
	T8	x		x
	T9		x	x
	T10		x	x
	T11		x	x
15–17	T12		x		x
	T13		x		x
	T14		x		x
	T15		x	x
	T16		x		x
	T17		x		x
	T18	x		x
	T19		x		x
	T20		x		x
	T21	x		x

) are Estonian heritage speakers living in Sweden who, based on self-reported ratings on a 5-point scale,4 have a good command of Swedish (M = 4.8), Estonian (M = 4.3), and English (M = 4.2).

The speakers were living in various parts of Sweden at the time of data collection (2022): the majority were based in or around Stockholm, while others resided in the southern and western parts of Sweden. Six were born in Estonia (henceforth referred to as the first generation, based on the classification of Raag 2010), and fifteen are Swedish-born (i.e., the second generation). First-generation speakers had lived in Sweden for an average of nine years. The longest period an Estonian-born participant had lived in Sweden was approximately 12 years, whereas the most recent arrivers reported having relocated to Sweden 6 or 7 years earlier. At the time of emigration, Estonian-born participants were aged between two and ten years (M = 4.8).

The participants have heterogeneous linguistic backgrounds: all have an Estonian-speaking mother, 10 have an Estonian-speaking father, and 11 listed other languages (mostly Swedish) as their father’s L1. Ten participants were raised in a monolingual Estonian household, others grew up in a multilingual household: eight in a bilingual (Estonian and Swedish) home, and three in trilingual (Estonian, Swedish, and another language) home.

Most of the speakers were acquainted with or related to another participant in this study; only a couple of speakers did not have ties to other participants. We made use of their connections in order to recruit participants and to arrange recording sessions in which they conversed with HL-speaking friends or acquaintances. The few who did not have ties to other participants may have had contact with older family acquaintances but lacked active communication with HL-speaking peers. It is also important to note that some of the participants had additional Estonian language support outside the home, either by attending an Estonian-medium kindergarten and/or school or via private lessons with an Estonian tutor. We do not observe the influence of this factor primarily because we lack information on how many participants received formal Estonian education. Several of the Stockholm-based participants (both first- and second-generation) attended the local Estonian school; we do not have data on how long they had done so.

3.2. Data

The analyzed corpus consists of transcriptions of nearly 20 h of recordings, ranging between 63 and 97 min (M = 81). The total corpus size is 174,917 spoken word tokens, including the investigator’s (the first author of this paper) utterances and occasionally those of others. The speech of the teenage participants included in this analysis comprises 149,459 tokens (see Figure 4); they predominantly conversed in Estonian, while around 3% of the spoken tokens contained non-Estonian words (mostly English and Swedish).

Before data collection, ethics approval was granted by the Research Ethics Committee at the University of Tartu (protocol No. 361/T-5). Spoken data were collected in two stages in 2022. Most of the recordings were conducted in person in Estonia or Sweden, and two were conducted online via a video calling application for logistical reasons.

Using a sociolinguistic approach (see Aalberse et al. 2019), data were elicited in the form of semi-structured dialogues. Where possible, participants conversed in pairs with a fellow Swedish Estonian peer. When this was not possible (if the speaker did not have local Estonian acquaintances in their age range or they were unavailable at the time of recording), they spoke with the investigator.5 To help the participants start and maintain the conversation, prompts were created for the recording sessions. These prompts were in the form of bilingual (Estonian and Swedish) question cards, which the participants could pick up and choose from. Notwithstanding the differences in conversational partners, the conversations revolved around similar, everyday topics, including school life, leisure activities, etc. Participants were free to move between the questions as they wanted and could converse in the language of their choice; nevertheless, throughout the recordings, participants mostly spoke in Estonian. The speakers were also encouraged to talk about topics not included in the prompts; such spontaneity occurred mostly in interactions with peers.

Upon completion, each recording was then transcribed using the ELAN transcription and annotation software (ver. 6.5; ELAN 2023). An automated speech-to-text tool was used to partially help with the speech transcription (Olev and Alumäe 2022). Due to the tool’s limitations of working with multilingual, nonstandard data, the automatically generated text was manually checked and corrected. The data were transcribed using a modified version of the transcription conventions of the Estonian Teen Language Corpus (see Vihman et al. 2022).

Datasets and Analysis

Out of the corpus of 18,038 utterances (containing only teen speech), we first manually detected and marked instances of non-standard inflectional morphology (for a total of 861 utterances, or around 5% of all utterances) and found that, of these, around two-thirds (N = 577, or 67%) involved case-marking. Since some utterances contain multiple divergences, we analyzed each instance separately; in total, these 577 utterances contain 672 instances of divergent nominal morphology. Hence, we selected two of the most prominent constructions with divergent case-marking to analyze more closely. Although the overall frequency of deviances is low, it is nevertheless important to investigate the extent to which the divergences show systematic patterns. This analysis can potentially signal which constructions are vulnerable to change in heritage Estonian. Additionally, it allows us to compare with earlier studies observing Estonian heritage speakers to examine whether morphosyntactic divergences in Estonian heritage speakers are as prominent in present-day generations as in earlier ones. Moreover, considering the corpus size, the frequency of occurrence was less reliable as an indicator than the regularity of patterns. Hence, we have chosen two of the structures, which most consistently recur with non-standard marking. Our estimations of relative frequency are detailed below.

We extracted two datasets for analyzing non-standard inflectional morphology, one for object marking (N = 243) and one for experiencer constructions involving the verb meeldima ‘to like/please’ (N = 78). The first dataset consists of all (and only) instances of non-standard object marking, manually identified during the transcription process by the first author and then subjected to two rounds of further inspection by each author. In instances of disagreement, consensus was arrived at through discussion between the authors.

In order to estimate the prevalence of non-standard usage in the two constructions of interest, we compared the dataset of divergent use to the rest of the corpus. Because the corpus is morphologically unparsed,6 extracting all examples of transitive object usage was not feasible. In order to estimate the prevalence of non-standard object case-marking, we first conducted a random sample analysis on a dataset consisting of 1002 utterances and manually coded whether each utterance contained a transitive object. We found that 208 utterances (or 21%) were coded as having overt transitive objects; 794 did not. If we apply this percentage to the whole corpus, then we postulate that 3778 utterances are likely to involve object marking; divergences (N = 243) hence make up around 6% of direct object contexts. For the second dataset, all utterances containing the verb meeldima (in any form) were extracted from the corpus and manually checked. This resulted in a dataset of 713 utterances containing all uses of the Experiencer ‘like’ construction, including both standard (N = 635, 89%) and non-standard (N = 78, 11%) usage. Each utterance was further coded manually, as detailed below.

For compiling and coding the object marking dataset, we first collected all instances of utterances with non-standard object marking and entered each as a separate item. For each example, we manually coded the word class of the object (e.g., noun, pronoun, noun phrase), the form produced (case and number), and code-switching (English, Swedish, or another language). We added the expected (standard Estonian) morphology (case and number) and its syntactic context. We further categorized the non-standard uses into two groups: the omission of object marking (using unmarked in place of case-marked nominals) and the use of overt, non-standard case-marking (substitution). In the latter case, we also coded whether the substitution involved using a more or less frequent case. To compare non-standard object marking with standard marking, we drew a sample dataset and identified utterances containing transitive objects, as described above, in order to estimate the proportion of such utterances in the corpus. Further details are provided in Section 4.1.

For experiencer constructions, we identified all utterances containing an experiencer construction with the verb meeldima. We coded each experiencer clause separately. In this analysis, we coded utterances for standard and non-standard usage; produced grammatical form of the Experiencer, verb, and Stimulus; verb form, verb agreement, and contexts where agreement is not expected (e.g., negated verbs). We then analyzed the data from two viewpoints: Experiencer form-verb agreement and Stimulus form-verb agreement. More details can be found in Section 4.2.

We used a qualitative and quantitative approach to identify recurring linguistic patterns within the data. For each of the observed structures, we coded divergences in a spreadsheet. In order to explain the connection between the use of non-standard morphology and the speakers’ background and because the participant sample size is small, we ran chi-squared tests7 in R (ver. 4.3.2; R Core Team 2024) to test the significance of two sociolinguistic factors: the speakers’ home language environment and generational belonging. We also examined Pearson residuals and Cramér’s V to elaborate further how the variables contribute to the use of non-standard forms. We did not explore the effects of age or gender, as the corpus has unbalanced age groups and far more girls than other genders.

4. Results

4.1. Non-Standard Object Marking

Overall, the most frequent contexts of non-standard nominal morphology (see also Section Datasets and Analysis) were instances of direct object marking (N = 243, or 36% of divergent nominal morphology). Based on our random sample analysis (see previous Section), we found that non-standard object marking occurred in around 6% of utterances with transitive object usage. The nominals are predominantly Estonian words or phrases (N = 220, or 91%), while the object consists of a code-switched element in 23 instances (9%; 13 English, 9 Swedish, 1 Spanish). Two main strategies were employed when using divergent object marking: case omission and case substitution. These strategies will be discussed in more detail in Section 4.1.1 and Section 4.1.2, respectively. Figure 5 illustrates the normalized frequencies of non-standard use in object marking.

4.1.1. Omission of Object Marking

In the vast majority of examples of non-standard object marking, standard object marking is omitted (N = 202, or 83%). Omission means that the speakers use base forms with no object marking morphology added, either unmarked nominative forms of Estonian words or words from other languages with no integration through Estonian morphological marking. Omission generally involves the use of the nominative instead of the genitive or partitive case (127 singular and 57 plural tokens); 49 of the 202 (24%) examples involve pronouns. In some cases, where Swedish and English mark accusative pronouns, this might allow us to disentangle language contact from simplification. However, 37 of the 49 pronouns involve demonstratives (singular and plural), which are uninflected in either Swedish or English. Generally, however, this shows a high preference for the nominative case, even when the noun bears overt plural marking, but not object marking, as in ex. (7). Here, the plural noun is in the nominative case although the verb requires a partial object marked with the partitive case.

Table 4. The omission of object marking: the expected form in standard Estonian is shown in the first column; the second column indicates the morphological form produced in the corpus; and the third column shows the frequency of each type of omission.

Expected		Produced	Number of Occurrences
gen sg, par sg	→	nom sg	121 (59.9%)
gen pl, par pl			6 (3.0%)
gen sg, par sg	→	nom pl	3 (1.5%)
par pl			54 (26.7%)
par sg	→	ø (eng, sve)	14 (6.9%)
par pl			4 (2.0%)
Total			202 (100%)

illustrates the frequencies of different types of object marking omission.

7.	me	tähista-me	ju	jõulu-d	[: jõul-e]8
	1pl.nom	celebrate-prs.1pl	emph	christmas-nom.pl	[: par.pl]
	Eesti-s
	Estonia-ine.sg
	‘After all, we celebrate Christmas in Estonia.’

It is worth noting that case omission occurred more prominently when marking partial (N = 164, 81%) rather than total (N = 38, 19%) objects. Examples (8) and (9) display a recurring pattern where the verb võtma ‘to take’ is followed by a nominative singular, instead of the expected genitive singular total object. This construction occurs 15 times in the corpus and displays the same conversational context throughout; in the examples, speakers ponder which discussion prompt question to discuss next and make a suggestion using võtame X ‘let’s take X’.

8.	kas	me	võta-me	viimane	küsimus
	q	1pl.nom	take-prs.1pl	last.nom.s	question.nom.sg
	[: viimase	küsimuse]
	[: gen.sg	gen.sg]
	‘Shall we take the last question?’

9.	okei	võta-me	järgmine	[: järgmise]
	int	take-prs.1pl	next.nom.sg	[: gen.sg]
	‘Okay, let us take the next (one).’

It is possible that examples like (9) are influenced by the imperative construction in Estonian, which takes nominative total objects (võta järgmine! ‘take the next one!’), but, in standard Estonian, the hortative first-person plural võtame does not follow imperative object marking, and indicative examples with pronouns like ex. (8) cannot be explained by the imperative pattern.

We also find several examples of partial objects with case omission instead of the expected partitive. In ex. (10), the speaker uses an object with overt plural but no partitive case-marking; a similar pattern occurs within examples (11) and (12) with the verbs vaatama ‘to look/watch’ and kuulama ‘to listen’. Example (13) also displays a similar omission, with the entire noun phrase millised Rootsi tähtpäevad ‘which Swedish holidays’ in nominative instead of partitive plural. This construction was frequently problematic for the speakers, as it was part of the written prompts that they read out and discussed. The verb tähistama ‘celebrate’ is a relatively low-frequency verb that occurred in specific contexts where speakers compared Estonian and Swedish holidays.

10.	me	mängi-me	rohkem	arvutimängu-d
	1pl.nom	play-prs.1pl	more	computer game-nom.pl
	[: arvutimäng-e]	ju
	[: par.pl]	emph
	‘We play more computer games, after all.’

11.	ma	ei	vaata	nii	palju	filmi-d	[: film-e]
	1sg.nom	neg	watch.cneg	so	many	movie-nom.pl	[: par.pl]
	‘I don’t watch that many movies.’

12.	tema	[: te-da]	ma	kuula-n	ikkagi	(.)
	3sg.nom	[: 3sg-par]	1sg.nom	listen-prs.1sg	still
	väga	palju
	very	much
	‘I listen to him very much.’

13.	ee	millise-d	Rootsi	tähtpäeva-d
	int	which-nom.pl	Sweden.gen.sg	holiday-nom.pl
	[: millise-id	Rootsi	tähtpäev-i]	sa:
	[: par.pl	gen.sg	par.pl]	2sg.nom
	sa	pere-ga	tähista-d
	2sg.nom	family-com.sg	celebrate-prs.2sg
	‘Uhm, which Swedish holidays do you you celebrate with your family?’

Of the 202 objects lacking marking, 23 (11%) were other-language insertions. Very few of the non-Estonian direct objects in our non-standard usage dataset carried overt marking: 21 of these insertions were unmarked or made use of the nominative case, and only two had a case substitution (see next Section). Although the proportion of code-switched objects is relatively low, it is useful to compare the integration of non-Estonian objects with the use of Estonian object marking. Whether omission in fact features more or less prominently among code-switched objects than Estonian-language objects remains to be explored in a future study. Example (14) exhibits the unmarked use of an English-language title: the speaker lists a TV show they watch (“Gossip Girl”) but does not add an inflectional ending to the foreign title (although it would be likely to take partitive case in standard usage). The direct object in this example, an English-language title, may show the influence of English, which takes an unmarked direct object, whereas the verb vaatama ‘to look/watch’ in Swedish takes a prepositional phrase (Estonian: ma vahel vaatan Gossip Girl’i; English: sometimes I watch Gossip Girl; Swedish: ibland tittar jag på Gossip Girl).

14.	ma	vahel	vaata-n	{Gossip_Girl}
	1sg.nom	sometimes	watch-prs.1sg	Gossip Girl.ø
	[: {Gossip_Girl-i}]
	[: par.sg]
	‘Sometimes I watch “Gossip Girl”.’

However, Swedish and English direct objects are also attested in the corpus with case-marking (as shown in the following example), similarly to how Estonian teenagers use English, as described by Vihman et al. (n.d.). Like the previous example, ex. (15) illustrates how the same construction with the verb vaatama ‘to look/watch’ is used in the corpus by other speakers, but this time with partitive marking. Here, all of the titles (in both Swedish and Estonian) are morphologically integrated.

15.	mina	vaata-sin	{Baby_Boss-en-it}
	1sg.nom	watch-pst.1sg	Baby Boss-def-par.sg
	ma	vaata-sin	Lotte-t	väga	palju
	1sg.nom	watch-pst.1sg	Lotte-par.sg	very	much
	vaata-sin		Pipi-t	väga	palju
	watch-pst.1sg		Pipi-par.sg	very	much
	‘I watched “Baby Boss”, I watched “Lotte” a lot, watched “Pipi” very much.’

Almost all code-switched objects with case omission are partial objects requiring standard partitive case-marking (N = 20). One such object can be seen in ex. (16), in which the repeated direct object includes a noun in Swedish (modersmål), where it is unmarked, and in Estonian (emakeel-t), where it is marked. This does not mean that speakers avoid marking code-switched nouns with partitive case; our dataset contains several examples in which code-switched direct objects are morphologically integrated with case-marking appropriate for Estonian. In ex. (17), not only is the code-switched English noun marked in the expected case, but it is also preceded by an indefinite quantifier, which agrees in case and number. In contrast, in two examples, speakers marked the code-switched object with partitive singular instead of the expected plural form (see ex. 18, in which the inserted element bears English plural and Estonian partitive singular marking).

16.	kooli	[: kooli-s]	saa-b	ju	nagu	ekstra	vali-da:
	school.gen.sg	[: ine.sg]	can-prs.3sg	emph	like	extra	choose-inf
	{modersmål}	[: {modersmål-i}]		emakeel-t
	mother tongue-ø	[: par.sg]		mother tongue-par.sg
	‘(At) school one can like extra choose mother tongue (classes).’

17.	ma	kuula-n	minge-id	s-	{sound-e}	TikToki-s
	1sg.nom	listen-prs.1sg	some-par.pl		sound-par.pl	TikTok-ine.sg
	‘I listen to some sounds on TikTok.’

18.	mu	ema	on:	(.)	tee-b	nagu
	1sg.gen	mother.nom.sg	be.prs.3sg		do-prs.3sg	like
	pilt-e	nagu	keha-st		mm
	image-par.pl	like	body-ela.sg		int
	{X-ray-s-i}		[: {X-ray-s-e}]
	X-rays-pl-par.sg		[: pl-par.pl]
	‘My mom is does like pictures like of the body, hmm, X-rays.’

4.1.2. Substitution of Object Marking

The remaining examples of non-standard object use (N = 41, or 17%) involve substitutions of standard case-marking with overt, non-standard marking. In these examples, speakers employ overt inflection that differs from the expected form, rather than using unmarked forms. We found that most (61%) of these substitutions used either genitive or partitive singular (13 and 12 times, respectively), which are both object cases. Miljan and Vihman (2023) report four times as many partitive as genitive objects in their spoken sample. While we do not have overall figures for the proportion of genitive and partitive objects, there are various possible explanations for these substitutions. The genitive case is part of the stem for most of the Estonian noun case paradigm, and, therefore, the form of the genitive is likely to be frequently encountered in the input. HL speakers may also favor a less complex form (genitive in place of partitive). On the other hand, partitive objects are likely to be much more frequent in the input, leading to a possible overgeneralization of partitive as an object marker. Finally, the DOM system is conditioned by a complex interplay of factors, and the substitution of both genitive and partitive cases may reflect an incomplete understanding of the system. Overall, these substitutions are relatively low in frequency. The rest of the examples of substitutions were significantly less frequent and idiosyncratic.

Table 5. Substitution of object marking, indicating the expected form in standard Estonian and the form produced in the corpus. The third column indicates whether the substituted form is more (+) or less (-) frequent than the expected, or standard form, and the fourth column shows the number of attested uses of each type of substitution in the corpus.

Expected		Produced	Frequency	Number of Occurrences
nom sg	→	gen sg	+	1 (2.4%)
par sg			-	10 (24.4%)
nom pl			-	1 (2.4%)
par pl			+	1 (2.4%)
nom sg	→	ø (eng, sve)	-	1 (2.4%)
gen sg			+	5 (12.2%)
par pl			+	6 (14.6%)
par sg	→	par pl	-	3 (7.3%)
nom pl			-	2 (4.9%)
par sg	→	ine sg	-	4 (9.8%)
par sg	→	com sg	-	3 (7.3%)
par pl	→	ade pl	-	1 (2.4%)
par pl	→	gen pl	-	1 (2.4%)
par pl	→	ine pl	-	1 (2.4%)
par sg	→	abl sg	-	1 (2.4%)
Total				41 (100%)

gives an overview of object marking substitutions.

Frequency alone does not seem to underlie the examples of object case substitution. In 23 examples (56% of the substitutions), the speakers used a less frequent form (e.g., partitive singular instead of nominative singular), while, in 18 examples (44%), the form used is more frequent than the expected form (e.g., genitive singular instead of partitive plural9). In nine examples (22%), object marking involved number substitutions; in these cases, the more frequent singular was usually used in place of plural. In three examples (7%), however, the speaker made use of the partitive case and substituted the less frequent plural for singular.

Some of the more frequent substitutions in the dataset are exemplified below. In (19), the speaker uses genitive instead of partitive for marking the partial object ajalugu ‘history’. Example (20) shows an interesting case of how a quantifier phrase with the noun keel ‘language’ is inflected as an object. In standard Estonian, the quantifier palju ‘many, a lot’ assigns the partitive case to a plural count noun, though numerals assign partitive singular to (semantically plural) count nouns. In ex. (20), the speaker uses partitive singular for partitive plural, showing a preference for a more frequent form (singular instead of plural).

19.	ma	arva-n	et	oleks	õpeta-nud
	1sg.nom	think-prs.1sg	that	be.prs.3sg.cond	teach-app
	ajaloo		[: ajalugu]
	history.gen.sg		[: par.sg]
	‘I think (that) I would have taught history.’

20.	kui	lahe	sa	oska-d	nii	palju
	how	cool.nom.sg	2sg.nom	know-prs.2sg	so	many
	keel-t		[: keel-i]
	language-par.sg		[: par.pl]
	‘How cool, you know so many languages.’

On the other hand, examples (21) and (22) show less frequent and more complex inflectional endings used in place of the standard. In (21), the relative pronoun kes ‘who’ is marked with comitative (kellega ‘with’) instead of partitive. Example (22) exhibits lack of agreement: while the inserted English adjective average shows (Estonian) partitive marking, the following (Estonian) noun iseloom ‘personality’ is marked with inessive case. Example (23) shows a total object marked with partitive instead of nominative plural (this occurred in two examples).

21.	Snapchat-is	on	nee-d	kelle-ga	[: ke-da]
	Snapchat-ine.sg	be.prs.3sg	3pl-nom	who-com.sg	[: par.sg]
	ma	proovi-n	kä-tte	saa-da
	1sg.nom	try-prs.1sg	hand-ine.sg	get-inf
	‘The ones I’m trying to reach are in Snapchat.’

22.	kui	sa	vaata-d		{average-it}	(.)	iseloomu-s
	if	2sg.nom	watch-prs.2sg		average-par.sg		personality-ine.sg
	[: iseloomu]		nagu	minu	rootsi		klassi-s
	[: par.sg]		like	1sg.gen	Swedish.gen.sg		class-ine.sg
	‘If you look at the average personality in my Swedish class.’

23.	jäta-me	ne-id	kaht-e	[: nee-d	kaks]
	leave-prs.1pl	3pl-par	two-par.sg	[: 3pl-nom	nom.sg]
	koju
	home.ill.sg
	‘Let us leave these two at home.’

In ex. (24), the speaker uses comitative instead of partitive singular. This utterance could be rephrased to govern a comitative complement if it included a different (derivationally related) verb: tegelema ‘to deal with’ instead of tegema ‘to do’. Assuming this is not a production error, such usage may illustrate that the two verbs are synonymous for the speaker, but that tegema (a much more frequent verb) is more accessible than tegelema. Secondly, the substitution may show cross-linguistic influence. The speaker may have used tegema as a counterpart for the Swedish verb syssla (which means the same as tegelema). Crucially, the Swedish equivalent of the utterance (jag sysslar med fotboll väldigt mycket) requires an adposition (med ‘with’): this may also have influenced the speaker to use the comitative form in Estonian. Likewise, ex. (25) shows the much less frequent ablative singular used in place of the expected partitive singular; in this case, there may be phonological reasons for the substitution.

24.	ma	tee-n	jalgpalli-ga	[: jalgpall-i]
	1sg.nom	do-prs.1sg	football-com.sg	[: par.sg]
	väga	palju
	very	much
	‘I do football a lot.’

25.	ma	tea-n	kõige	rohkem	nagu
	1sg.nom	know-prs.1sg	most	more	like
	inglis		keele-lt		[: keel-t]
	English.nom.sg		language-abl.sg		[: par.sg]
	‘I know, like, the English language the most.’

4.1.3. Object Marking: Discussion of Factors Affecting Usage

Divergences in object marking are estimated to occur in roughly 6% of uses of direct objects. Our findings show that the omission of object marking was prominent, occurring in 83% of the non-standard examples of direct object marking (predominantly with partial objects), wherein speakers used unmarked nominative forms (N = 184) or unmarked, other-language insertions (English and Swedish; N = 18). All the cases of omission are compatible with all the explanatory factors: frequency (nominative is more frequent overall than other cases), simplification (nominative is unmarked and, hence, simpler than other cases), and cross-linguistic influence (Swedish and English only use object case-marking with pronouns). The remaining 17% of the observations involved substitutions of standard case-marking with non-standard marking. Substitution occurred in favor of both more (N = 13) and less (N = 28) frequent, as well as more (N = 13) and less (N = 28) complex inflectional endings (with unmarked forms being less complex than marked ones, e.g., that involve stem changes). This finding shows that, when speakers opt for using inflection (instead of simplifying or omitting object marking), frequency plays a less significant role than morphological complexity. Overall, however, omission was much more often used than substitution, meaning that frequency or simplification are equally good explanatory candidates. This usage could also be the result of cross-linguistic influence, as seen when the observed examples (such as ex. 8–11) resemble Swedish and/or English, rather than standard Estonian constructions (particularly when partial objects were marked with nominative instead of partitive, e.g., Estonian: me mängime videomäng-ud, pro me mängime videomäng-e; English: we play videogame-s; Swedish: vi spelar videospel). Speakers displayed awareness of the Estonian morphological system, and the inflectional ending was not always omitted; they often opted for a more straightforward form (as seen with quantifier phrases). Although we found a variety of complex substitutions, these were infrequent and unevenly distributed. Therefore, they may reflect individual differences or production errors.

In addition to these input-related factors, we also looked at the effect of sociolinguistic background variables on the use of non-standard object marking. We ran chi-squared tests to measure the significance of the home language environment and the generation of speakers in the use of non-standard object marking. The results suggest that the speakers’ generational belonging influences the use of divergent forms (χ² = 14.016, df = 1, p-value < 0.001), although the association is very weak (Cramér’s V = 0.028). Pearson residuals revealed that divergently marked objects are more likely to be produced by speakers born in Sweden. This suggests that the different, and decreased, input to which second-generation speakers were exposed in their childhood influences their use of grammatical forms in their teenage years. In contrast, a chi-squared test showed no effect of the speaker’s home language environment on object marking (χ2 = 1.9343, df = 1, p = 0.1643).

Our findings align with previous studies of the language use of Swedish Estonians that evidence cross-linguistic influence and morphological simplification in object marking (Raag 1983, p. 42). Our data further confirm that object marking is one of the more complex constructions in Estonian for HL speakers to acquire (see also Maandi 1989; Laagus et al. 2004; Teral 2007). While object marking as a structure remains vulnerable, the overall proportion of divergences is not as prominent among the participants in the current study as in previous studies. The language use of today’s heritage speakers possibly reflects how opportunities for active HL maintenance have become wider than in the previous decades.

4.2. Non-Standard Experiencer Constructions

The Estonian experiencer construction of interest here, with the verb meeldima ‘to like/please’, consists of three interconnected elements: an Experiencer noun phrase, marked with adessive or allative case (similar in meaning to dative); a Stimulus, which can be a nominative noun or noun phrase, an infinitival verb phrase, or a relative clause; and a verb agreeing with the nominative Stimulus or bearing default third-person singular marking in the absence of a nominative subject (see also Section 4.2.1.). These are most frequently, but not necessarily, in the order of Experiencer-Verb-Stimulus. Substituting one of these with a different form may lead to changes in the other elements, as verb agreement is conditioned by nominative case; if the Experiencer is expressed with a nominative NP (as it is in Swedish), or if the Stimulus is not in nominative, this has implications for the rest of the construction. Figure 6 shows the proportions of standard and non-standard usage for the meeldima experiencer construction in the whole corpus.

From the corpus, we extracted 713 utterances with the verb meeldima, out of which 78 (11%) deviated in form from the standard construction. Verb agreement depends on the presence of a nominative NP. As the dataset includes both Experiencer and Stimulus participants expressed as nominative noun phrases, we can only look at verb marking in the context of participant marking. Hence, we analyze the marking of each participant argument (Experiencer, Stimulus) separately, together with verb inflection, as shown in

Table 6. Experiencer form in meeldima-constructions. Experiencer arguments were marked in the corpus with the standard adessive/allative (ade/all), the non-standard (*) nominative (nom), or genitive (gen) case, or were omitted from the construction (ø). Standard verb agreement in all cases means that the verb does not agree with the Experiencer, but instead either agrees with the Stimulus or bears default 3sg agreement.

Experiencer Form	Standard Verb Agreement	Non-Standard Verb Agreement	NA (Negative Clauses)	Total
ade/all	454	39	155	648 (90.9%)
* nom	3	3	3	9 (1.3%)
* gen	1	0	0	1 (0.1%)
ø	16	6	33	55 (7.7%)
Total	474 (66.5%)	48 (6.7%)	191 (26.8%)	713 (100%)

and

Table 7. Stimulus form in meeldima-constructions. When the Stimulus was a noun phrase, it was used with either standard nominative (nom) or non-standard (*) partitive (par) case, or as a code-switch (English or Swedish). The Stimulus also occurred as an infinitival verb phrase (inf), a relative clause, or was omitted from the construction (ø). Standard verb agreement in all cases means that the verb does not agree with the Experiencer, but instead agrees with the Stimulus or bears default 3sg agreement.

Stimulus Form	Standard Verb Agreement	Non-Standard Verb Agreement	ø (Negative Clauses)	Total
nom	241	35	108	384 (53.9%)
* par	12	4	10	26 (3.6%)
inf	108	2	12	122 (17.1%)
na	68	2	44	114 (16.0%)
Relative clause	27	0	12	39 (5.5%)
Code-switch (English/Swedish)	17	6	5	28 (3.9%)
Total	473 (66.3%)	49 (6.9%)	191 (26.8%)	713 (100%)

.

Table 5 and Table 6 give an overview of the use of inflection in utterances with meeldima. The form of the argument is in the left-hand column; the remaining columns display the use of standard or non-standard verb inflection and negation (in which case the verb does not bear person/number marking), along with total frequencies. From the tables, we can see that non-standard verb agreement and argument marking are relatively infrequent. In examples involving Experiencer arguments without negation (N = 522), 91% (N = 474) exhibit standard and 9% (N = 48) non-standard verb agreement usage. These proportions are the same with Stimulus arguments with no negation (N = 522), where standard verb agreement is more frequent (91% or 473 examples) than non-standard (9% or 49 examples). Experiencer and Stimulus arguments predominantly make use of the standard form: 648 (or 98%) of 658 nominal Experiencer arguments use the standard adessive/allative case, and 384 of 410 (or 94%) nominal Stimulus arguments use the standard nominative case.

4.2.1. Non-Standard Experiencer Marking

As shown in Table 6, the Experiencer argument (see also Figure 7) is marked in three ways: with the standard allative or adessive case10 (N = 648, or 91%), the non-standard nominative (N = 9, 1%), and one token of non-standard genitive (N = 1, 0.1%) case (overall, 1.4% of Experiencers made use of a non-standard case); when the construction had no overt Experiencer (N = 55, 8%), we determined the person and number of the Experiencer referent based on the surrounding context.

In three examples with a nominative Experiencer, there is no verb agreement, and the verb is in the default third-person singular form. This can be seen in examples (26) and (27), in which the Experiencer is in first- or second-person singular, respectively. In the case of third-person singular NPs, the use of third-person singular as a default form is indistinguishable from verb agreement. However, two examples have a plural Experiencer and a singular Stimulus, as in ex. (28), where the verb agrees with the Stimulus but not the Experiencer, despite the Experiencer argument’s preverbal position and adjacency to the verb.

26.	et	ma	[: mu-lle]	nagu	meeldi-b	ve-
	that	1sg.nom	[: 1sg-all]	like	like-prs.3sg
	kirjuta-da
	write-inf
	‘That I like like to write.’

27.	sa	pea-d	ise	lissalt	otsu-ma	[: otsusta-ma]
	2sg.nom	must-prs.2sg	self	just	decide-inf	decide-inf
	kui	sa	[: su-lle]		meeldi-b
	if	2sg.nom	[: 2sg-all]		like-prs.3sg
	‘You just have to decide yourself if you like.’

28.	et	(.)	ameeriklase-d	[: ameeriklas-tele]		meeldi-b	rohkem
	that		American-nom.pl	[: all.pl]		like-prs.3sg	more
	briti		kui	briti	või	briti
	British.gen.sg		than	British.gen.sg	or	British.gen.sg
	inglis		keel
	English.nom.sg		language.nom.sg
	‘That Americans like more British than British or British English.’

In three examples, the nominative Experiencer does trigger verb agreement, such as ex. (29), with a first-person singular pronominal Experiencer.

29.	rootsi	keel-t	ma	[: mu-lle]	ikkagi
	Swedish.gen.sg	language-par.sg	1sg.nom	[: 1sg-all]	still
	meeldi-n	[: meeldi-b]	kooli-s
	like-prs.1sg	[: prs.3sg]	school-ine.sg
	‘I still like the Swedish language at school.’

Furthermore, two utterances contain a nominative Experiencer used with a negative verb. As in standard Estonian, there is generally no agreement in these examples (see ex. 30), the verb having a bare form with the negative particle ei ‘no’. In ex. (30), both arguments are coded for a canonical transitive clause, with a nominative Experiencer and partitive Stimulus. The adverbial placement between the negative particle and the lexical verb is unusual in Estonian and may evidence cross-linguistic influence, as adverbs follow negation in both Swedish and English.

30.	et	se-da	ma	[: mu-lle]	ei	tõesti
	so	3sg-par	1sg.nom	[1sg-all]	neg	really
	meeldi		sellepärast	nagu
	like.cneg		because	like
	‘So, that I do not really like because like.’

In two examples, the speakers make use of an unusually marked, complex combination, with a non-standard inflected verb following the negation particle. In ex. (31), the verb is marked for third-person plural (possibly due to the surrounding context), while the Experiencer is third-person singular; in ex. (32), the negative verb is overtly marked for third-person singular, though, in standard Estonian, the negative construction uses an unmarked verb stem. In (32), the parallel coordinated verbs (‘what I like and dislike’) also show the overuse of inflection in a negative context.

31.	et	kui	ta	ei	meeldi-vad	[: ta-lle	ei	meeldi]
	that	if	2sg.nom	neg	like-prs.3pl	[: 3sg-all	neg	cneg]
	siis
	then
	‘That if he/she does not like (them) then.’

32.	see		=	ei	ole	nagu	mis	mu-lle	meeldi-b
	3sg.nom			neg	be.cneg	like	what	1sg-all	like-prs.3sg
		ja	ei	meeldi-b		[: meeldi]
		and	neg	like-prs.3sg		[: cneg]
		‘It is not like what I like and dislike.’

Finally, we have one deviation in which the Experiencer is marked with the genitive case. As with standard adessive/allative Experiencers, in this example (33), the genitive Experiencer is followed by a third-person singular (past-tense) verb agreeing with the nominative Stimulus subject. The genitive marking in this example may simply be a result of a production error or rapid speech, as the forms of the standard adessive (mul) or allative (mulle) have the genitive as their stem.

33.	nagu	mu	[: mu-lle]	meeldi-s	see
	like	1sg.gen	[: 1sg-all]	like-pst.3sg	3sg.nom
	‘Like I liked it.’

4.2.2. Non-Standard Stimulus Marking

As shown in Table 7, if the Stimulus (see also Figure 8) is a noun or noun phrase (N = 410), 94% (N = 384) of these occur in nominative form, while 6% (N = 26) are marked with non-standard partitive (among the latter, only two examples co-occurred with the non-standard nominative Experiencer). The Stimulus also occurred as an infinitive verb phrase (N = 122), a relative clause (N = 39), or a code-switched (13 English and 15 Swedish) noun or infinitive verb, all of which are standard forms. In 114 utterances, the Stimulus was absent from the construction.

In the context of verb agreement, instances where the verb and Stimulus did not agree in person and number were considered examples of non-standard usage. In nearly all the non-standard uses, the third-person singular form of the verb meeldima (i.e., meeldib) acted as the default form. While this is the expected form to be used with third-person singular Stimuli, as well as nonfinite verb complements, other person-number categories require different verb marking. In a total of 110 utterances, the Stimulus was in third-person plural or first-person singular11 and, hence, required marking other than the third-person singular. Of these, speakers used non-standard verb agreement in 28 examples, using the third-person singular form instead. For instance, we can see this pattern with nominative plural Stimuli, as in examples (34) and (35).

34.	mu-lle	mm	meeldi-b	[: meeldi-vad]	ainult
	1sg-all	int	like-prs.3sg	[: prs.3pl]	only
	põhimõtselt		draakonikommi-d
	basically		dragon candy-nom.pl
	‘I, hmm, basically only like dragon candy.’

35.	mu-lle	meeldi-b	[: meeldi-vad]	kõik	filmi-d
	1sg-all	like-prs.3sg	[: prs.3pl]	all	movie-nom.pl
	‘I like all movies.’

Interestingly, there was a consistent lack of agreement between the verb and plural Stimulus in two specific contexts. The default third-person singular verb form was used when the Stimulus involved a code-switch in plural form (N = 5), as in ex. (36), or was marked with partitive plural (N = 3), as in ex. (37); such constructions were used by different speakers. In the case of partitive plural, this construction can be seen as being more consistent with standard Estonian morphosyntax, since a partitive plural noun does not trigger verb agreement in any context, yet, in this construction, the standard form would have a nominative plural Stimulus and a plural verb.

36.	mu-lle	meeldi-b	[: meeldi-vad]	{dumpling-s}
	1sg-all	like-prs.3sg	[: prs.3pl]	dumpling-pl
	‘I like dumplings.’

37.	mu-l	meeldi-b	palju	inimes-i
	1sg-ade	like-prs.3sg	many	person-par.pl
	[: meeldi-vad	palju-d	inimese-d]
	[: prs.3pl	nom.pl	nom.pl]
	‘I like a lot of people.’

We also identified a total of 11 utterances with the indefinite pronoun kõik ‘everything/everyone’ as the Stimulus. Despite the pronoun having an identical form in singular and plural, it is usually used in standard Estonian with a plural verb. We found that, with only one exception, speakers used a singular verb with this pronoun, as shown in ex. (38):

38.	mu-lle	meeldi-b	[: meeldi-vad]	kõik
	1sg-all	like-prs.3sg	[: prs.3pl]	everyone
	nagu	.hhh (.)	Hovet	Adel
	like		Hovet.ø	Adel.ø
	‘I like everyone, like Hovet, Adel [band/musician names].’

While third-person singular verb marking is highly frequent in all the non-standard utterances, we also find six examples where the non-standard agreement involves a third- person plural verb. These are all in contexts of semantic plurality, e.g., the Stimulus is an infinitival verb phrase containing a plural object (as in ex. 39) or a nominative singular collective noun (ex. 40, jalgpall ‘football’ seems to refer to various activities related to football), which are both contexts where standard verb agreement would typically be used.

39.	kas	su-lle	meeldi-vad	[: meeldi-b]	kingitus-i	saa-da
	q	2sg-all	like-prs.3pl	[: prs.3sg]	gift-par.pl	get-inf
	‘Do you like receiving gifts?’

40.	sellepärast	mu-lle	meeldi-vad	[: meeldi-b]	jalgpall
	because	1sg-all	like-prs.3pl	[: prs.3sg]	football.nom.sg
	‘Because I like football.’

In 26 utterances, speakers marked the Stimulus with the non-standard partitive case. In half of these examples, the partitive Stimulus is the demonstrative pronoun seda ‘this’, (ex. 41, N = 6) or relative pronoun mida ‘what’ (ex. 42, N = 7):

41.	mu-lle	ei	meeldi-nud	se-da	[: see]	väga
	1sg-all	neg	like-app	3sg-par	[: 3sg.nom]	very
	‘I did not like this very much.’

42.	ma	ei	tea	tegelikult	mi-da	[: mis]
	1sg.nom	neg	know.cneg	actually	what-par.sg	[: nom.sg]
	mu-lle		Eesti-s		ei	meeldi
	all-1sg		Estonia-ine.sg		neg	like.cneg
	‘I do not know actually what I do not like in Estonia.’

About one-third of the constructions containing a partitive Stimulus occur in negative contexts, which makes the question of verb agreement irrelevant, as evidenced by ex. (41)–(44). Here, the partitive may also be conditioned by analogy with other constructions in Estonian where negation triggers the partitive case (as with transitive objects and existential/presentational subjects; see (Lindström 2017, p. 192)).

43.	mu-lle	ei	meeldi	ver-d	[: veri]	ka
	1sg-all	neg	like.cneg	blood-par.sg	[: nom.sg]	also
	‘I also do not like blood.’

Six of the examples involving non-standard verb agreement have Stimuli from English and Swedish. Most of these bear plural marking from the source language and are preceded by a singular verb (ex. 44). In one example (45), however, a singular Stimulus is used with a plural verb.

44.	mu-lle	nii	meeldi-b	[: meeldi-vad]	see	[: nee-d]
	1sg-all	so	like-prs.3sg	[: prs.3pl]	3sg.nom	[: 3pl-nom]
	nagu	{kötbull-ar}	[: {kötbull-ari-d}]	koos	{potatismos}
	like	meatball-pl	meatball-pl-nom.pl	with	mashed potatoes.ø
	[: {potatismosi-ga}]
	[: com.sg]
	‘I really like it, like meatballs with mashed potatoes.’

45.	kui	ma	ole-n	jõusaali-s	siis
	if	1sg.nom	be-prs.1sg	gym-ine.sg	then
	mu-lle	meeldi-vad	[: meeldi-b]	{hardstyle}
	1sg-all	like-prs.3pl	[: prs.3sg]	hardstyle.ø
	‘If I’m at the gym, then I like hardstyle (music).’

4.2.3. Experiencer Constructions: Discussion of Factors Affecting Usage

We found that 11% of meeldima-constructions involved non-standard usage; it is worth highlighting that, while object marking may vary in Estonian youth language usage, divergences involving ‘like’ constructions are extremely rare. We observe divergences in both Experiencer and Stimulus marking as well as verb agreement. Experiencers occurred in the standard allative/adessive, as well as the non-standard nominative and genitive case, or were omitted. Stimuli predominantly occurred as noun phrases marked with either the standard nominative or non-standard partitive case. Stimuli also occurred as infinitival verb phrases, relative clauses, code-switched noun phrases, or were omitted. Third-person singular is the most frequently used verb form in non-standard usage. This finding is compatible with two interpretations, as either marking verb agreement or the default form. These interpretations can be distinguished in cases where the Stimulus argument is not third-person singular and so an agreeing verb would take a different form; however, most of the Stimulus arguments are third-person singular. Among cases where the Stimulus was not third-person singular (N = 110), and, so, the default form interpretation was unavailable, we found a relatively high rate (26%) of non-agreement, indicating that the verb form may often be used as a default. Non-standard marking was also found with plural marking, if rarely, e.g., when the infinitival verb phrase contains a plural object. The verb form’s frequency might also be related to cross-linguistic influence, since, in Swedish and English, the number and person of the Stimulus do not influence the verb form.

The nominative case is more frequent and morphologically simpler than the standard Experiencer-marking adessive/allative cases, but it is also analogous to the Swedish and English transitive experiencer constructions. Of the non-standard deviations, 11 involve the use of less frequent and more complex forms than the standard (e.g., partitive used instead of nominative, plural marking on verb instead of singular). Factors driving this pattern will be discussed in more detail in Section 5.

When exploring the effect of sociolinguistic background variables on the use of non-standard meeldima-constructions, the chi-squared test showed that generational belonging did not play a statistically significant role in the use of divergent constructions (χ2 = 0.67214, df = 1, p = 0.4123). On the other hand, the speaker’s home language environment suggests significance (χ2 = 3.8472, df = 1, p = 0.04983; just at the significance threshold of p = 0.05) and indicates a possible connection between a speaker’s linguistic environment at home and the non-standard use of the meeldima-construction. Pearson’s residuals showed that speakers from multilingual families are more likely to use non-standard constructions; however, Cramér’s V (= 0.073) indicates that the association between the observed variables is very weak. These findings may suggest that more multilingual linguistic input can lead to higher variation in using experiencer constructions. However, other factors not observed in this study may also play a role, for example, whether the speaker has attended an Estonian-medium school or had private lessons of Estonian, independently of whether their home is monolingual Estonian or multilingual.

5. Discussion and Conclusions

In this study, we investigated the use of non-standard morphosyntax in the speech of adolescent Estonians in Sweden, addressing the following research questions: (1) how do speakers make use of non-standard inflection in object marking and experiencer ‘like’ constructions, (2) does the speakers’ use of these constructions evidence cross-linguistic influence, simplification, or the effects of input frequency, and (3) how do the speakers’ home language environment and generational belonging explain the use of divergent forms? Although the proportion of instances of non-standard usage is relatively low in the corpus as a whole (occurring in around 5% of utterances), it is important to investigate the details of divergent usage, as low overall rates of divergence may conceal consistent patterns of divergence in particular contexts (e.g., Aguado-Orea and Pine 2015). Certain constructions are more vulnerable to deviation, and these may signal the effects of both decreased input and language contact. We focused on two structures—object marking and experiencer meeldima ‘to like/please’ constructions—with the aim of investigating factors driving the use of divergent forms.

With respect to the first and second research questions, we conducted separate ana-lyses on the object marking and experiencer constructions. We identified 243 instances of non-standard object marking and estimated that they represent approximately 6% of all transitive direct objects in the corpus. In the dataset of utterances containing non-standard nominal morphology (N = 672), direct object marking was the most frequent (36%) phenomenon. This finding corroborates previous studies (e.g., Torn 2003; Teral 2007; Raag 2010), which have also reported that heritage Estonian speakers struggle particularly with object case-marking, and that they tend to simplify this construction. We grouped object marking divergences into two patterns. The vast majority (83%) of cases constituted the omission of marking, meaning that speakers used unmarked or nominative forms instead of the expected partitive or genitive case (moreover, divergent object marking occurred nearly four times more with partial than total objects). In less than a fifth of the examples of non-standard object marking (17%), speakers substituted the expected case with another overtly marked case, often in favor of less frequent, more complex cases (such as partitive instead of nominative or genitive). Code-switched objects involved omission rather than substitution, which is unexpected in the light of previous studies exploring the morphological integration of code-switched elements in Estonian (Igav 2013; Kask 2021). However, this is not a general trend for how the speakers in this study integrate non-Estonian insertions. Although we were unable to outline how code-switched objects with non-standard marking compare with code-switches using standard marking, a recent analysis (Korkus 2024) found that teenage HL participants resemble Estonian teenagers (as analyzed in Vihman et al. n.d.) in generally integrating code-switched nouns. Nevertheless, Estonian teens are significantly more likely to do so than Swedish Estonian teens.

Regarding the second research question, simplification plays a major role overall in object marking (with speakers favoring bare forms over inflected ones). This may result from frequency and the complexity of the standard form, as well as the effects of language contact. The prevalence of object marking omission reflects the speakers’ preference for a less marked and morphologically simpler form (likely due to cross-linguistic influence, as objects are typically unmarked in both Swedish and English). Unmarked nominative forms are the most common, despite being less frequent in object marking contexts (importantly, however, the nominative case is used for marking objects, and nominative objects were found to be more frequent than genitive objects in spoken Estonian). It is interesting to note that a relatively high proportion of examples (57 out of 202) involved plural objects in the nominative case. This finding indicates sensitivity to the Estonian system, in which plural total objects do occur in nominative case; hence nominative is used proportionally more with plural than singular objects. This usage may lead to the participants’ overuse of unmarked plural objects (here we may interpret overuse as another form of simplification). Nevertheless, total objects (marked with nominative or genitive case) tend to make up a smaller proportion of object marking than partitive objects in standard Estonian. We also found that, when speakers opt to substitute the standard object case with another form, morphological complexity might play a more significant role than frequency. The single most frequent substitution (in a quarter of overt object case substitutions) is that of genitive instead of partitive case-marking: this is often a simpler case to form (partitive is much more irregular than genitive), even though genitive is far less frequent as an object case marker.

From a total of 713 utterances with the meeldima-construction, 78 (11%) included non-standard usage in the marking of one of the arguments and/or verb agreement. Experiencer arguments are mostly in the standard allative/adessive case, with 1.4% in non-standard nominative and/or genitive case. In about half of the examples containing a non-standard Experiencer, the verb agrees with the Experiencer, diverging from the standard, where the verb agrees with the Stimulus. The Stimulus occurs mostly in standard forms (nominative subject, infinitival verb phrase, or relative clause), but 6% of nominal Stimuli were in non-standard partitive case, suggesting speakers struggled more with marking the Stimulus than the Experiencer. Regarding Stimulus-verb agreement, we found that the third-person singular verb form was used in both expected contexts and contexts requiring other forms (e.g., with third-person plural Stimuli), indicating a simplification strategy. Lack of agreement does not always involve the third-person singular form; in a small number of examples, plural marking is used.

Returning to the second research question, experiencer constructions reflect more prominent effects of likely language contact than object marking. Speakers make various adjustments to the non-canonical Experiencer construction, most of which reflect the transitive argument structure in the equivalent Experiencer constructions in Swedish and English. They simplify the verb form, suggesting influence especially from the Swedish construction, in which the verb ending remains unchanged across all person–number categories. Non-standard Experiencers do not occur frequently, but the use of nominative Experiencers again suggests cross-linguistic influence. The same can be said about the partitive-marked (object-like) Stimulus: although Swedish does not use this case, it has a rigid word order in transitive constructions—including the ‘like’ construction—which may lead speakers to treat these postverbal constituents as objects. They generally lean toward the use of more frequent and morphologically less complex forms. In the case of substitution with a less frequent form (such as using a partitive Stimulus in place of the frequent, unmarked nominative, or marking the verb plural instead of the frequent, unmarked singular), we find language contact effects again to be the most likely explanation. Constructions with great structural differences from the societal language are particularly vulnerable to change, even when they are in frequent use. For heritage speakers, they may take more time to acquire and do not achieve the same level of entrenchment as constructions with parallels across the two languages.

The second question in this study sought to determine the effects of cross-linguistic influence, simplification, and input frequency. Overall, our findings suggest that all three factors (language contact, simplification, and frequency) influence the use of non-standard morphosyntactic forms. The observed factors often occur together, each contributing to a tendency to omit marking or use default marking. This is evidenced by how the most frequent (and default) forms (e.g., nominative singular nouns or third-person singular verbs) are generally morphologically less complex than the alternatives. Furthermore, both contact languages (Swedish and English) have simpler morphology than Estonian has, particularly in the two constructions under consideration. Regarding cross-linguistic influence, the effects of English seem to be found more on the lexical than the morphosyntactic level, while Swedish influence is present on both levels: this distinction is largely due to how English exposure and acquisition contrasts with Swedish, in terms of age of exposure, the context and trajectory of acquisition, as well as everyday usage contexts. This is evidenced by the fact that, unlike Swedish and Estonian, English is acquired mostly via technological channels (e.g., TV, social media) and not necessarily from native English speakers. Since most participants were born in Sweden, it is likely they began acquiring Swedish much earlier than English. Moreover, on a lexical level, English insertions are considerably more frequent (53.7% of all code-switching instances) than Swedish insertions (45.0%) (Korkus 2024).

Our quantitative analysis addressed the third research question by revealing how the speaker’s generation and home language environment explains non-standard use of object marking and experiencer constructions. We found a statistically significant, though weak, association between divergent object marking and the speaker’s generational belonging: the results indicate that speakers born in Sweden are more likely to produce divergently marked objects. This finding highlights the role of Estonian input in acquiring object marking constructions, noting that the linguistic input Swedish-born speakers grow up with is significantly different (possibly both quantitatively and qualitatively) from the input of Estonian-born speakers who were initially raised in an environment where Estonian was the dominant language. Moreover, second-generation speakers have been exposed to Swedish for longer than the first generation, thus being subject to more cross-linguistic influence (especially from Swedish). This finding also aligns with Montrul and Bateman (2020), who suggest that simultaneous (second-generation) bilinguals are more vulnerable than sequential (first-generation) bilinguals.

A weak connection was also found between the speakers’ home language environment and the use of non-standard experiencer constructions: those from multilingual homes were more prone to divergent usage. This result may reflect influence from other languages spoken at home, as well as an effect of reduced Estonian input. From the available data about the participants’ language history, we are unable to disentangle the possibly different effects of input quantity, quality, and timing, although we believe all three to have some effect on their divergent usage. In multilingual families, Estonian is acquired alongside another language; therefore, certain less frequent and more complex constructions may be acquired later than their non-Estonian counterparts. Additionally, the speakers from multilingual households (all of whom are, in fact, Swedish-born) may be exposed to reduced or less diverse Estonian input. Thus, individual differences play a substantial role. If this were the case, it would be worth looking into other experiencer constructions to compare usage across other verbs and across languages. Unlike the transitive structure, the meeldima-construction is much more specific (lexically specified) and less frequent, in general language use (e.g., in the Phonetic Corpus of Estonian Spontaneous Speech, it is the 42nd most used verb; see Lippus 2019); hence, it takes longer to acquire and become entrenched for speakers (thus also being morphologically complex; see Rinke et al. 2024). Based on data from the longitudinal Vija and Zupping child language corpora available on CHILDES (MacWhinney 2000; Vija 2004; Zupping 2015), children may begin to use this construction productively as late as three years old. However, in interpreting tests of significance, we must take into consideration that the corpus is unbalanced and limited in size; yet, despite the small sample size, these findings are suggestive for the variable effects of sociolinguistic background factors.Adolescence is a critical time in shaping speakers’ language usage, when young people become more embedded in their social environment and the influence of the family language may weaken. The examination of which constructions are subject to divergent usage, either driven by cross-linguistic influence or heritage language input, can provide insight into heritage language development both individually and societally. Although generally teenagers are noted for being linguistic innovators who diverge from their parental language and have thus been referred to as the ‘drivers of change’ (Eckert 1997), we do not find linguistic innovation to be a prominent feature of young heritage speakers. Instead, our findings indicate similar usage patterns among the Swedish Estonian teenagers in our study and earlier generations of Swedish Estonians (most of whom had reached adulthood when being recruited for research). While previous studies on heritage Estonian speakers suggest younger generations are more prone to divergences, our analysis finds that, overall, non-standard usage is relatively infrequent (though systematic); usage patterns suggest which constructions might be susceptible to a potential language change. Such stability may be attributed to the increased HL maintenance opportunities (particularly through the internet) available to today’s youth compared to earlier generations. Additionally, it is important to note that most of the participants of this study used Estonian outside of their home. These teenage HL speakers have nearly ideal circumstances for language maintenance: they have frequent contact with multiple HL speakers and easy access to a country where it is spoken as the majority language.

This study is the first to investigate present-day Swedish Estonian teenagers’ HL usage. Future research should follow up this study with a more detailed analysis in two directions: the small sample here is suggestive but calls for a fuller sociolinguistic investigation, and the HL usage investigated here also calls for a closer look at the linguistic and discourse context of non-standard examples. Speakers’ individual differences and idiolects also need to be probed more thoroughly. Traditionally, adult speakers’ language has been used as the point of comparison, but taking adolescent speech from the country of origin (Estonian teenagers) as the baseline may provide a more informative framing for data comparison. This study aligns with the literature in terms of what factors drive non-standard usage. Even when divergences are not highly prevalent, constructions that are more vulnerable show systematic signs of simplification and language contact in adolescent heritage speaker usage, regardless of their generation or home language environment.