A Comparative Perspective on Language Shift and Language Change: Norwegian and German Heritage Varieties in North America

Abstract

This study evaluates the relationship between language shift and linguistic change in multigenerational immigrant communities, focusing on North American Norwegian (NAmNo) and German heritage varieties. The research synthesizes current findings on how language shift impacts linguistic structures in moribund heritage varieties. Methods include a qualitative review of diachronic studies, comparing data from different periods to assess changes in tense morphology, language mixing, compositional definiteness, possessive placement, verb placement, argument placement, and phoneme variation. Results indicate that the last generation of heritage speakers demonstrates increased linguistic innovation and variation compared to earlier generations. Key findings show that language shift leads to different input quality and quantity, affecting grammatical stability. The study concludes that sociocultural changes, such as verticalization and domain-specific language use, significantly influence heritage language maintenance and loss. These insights contribute to understanding the dynamics of language shift and its role in heritage language change, offering valuable comparative perspectives across different immigrant communities.

1. Introduction

Research on heritage language grammars has established that language change can emerge even in the first generation of heritage speakers (Montrul, 2016; Polinsky, 2018).1 However, recent diachronic studies on multigenerational heritage varieties indicate that the trajectory of change differs across generations of North American Norwegian (NAmNo) and German heritage varieties. In particular, the last generation of speakers demonstrates increased linguistic innovation, variation, and change across multiple variables compared with earlier generations (Salmons, 1994; Johannessen, 2018; van Baal, 2022; Lykke, 2022; Eide & Hjelde, 2023).2 This paper provides an evaluation of the relationship in multigenerational immigrant communities between language shift and linguistic change, by providing a novel synthesis of the current state of research. We focus especially on the claim that language shift is an underlying cause of subsequent linguistic change in moribund heritage varieties, which has been made for NAmNo (van Baal, 2022; Lykke, 2022; Eide & Hjelde, 2023; see also Johannessen, 2018).

The need for a review arises because structurally oriented work makes claims about the correlation between language shift and language change but often lack the sociolinguistic component and/or comparative empirical scope needed to credibly make such claims (e.g., Lykke, 2022). The argued key explanatory factor is the shift in sociolinguistic context over time, with a significant reduction in domains where the heritage languages were spoken. Eide and Hjelde (2023) operationalize the effects of language shift in terms of input and output in different social domains, arguing that these shifts impact linguistic structures to varying degrees.3 Their argument aligns with broader discussions in heritage language research (see Benmamoun et al., 2013; Polinsky & Scontras, 2020). The availability of diachronic data for NAmNo and German varieties in the USA offers a rare opportunity to compare such claims about the effect of language shift upon linguistic change for two multigenerational immigrant communities.

There are many sociohistorical commonalities between NAmNo-speaking communities and the German-speaking communities we discuss.4 Because of the large numbers of these immigrants, many communities initially remained inward-looking and relatively independent of the English-speaking majority or other immigrant groups (Munch, 1949; Salmons, 1983; Johannessen & Salmons, 2015; Johannessen, 2018, pp. 449–450; Bousquette, 2022). Institutions were organized by and for community members, and the use of the immigrant language was maintained across different domains for multiple generations. However, with the onset of a larger sociocultural change in the early 20th century, the local communities became more connected to the majority society. Power over local institutions was (or had to be) transferred to larger state or national governments and such societal change is a crucial factor for language shift, according to verticalization analyses (Salmons, 1983, 2005; Brown, 2022b). Numerous recent studies (e.g., Johnson, 2018, 2022; Natvig, 2022a; Bousquette, 2022) have detailed how such verticalization and concomitant, widespread sociocultural changes led to language shift within immigrant communities. Language shift results in certain domains (typically education, church, and media) swiftly being occupied by the majority language, followed by a slower shift in the community and home language domains. Although the idea that language shift to the majority language and language change in the heritage language go hand in hand is not new (Boas, 2002, 2009; Eide & Hjelde, 2023; Lykke, 2022; Rocker, 2022; van Baal, 2022), to our knowledge, there has been no attempt to integrate sociolinguistic studies from a verticalization perspective with studies exploring linguistic change using diachronic data. Communities included in this paper were chosen based on the availability of studies analyzing their sociocultural development from a verticalization perspective exploring language shift and studies using diachronic data analyzing language change from the same community.

To frame our discussion, we adopt Bousquette’s (2020) distinction between preshift multilinguals and early shift bilinguals. Preshift multilinguals acquired and used the heritage language across multiple societal domains, whereas early shift bilinguals primarily spoke the heritage language at home or within a restricted community context. Early shift bilinguals constitute the last generation of proficient speakers and can be classified as heritage speakers in the narrow sense (cf. Polinsky, 2018, pp. 4–9). These speakers acquired the majority language early—either simultaneously or sequentially—and eventually became dominant in it. They also lacked dense social networks of fellow heritage speakers, leading to decreasing heritage language use over time. This shift distinguishes early shift bilinguals from preshift (multilingual) speakers.

Although previous research, as mentioned, has emphasized the role of societal factors in heritage language change, no comprehensive cross-linguistic review exists. To evaluate the role of language shift in linguistic change in these communities, we discuss all existing diachronic studies of NAmNo and three American German varieties. The varieties are selected because of the availability of both historical sociolinguistic and diachronic linguistic studies. NAmNo has been extensively documented (e.g., Johannessen, 2018; Eide & Hjelde, 2023), and as Eide and Hjelde (2023) note, it presents a unique opportunity for understanding multigenerational heritage varieties due to its well-documented linguistic and immigrant community histories. Studies of American German varieties (e.g., Texas German, Wisconsin German, and Iowa Low German), though less comprehensive, offer valuable comparative insights. By comparing these varieties, we evaluate whether similar societal developments have similar effects on language change in different varieties. The joint discussion enables a novel evaluation of whether the trends are generalizable beyond heritage Scandinavian or a single variety or speaker community. Furthermore, by introducing diachronic data on language change and sociolinguistic studies on verticalization and language shift (see Section 2), we further elucidate the role that language use in different societal domains plays in linguistic change in heritage varieties. These findings on change are additionally relevant to the methodological discussion about establishing the baseline in the study of multi-generational heritage varieties (see D’Alessandro et al., 2021, pp. 4–5; Kinn & Larsson, 2022; Lykke, 2020, pp. 100–101; van Baal, 2022, for discussion).

This article is a qualitative review, synthesizing findings from existing studies under a unified approach. The remainder of the paper is structured as follows: Section 2 outlines the sociohistorical backgrounds of the Norwegian and German communities under study, considering language shift through the lens of the verticalization model. Section 3 presents findings from diachronic studies on language change among heritage speakers. Section 4 discusses the implications of these findings, linking them to general explanations of heritage language change. Finally, Section 5 concludes the article.

2. Sociolinguistic History and Verticalization

2.1. Theoretical Framework: Verticalization and Language Shift by Domain

The verticalization approach (Brown, 2022b; Salmons, 1983, 2005) is used to describe sociocultural change in immigrant communities and to explain language shift starting in the early 20th century. The approach is based on sociological work by Warren (1963, p. 53) who found that small, rural communities in the USA showed intense sociocultural change in the early to mid-20th century due to the diversification of community interests, urbanization, the transfer of authority to non-local groups and government, and growing connections to the majority society. Adapted to minority language groups, these profound sociocultural changes were found to impact language use (see Lucht et al., 2011; Frey, 2013). In short, it seems that minority languages are maintained as the communal language if the community is more autonomous and in charge of their own institutions. Once local control over institutions is transferred to state or federal agencies and community members are increasingly looking for employment outside the community, gradual language shift to the majority language follows in the public domains for these communities (Bousquette, 2022; Brown, 2022b; Natvig, 2022a; Salmons, 1983, 2005).

The verticalization model has been tested with multiple different communities in the North American context, including Wisconsin German (Lucht et al., 2011; Frey, 2013; Bousquette, 2022), Wisconsin West Frisian (Bousquette & Ehresmann, 2010; Bousquette, 2022), American Norwegian (Moquin, 2025; Natvig, 2022a), American Finnish (Johnson, 2018, 2022), Pennsylvania Dutch (Brown, 2022a),5 and Iowa Low German (Rocker, 2022). While these studies focus on different institutions and areas of social change in the respective community, there is agreement that sociocultural and economic change are the main factors leading to language shift to the majority language. Language use in a minority language community is typically affected by language use in school education, church service and congregational life, and written media (especially newspapers). Similarly, community members’ occupations or life modes impact language use: as long as most community members work on their own farms or within their own households, maintenance of the minority language is more likely (Bousquette, 2022, pp. 83–84; Natvig, 2022a, p. 113). Importantly, even if a standard variety or the majority language is used in certain domains (e.g., education or church), the heritage language is still the dominant language used as the spoken community language and within the home. Bousquette refers to this stage as “preshift multilingualism” (Bousquette, 2020, p. 514), indicating the use of the majority language (English), (possibly) a standard variety of the immigrant language, and a vernacular heritage variety.

However, due to sociocultural changes such as the introduction of larger school districts and non-locally oriented church congregations, using the heritage language in institutions or social gatherings became less common throughout the 20th century. Concurrently, community members were increasingly employed in positions outside of the farm or the community. Together, these factors lead to a widespread shift to the majority language (Bousquette, 2022; Natvig, 2022a). Simultaneously, accepting employment outside of the community changes the formerly tight-knit, multiplex communal ties in favor of a wider system of loosely connected networks, which may be more conducive to language change (Bousquette, 2022, p. 83; J. Milroy & Milroy, 1985; L. Milroy & Milroy, 1992). Thus, since the community itself is changing and many of the domains that previously supported the use of the heritage language now are increasingly occupied by the majority language, the use of the heritage language slowly retreats to the home domain. Bousquette refers to this stage as “early shift bilingualism” (Bousquette, 2020, p. 514), showing that speakers who grew up in this changing community often learned the heritage language at home, but that the heritage language was more rarely spoken in the community, and that speakers no longer learned a standard variety of their heritage language. This makes early-shift bilinguals heritage speakers in the narrow sense, i.e., unbalanced minority-language bilinguals as described by e.g., Polinsky (2018, pp. 4–9).

Therefore, due to communal changes, the last generation of active speakers notably differs in the quality and quantity of input they received in different domains during language acquisition as children. So far, the verticalization approach has been used to describe sociocultural changes in minority language communities leading to language shift. It has not directly been used to describe language change, which we will attempt by incorporating studies using diachronic data from the communities under investigation. In the following sections, more detailed accounts of the sociolinguistic history and language use of the Norwegian and German communities in the USA are laid out based on comparable language domains (church, school, newspapers, and workplace) followed by a summary of the main trends.

2.2. Establishment of Communities and Language-Bearing Institutions

Norwegian and German have long existed as heritage languages in the United States, brought by large waves of immigration between 1820 and 1930 (Haugen, 1953, pp. 23–33; Jacob, 2002, p. 37; Lovoll, 1999, pp. 23–38; Luebke, 1990, p. 95).6

North American Norwegian (NAmNo) emerged as Norwegian immigrants founded communities and established language-bearing institutions such as schools, the Norwegian Lutheran Church, and a Norwegian-language immigrant press (Haugen, 1953, p. 101; Lovoll, 1999, pp. 74, 98–102, 143–189, 328). Some Norwegian settlements, such as those in Vernon County, Wisconsin, were to a large degree socially and economically independent from the English-speaking society until the 1940s (Munch, 1949, pp. 782–784). Although in-depth studies do not exist for all NAmNo communities, Natvig’s (2022a) detailed study of verticalization processes leading to language shift from Norwegian to English in Ulen, Minnesota, may serve as an approximation for the developments in other communities. Most Norwegian-speaking communities in North America followed a trajectory similar to Ulen and shifted to English during the 1930s and 1940s. The current generation of speakers are, with only a few exceptions, born in the 1930s and 1940s. Most domains were already shifting to English during that time, and the current generation makes up the last generation of NAmNo speakers.7

Similarly, German-speaking communities flourished in the mid- to late 19th century as approximately 5.8 million German immigrants settled in the U.S., particularly in rural areas and in the Midwest where they often outnumbered other ethnic groups (Dolmetsch, 1976, pp. 190–191; Jacob, 2002, p. 37; Luebke, 1990, p. 95; Schwartzkopff, 1987, p. 17). Importantly, these settlers originated from a large area in Central Europe and were culturally and linguistically far from homogeneous (see Putnam & Salmons, 2015, p. 29), even though they may appear in US Census data as “German”. Thus, their dialects were not necessarily mutually intelligible, which often encouraged settlers to form communities with other immigrants from the same place of origin (Langer, 2008, p. 501).8 In those cases, the spoken vernacular often co-existed with a more standard-like variety and the majority language (Nützel & Salmons, 2011). In other cases, when many groups of immigrants speaking different German varieties lived in the same community, cases of dialect levelling and koineization have been observed (Drake & Kramer, 2014; Lindemann, 2019; Louden, 2011; Nützel & Salmons, 2011). Studies of language shift in communities such as Texas German (Boas, 2002, 2009; Boas & Fingerhuth, 2017; Salmons, 1983), Wisconsin German (Bousquette, 2020, 2022; Frey, 2013; Louden, 2011; Lucht et al., 2011; Nützel & Salmons, 2011; Wilkerson & Salmons, 2008), and Iowa Low German (Kehlenbeck, 1948; Mertens, 1994; Rocker, 2022; Saathoff, 1930; Webber, 1998) have provided insights into sociolinguistic developments, including verticalization and diachronic language change, offering a comparative perspective on heritage language maintenance and loss.

2.3. Language Use and Shift in the Church

Norwegian immigrants in the 1800s established a robust Norwegian Lutheran Church presence in the USA. This church served not only a spiritual function, but also as a central institution for preserving the Norwegian language and cultural traditions (Haugen, 1953, p. 101; Lovoll, 1999, pp. 74, 98–102, 143–189, 328). The church was an important institution for the Norwegian communities (Lovoll, 1999, pp. 143–170), which was linked to Norwegian-American group identity and, consequently, language preservation (Lovoll, 1999, p. 100). However, organizational shifts, including the merger of Lutheran synods in the 1920s and growing integration into American Lutheran congregations, weakened local control and accelerated language shift to English (Natvig, 2022a). Nationally, the use of Norwegian in services in the US declined from 73% in 1918 to 50% in 1925. Additionally, church records switched to English in 1926. The church provided religious education and, importantly, literacy in Norwegian, but this ended in the 1930s (Lovoll, 1999, pp. 327–328). By the mid-1940s, a mere 7% of services were held in Norwegian, and in 1946, the church removed “Norwegian” from its name, marking the completion of the linguistic shift (Haugen, 1953; Lovoll, 1999; Natvig, 2022a).

German churches followed a similar trajectory. German Catholic and Lutheran congregations initially maintained the use of German through religious instruction, hymns, and social interactions (Dippold, 2002; Goldberg, 1990; Saathoff, 1930).9 In the beginning, Catholic churches supported language retention by operating German-language schools and publishing religious materials, but already in the 1890s, the American Catholic Church moved toward linguistic unification, discouraging ethnic distinctions and promoting English as the primary language of worship (Salmons, 2005, p. 139).10 Similarly, Lutheran churches, once divided by ethnicity, experienced decreasing institutional support for immigrant languages by the early 20th century (Goldberg, 1990, pp. 10–14; Dietz, 1949). The unification of synods into larger, English-dominant groups further contributed to language shift (Salmons, 2005, p. 140). While some congregations maintained German into the early 20th century, many transitioned to alternating English and German services in the 1920s and 1930s, and by the 1950s, German services had largely disappeared (Seeger, 2006; Rocker, 2022).

In both Norwegian and German congregations, the loss of local control and broader societal pressures led to a gradual shift away from heritage languages in the 1920s and 1930s, with English becoming the dominant language of worship and religious instruction by the 1950s.

2.4. Language Use and Shift in Education

Despite the prevalence of Norwegian in many Midwest communities, local schools often used English as the language of instruction from the outset. However, these early, community-centered schools facilitated social interaction among Norwegian-speaking children, reinforcing language maintenance through playground use and friendships (Natvig, 2022a). In Ulen, MN, district mergers in the 1940s led to children attending school with peers from different backgrounds, weakening close-knit social ties and reducing the social pressure to maintain Norwegian (Natvig, 2022a). This pattern likely extends to other North American Norwegian-speaking communities, though further research is needed to determine precise timelines.

Many German immigrants placed a high value on education and literacy in German, with local and church-based schools supporting German instruction throughout the early 19th century. In Wisconsin, local schools operated with considerable autonomy, often providing education in standard German, despite laws mandating English instruction, which were either ignored (1854) or repealed after resistance (1889) (Salmons, 2005, p. 141; Kellogg, 1918, p. 4). Texas also passed English-only education laws in 1909 and 1918, contributing to a slow shift from bilingualism toward English usage in the public domains. Among Low German-speaking immigrants, English was often adopted early as the educational language mirroring the prior diglossia, where standard (High) German was used for formal contexts. Thus, English became the new language of literacy and education, while the heritage variety retained the function as spoken vernacular (Rocker, 2022; Seeger, 2006; see also Bousquette & Ehresmann, 2010, p. 271 for West Frisian). However, the small local schools still provided children with the opportunity to meet other children who spoke the immigrant language and thus may have contributed to initial language maintenance (similar to the situation in Norwegian communities, as studied by Natvig, 2022a).

Importantly, whether the language of instruction was English or a standard heritage variety, local schools initially served as gathering places that supported language maintenance in both Norwegian and German communities. However, as state governments expanded their influence over education, the shift to English in this domain became more pronounced, especially by the 1930s and 1940s, further contributing to the language shift of the communities. As younger community members increasingly became illiterate in the heritage language, the effects were also felt in the heritage language press, as shown below.

2.5. Language Use and Shift in the Press

The Norwegian immigrant community supported a vibrant Norwegian-language press with about four hundred different publications, large and small. Three immigrant newspapers were especially successful, namely Skandinaven (Chicago, 1866–1941), Minneapolis Tidende (Minneapolis, 1887–1935), and Decorah-posten (1874–1972, Decorah, IA). These newspapers had a wide appeal, indicating that Norwegian literacy was a widespread phenomenon (Lovoll, 1999, pp. 173–189). However, as the 20th century progressed, and with societal shifts towards English, these newspapers faced declining readership and eventual discontinuation, reflecting broader community trends towards assimilation. Even two of the largest Norwegian-language newspapers (Skandinaven and Minneapolis Tidende) were discontinued in 1930s and 1940s, providing another indication of the communal language shift at that time.

Similarly, German-language newspapers initially thrived, with a large readership in states such as Pennsylvania, Nebraska, Wisconsin, and Texas (Dolmetsch, 1976, p. 192; Condray, 2015; Schach, 1984; Etzler, 1954; Ramsden, 2016). In the late 19th century, German-language newspapers accounted for 80% of all non-English publications in the U.S., peaking at nearly 800 titles in 1893–1894 (Wittke, 1973, p. 208; Dolmetsch, 1976, p. 187).11 These newspapers provided a mix of local, national, and international news, political commentary, homeland reports, advertising, and fiction. While initially popular among first-generation immigrants, later generations increasingly turned to English-language publications, contributing to a decline in German newspaper circulation. World War I further exacerbated this decline due to anti-German sentiment and legal restrictions on the public use of German (Kloss, 1966, p. 237; Wittke, 1973, p. 244).12 The generational language shift, coupled with these additional challenges, led to a significant reduction in the number of German-language newspapers, decreasing from 554 in 1910 to 234 in 1920, to 146 in 1930, to 90 in 1940, to 41 in 1960 and a mere 20 in 1984 (Haller, 1988, p. 190). Many newspapers either switched to English or ceased publication due to declining profitability. The decline in both Norwegian and German immigrant newspapers highlights the broader language shift trends within these communities.

2.6. Language Use and Shift in the Workforce

Many Norwegians initially settled in rural areas in the Midwest engaging in farming. Some Norwegian settlements were, to a large degree, socio-economically independent of the English-speaking society until the 1940s, as shown in Vernon County, WI (Munch, 1949, pp. 782–784). In such settlements, Norwegian was the language of the workplace, with communities forming around shared labor on the family farms and in small local businesses. In the area surrounding Westby and neighboring Coon Valley in Vernon County, Wisconsin, there are reports of non-Norwegians learning the local Norwegian dialect to live in the area (Haugen, 1953, p. 505; Munch, 1949, p. 784). However, as the 20th century progressed, economic developments and industrial advancements began to reshape the labor landscape. Norwegian communities became more externally oriented because of the Great Depression, the subsequent New Deal programs, and the concurrent mechanization of farming (Natvig, 2022a). As Norwegian-speaking workers entered English-dominant workplaces, the use of Norwegian in professional settings diminished.

German immigrants also contributed to both rural agriculture and urban industries, particularly in cities like Milwaukee, where German was often used in instruction, collaboration, and negotiation (see Lucht et al., 2011 and citations therein). In rural communities, most people worked on their own family farms or at businesses within the community, which meant that the heritage variety was used in the workplace and to conduct business. However, economic pressures, urbanization, and industrialization weakened the role of German in labor settings. Increased social and economic mobility encouraged community members to leave rural, German-speaking communities for urban job opportunities and higher education (Salmons, 1983, p. 188). Simultaneously, non-German-speakers moved into these areas, further reducing the usage of German in public domains such as businesses and the workforce (Salmons, 1983, p. 188). The shift from Norwegian and German to English in professional contexts reflects broader patterns of sociocultural change that changed the makeup of many immigrant communities and lead to increasing use of English across domains.

2.7. Summary: Language Shift in American Norwegian and American German Communities

In summary, we see that shift to English in most public domains occurred between the 1910s and 1940s in both Norwegian and German communities in America. The gradual decline in Norwegian and German language usage in schools, churches, press, and workplaces was driven by decreasing local control over institutions and increased social mobility to pursue work opportunities outside the community. Although the shift may have slightly varied by domain and community, it was mostly complete in the public domains by the 1950s. The last generation of both Norwegian and German heritage speakers was born between 1930 and 1940 with few exceptions (see Salmons, 2022) for similar findings from different speech communities in the Midwest). As such, these heritage speakers acquired the language in their homes but had limited access to heritage language interlocutors, received input in fewer domains and experienced a lack of academic support for their heritage language. From school age or earlier, they started acquiring English and eventually became English-dominant at some point in their lives. Today, the NAmNo and most German varieties in the USA are moribund because this last generation did not pass the heritage language on to their children or grandchildren. For a more condensed comparison of the sociohistorical and linguistic developments in the two communities, please see Appendix A. In this table, we summarize important communal events affecting language use by domain, hence illustrating the gradual language shift. In the remainder of this paper, we will focus on language change observed diachronically in these varieties and discuss how language shift affects that change.

3. Diachronic Studies

There is a body of work devoted to understanding both sociocultural and linguistic change in the communities and language of the German and Norwegian immigrant communities under discussion. In this qualitative literature review, we aim to provide a novel synthesis of the present research to evaluate the degree to which language shift affects language change in moribund heritage varieties (see 1, compare claims in van Baal, 2022; Lykke, 2022; Eide & Hjelde, 2023). We have elected to discuss NAmNo and three specific heritage German varieties because of the existence of both diachronic linguistic and historical sociolinguistic studies of them. We discuss all studies with diachronic data which enable comment on the correlation between language shift and language change, which are available at the time of writing.13 Importantly, these heritage varieties are not characterized by change or anything resembling a collapse of the grammatical system, and there are clearly demonstrated cases of high stability (Lykke, 2020; Larsson & Kinn, 2022; Louden, 2006). However, only the cases where change is found in the present-day varieties can be used to comment on the relation between change and shift. This naturally guides the selection of studies discussed below.14

All data presented in this section are retrieved from the relevant cited sources. As noted before, NAmNo, with the Corpus of American Nordic Speech (CANS) providing spontaneous speech data spanning 80 years, is in a rather unique position in the study of moribund heritage varieties. While diachronic data are thus less readily available for most German varieties, the data we present below still offer a valuable cross-linguistic comparison.

The discussed studies are of linguistic innovation, innovation spread, and other distributional change in language use post-dating migration. By innovation, we understand the occurrence of a novel linguistic feature which is not part of the baseline (see H. Andersen, 1989), i.e., (ideally) the input of the heritage speakers. Examples of innovations can be the delabialization of [y] into [i] in positions with [y] in the baseline, overgeneralization of inflection into new contexts or the innovated placement of a finite verb in the third clausal position where such placement is not licit in the baseline. By distributional change, we understand the altered frequency of use of a linguistic variable. The spread of an innovation is one kind of such change. Another kind is change in usage frequency where a grammar allows for optionality. An example can be the decreased use of non-subjects as the initial constituent of main clause declaratives when a grammar allows fronting of non-subjects (as in both Norwegian and German, see below).

3.1. North American Heritage Norwegian

In the following section, we present NAmNo data and findings from previous studies which enable us to comment on the relation between language shift and language change. All available diachronic studies which analyze data from 1942 (or earlier), in comparison with data from 2010 and onwards, are included in our discussion. This limits us to a discussion of the following variables: tense morphology, language mixing, compositional definiteness (CD), possessive placement, verb placement (V2), argument placement, and the variation and stability of the phoneme /r/ (references in the discussion below).15

All of the studies of NAmNo presented here rely, to some degree, on the Corpus of American Nordic Speech (CANS, Johannessen, 2015). CANS contains spontaneous speech data, and audio and video recordings, which are transcribed, morphosyntactically tagged and searchable with the user interface Glossa (Nøklestad et al., 2017). Importantly for our discussion, the data stem from three different time periods: 1942, 1987–1992, and 2010–2016. We refer to these recordings as early NAmNo or CANS-1942, CANS-1987–1992, and, lastly, present-day, late or post-2010 NAmNo, respectively. Some of the referred studies use data which are not from CANS, namely Eide (2019), Lykke and Hjelde (2022), and Eide and Hjelde (2023).16

Correlating the available corpus data with the sociolinguistic history outlined above, we see that CANS-1942 is a sample of NAmNo spoken by speakers who lived most of their lives in a preshift community. The data from 1987–1992 are somewhat different. The speakers were born between 1909 and 1915 and thus experienced the shift to English in late adolescence or early adulthood, and have lived 40 to 50 years in English-language communities.17 CANS-1987–1992 is also used more sparsely in the referred studies than is CANS-1942 and post-2010 NAmNo data. The post-2010 speakers are predominantly early shift bilinguals, who only experienced Norwegian as a community language in early childhood, if at all.18

3.1.1. Tense Morphology

Synchronic work on late NAmNo tense morphology finds a high degree of stability, but also some trends of innovation (Lykke, 2020; Natvig et al., 2023). Norwegian has two weak classes of verbal inflection, verbs inflecting like kaste “throw”, pret. kasta (a-class) or like prøve “try”, pret. prøvde (Te-class).19 In Lykke (2020), the strongest innovative trend is class overgeneralization, where the inflection of the weak classes spreads into new contexts, either from one weak class to another, or to verbs with strong or preterit-present inflection. An example of such overgeneralization is provided below, with an unchanged form of the preterit-present modal måtte “must” in 1a and an innovative form in 1b:

(1)	a.	vi	mått-e	læ-	vi	mått-e	lære	enngelsk	på	skuurn
		we	must-pret	læ-	we	must-pret	learn.inf	English	at	school.the
		‘We had to learn English at school.’      (Coon_valley_WI_06gm, CANS-2010)

	b.	i	tr-	tre	mane	så	må-dde	ru	havvreste
		in	thr-	three	months	so	must-pret	you	harvest
		‘In three months you had to harvest.’      (Westby_WI_06gm, CANS-2010, Lykke, 2020, p. 180)

In addition to class overgeneralization, Lykke (2020) finds innovated syncretism in present-day NAmNo. Specifically, present tense and past participle forms are used innovatively in preterital contexts (Lykke, 2020, pp. 209–219). In more recent work, Natvig et al. (2023) study all 21 present-day speakers from Coon Valley/Westby in a joint analysis,20 with empirical findings that align with Lykke’s (2020). Notably, Natvig et al. (2023) describe a variable innovated syncretism in the past domain of the Te-class. In 2a, an unchanged preterit is displayed, and 2b shows an example of the innovative trend from late NAmNo, analyzed by both Lykke (2020) and Natvig et al. (2023):

(2)

a.

e

lær-de

fele

me

bærre

#

å

lye

på

#

tep

Rekårdings

I

learn-pret

violin

with

only

#

to

listen

on

#

tape

recordings

‘I learnt to play the violin only by listening to tape recordings.’

(Coon_valley_WI_06gm, late NAmNo)

	b.	e	e	#	læR-d	enngels	fysst
		I	I	#	learn-part	English	first
		‘I learnt English first.’
		Conservative:/læ:r-te/or/læ:r-de/21
		(Coon_valley_WI_12gm, late NAmNo)

In contrast to the present-day speakers, no evidence of post-migration innovation has been identified in the tense morphology of early NAmNo. Haugen (1953, pp. 396–399, 454–456) primarily discusses language mixing, and his discussions do not indicate inflectional change. In a more recent diachronic study of NAmNo tense morphology, Lykke (2022) studies all speakers from Westby/Coon Valley recorded in 1942. The study selects heritage speakers of the same dialects as those spoken in Coon Valley/Westby after 2010.22 The material thus delimited makes for good diachronic comparison with the present-day data from the same location studied by Lykke (2020) and Natvig et al. (2023). Lykke (2022) finds no trends of change in the 1942 dataset. This entails that neither overgeneralization of weak class inflection (see 1b) nor innovated syncretism (see 2b) was found in 1942, nor was there other inflectional change. Lykke (2022) consequently argues that the innovations in the present-day speakers’ language must have arisen with them and not before.

3.1.2. Language Mixing

We now turn to the first of three variables concerning noun phrase (NP) syntax. In work on noun phrase-internal language mixing, Riksem (2017) finds change in the marking of definiteness and number in NPs between the early NAmNo described by Haugen (1953) and late NAmNo (see also Riksem, 2018; Riksem et al., 2021; however, the latter is less diachronically oriented). In early NAmNo language mixing, English roots commonly receive Norwegian number affixes, as demonstrated in 3a–c below (see Riksem, 2017, p. 13). Although Haugen (1953, pp. 449–450) does not make explicit quantitative statements about early NAmNo, the pattern exemplified in 3a–c is, according to him, the clearly predominant pattern:

(3)	a.	creek-ar
		creek-indef.pl.m
		(Haugen, 1953, p. 450)
	b.	field-er
		field-indef.pl.f
		(Haugen, 1953, p. 757)
	c.	team-0
		team-indef.pl.n
		(Haugen, 1953, p. 450)

In addition to such use of Norwegian number affixes, however, Haugen (1953, pp. 450–451) also observes the use of the English plural affix -s in language mixing in early NAmNo. He identifies distinct patterns among two speaker groups: Norwegian–English childhood bilinguals and adults acquiring English as a second language (L2) (or “pre-bilingual borrowers” (Riksem, 2017, p. 13)). According to Haugen, adult L2-acquirers often reanalyze -s as part of the root, because they use it in both sg. and pl. contexts in conjunction with appropriate Norwegian affixes, e.g., car-s-ar “car-s-INDEF.PL.M” and bean-s-en “bean-s-DEF.SG.M”. The other speaker group, the Norwegian–English childhood bilinguals, however, use -s as an indefinite plural affix, as an alternative to functionally parallel Norwegian affixes. The use of -s is mostly limited to indefinite phrases in early NAmNo (Haugen, 1953, pp. 450–451; see Riksem, 2017, p. 13).

In a study of data from 1986 and 1987, Hjelde (1992, pp. 82, 91–94) discusses the plural allomorphy in language mixing of that NAmNo period.23 Like Haugen, Hjelde reports the use of both Norwegian plural affixes (equivalent to those in (3), with predictable dialectal variation) and the use of English -s in language mixing. Hjelde (1992) does not report about a contrast between his data and the findings of Haugen with regards to language mixing (1953).24 In qualitative terms, Hjelde (1992, p. 82) does comment that use of plural -s occurs commonly with English-origin nouns. In light of language mixing in present-day NAmNo, Hjelde’s comment may reflect a gradual change towards the present-day situation, but the empirical picture is not clear.

In post-2010 NAmNo, the patterns with regards to number have changed, when compared to Haugen’s (1953) study of early NAmNo. Out of a total of 175 Norwegian plural phrases with language mixing in Riksem’s (2017, p. 15) dataset, 103 have pl. -s, 37 have a Norwegian affix (see examples in 4), and 35 have zero-marking of plural. The relation in late NAmNo is then the inverse of that in early NAmNo, as Riksem comments. English pl. -s is now used much more than Norwegian affixes with English loan nouns. Secondly, Riksem argues that a zero-marking of plural has arisen, which is not reported by Haugen (1953), exemplified in (4) (Riksem, 2017, p. 16):

(4)	a.	fem	seks	hour_
		five	six	hour
		‘Five or six hours.’

	b.	mange	memorial_
		many	memorial
		‘Many memorials.’

To this, we will add that although Riksem (2017, p. 16) claims that the zero-marking exemplified in (4) is not reported by Haugen (1953), there is some difficulty with her reasoning. All Norwegian varieties have zero-marking of plurals with a large class of neuter nouns, e.g., land-0 “land-INDEF”, which is syncretic for singular and plural. This kind of inflection on loan words is reported by Haugen (1953, p. 450, see 3c). If the examples in (4) are to constitute innovations, they must not be assigned neuter gender. Riksem (2017) does not comment on the gender of the roots with zero-marking of plural, and the question about whether this marking is innovative thus remains open.

Turning to definiteness-marking in language mixing, Haugen (1953, p. 451) reports that the use of a Norwegian definite suffix is obligatory in early NAmNo and that the English definiteness marker, the, is not used. In 5a–c, we see examples of English roots with Norwegian definite inflection suffixes (see Riksem, 2017, p. 13):

(5)	a.	railroad-en
		railroad-def.sg.m
		(Haugen, 1953, p. 590)
	b.	field-a
		field-def.sg.f
		(Haugen, 1953, p. 575)
	c.	det crew-et
		the.n crew-def.sg.n
		(Haugen, 1953, p. 571)

In present-day NAmNo, definiteness is to a large degree still marked as in early NAmNo, exemplified in 6a–c, but Riksem (2017) reports two trends of change. Firstly, she finds omission of the definiteness suffix. Secondly, she reports the use of the English determiner the, in present-day NAmNo, which was unknown in early NAmNo. Riksem’s (2017, p. 18) examples of use of the in both sg. and pl. contexts are provided in 6a–b and in a modified definite phrase in 6c. Definiteness in modified definite phrases is discussed further below, with findings from van Baal (2020, 2022).

(6)	a.	the	ungdom
		the	youth.indef.sg
	b.	the	penger
		the	money.indef.pl
	c.	the	gamle	kirke
		the	old.def	church.indef.sg

In language mixing, there are commonalities between the bilinguals of 1942 NAmNo and post-2010 NAmNo in that English plural -s is used in both periods. The periods clearly differ, however, because in late NAmNo -s has become more common than Norwegian plural affixes in language mixing. Moreover, omission of the Norwegian definiteness suffix and use of English the, separates 1942 NAmNo from 2010 NAmNo. The bilinguals in early and late NAmNo thus behave differently with regards to language mixing, marking a change that has occurred between 1942 and 2010 (Riksem, 2017).

3.1.3. Compositional Definiteness (CD)

Compositional, or double, definiteness is a trait of Norwegian NP syntax.25 The example in 7 shows that modified NPs require both a determiner and a suffix on the noun head marking definiteness, in addition to marking on the adjective. With some crucial exceptions discussed below, such marking of definiteness is obligatory in Norwegian when the definite noun is modified by a numeral or an adjective, as in 7:

(7)	det	norsk-e	språk-et	i	Amerika
	the.n.sg	Norwegian-def	language-def.n.sg	in	America
	(spring_grove_MN_19gm, CANS-2010, van Baal, 2022)
	‘The Norwegian language in America’

The exceptions to this general rule of compositional definiteness (CD) marking are important to understand the developments in NAmNo. The exceptions to the rule allow omission of the prenominal determiner from the phrase (as in 8). Such omission of the prenominal determiner occurs with a number of different specific adjectives and types of adjectives, including superlatives (as in 8), ordinal numerals, and directional terms among others (see van Baal, 2020, pp. 36–39) for discussion and documentation of language use).

(8)	det	er	kanskje	best-e	plass-en	av	alt
	that	is	probably	best-def	place-def.m.sg	of	all.n
	‘That is probably the best place of all.’
	(van Baal, 2020, p. 37)

Crucially, exceptions with omission of the prenominal determiner are more frequent in spontaneous speech than the modified NPs following the general rule of CD. van Baal (2022) makes usage estimations for European Norwegian based on three corpora: NDC (Johannessen et al., 2009), LIA26, and NoWaC (Guevara, 2010). NDC is a speech corpus of early 2000s dialectal Norwegian, LIA is a speech corpus of spoken dialectal Norwegian from the 1900s, and NoWaC is a large written language corpus of Norwegian Bokmål from 2009 to 2010.

van Baal (2022) comments that her European Norwegian data, which we repeat in

Table 1. Frequency of modified definite phrases with CD and exceptions without prenominal determiner in European Norwegian (van Baal, 2022).

	NDC	LIA	NoWaC
Modified definite phrases	4560	7505	3,459,060
Phrases with CD	1049 (23%)	2585 (34.4%)	2,300,981 (66.5%)
Exceptions	3511 (77%)	4920 (65.6%)	1,158,079 (33.5%)

, are not to be understood as a baseline for post-2010 NAmNo, but that they show that omission of the prenominal determiner is common across Norwegian. What is more, she shows that exceptions to CD are markedly less frequent in written Norwegian than in European Norwegian spoken varieties. van Baal (2022) points out that post-2010 NAmNo speakers have predominantly received spoken Norwegian input (see also Eide & Hjelde, 2023). She argues that the high occurrence of exceptions to CD lays the ground for change in post-2010 NAmNo.

In present-day NAmNo, both Anderssen et al. (2018) and van Baal (2020) find that use of CD is much lower than in European Norwegian.

Table 2. Types of modified NPs in obligatory CD contexts in three periods CANS-1942, CANS-1987–1992, and post-2010 NAmNo *.

	1942	1987–1992	ALW 2018	van Baal (2020)
Compositional definiteness	31 (70%)	7 (70%)	93 (39%)	143 (21%)
Without determiner	5 (11%)	0	113 (48%)	339 (49%)
Without suffixed article	7 (16%)	0	31 (13%)	35 (5%)
Bare phrase	1 (2%)	2 (20%)	0 (0%)	123 (18%)
Adjective incorporation	-	1 (10%)	-	47 (7%)
Total phrases	44	10	237	687

* ALW 2018 is the spontaneous speech data of Anderssen et al. (2018) from CANS; the column marked van Baal (2020) comes from van Baal’s (2020) experimental data. The historical data are from van Baal (2022), and the table itself is reproduced from there.

, repeated after van Baal (2022), summarizes all findings on contexts where CD is obligatory in European Norwegian. With regard to comparability, note that Table 2 does not have an entry for licit omission of the determiner comparable to Table 1. Table 2 shows that post-2010 NAmNo has a CD usage ranging between 21 and 39% in contexts where CD is obligatory in European Norwegian.27 Van Baal categorizes the innovative non-CD instances into four groups, namely phrases without the determiner, phrases without the suffixed article, bare phrases, and phrases with adjective incorporation. Of these, adjective incorporation is a further exception to CD found in certain dialects and does not constitute change. As Table 2 shows, the most important innovative category is rather phrases lacking the determiner, which make up 48–49% of the data in post-2010 NAmNo. Phrases which lack the preposed determiner where it was previously obligatory thus make up the most numerous category in post-2010 NAmNo. Overall, the omission of the determiner is, at this point, more common than its use. van Baal (2022) argues that this change in language use reflects the merging of two grammatical categories. Post-2010 NAmNo allows omission of the prenominal determiner with any adjective, where this is not the case in other Norwegian varieties, including early NamNo.

Importantly for the relation between language shift and language change, van Baal (2022) shows that change affecting CD is not present to the same extent for CANS-1942 or CANS-1987–1992 (compare Table 2). Firstly, while the data from 1942 and 1987–1992 are sparse, innovations are even more so; especially the most common trend post-2010, omission of the determiner. In 1942, there are some instances of determiner omission, but the occurrence is low. More common in 1942 is omission of the suffixed article, of which an example is displayed in 9:

(9)	den	eldst-e	bror	min
	the.sg	oldest-def	brother	my
	‘my oldest brother’
	Baseline: den eldste bror(en) min
	(westby_WI_24gm, CANS-1942, van Baal, 2022)

This is an example of a phrase which cannot certainly be considered innovative. In the homeland dialects, so-called bare kinship terms are licit (see Kinn, 2021 for a study of post-2010 NAmNo). van Baal (2022) thus comments that among the seven tokens lacking the postpositioned article from CANS-1942, not all are securely innovative. To this, we will add that written language may be relevant to these data. Danish does not have the postpositioned article in modified definite phrases (for an analysis, see Julien, 2002). In the early NAmNo period, Danish was the written language used by Norwegian immigrants in the US (Haugen, 1953, p. 124). This is relevant because the preshift generation of CANS-1942 had more exposure to written language (and standardized spoken language) (see Eide & Hjelde, 2023). Written input may thus serve to increase omission of the postpositioned article, in our opinion. Written input in Norway was also in Danish at the time, which means that this influence from Danish possibly may have been part of variation brought from Norway.28

In CANS-1987–1992, van Baal (2022) finds no evidence for omission of the determiner, the most prevalent trend in the post-2010 data (see Table 2). There are only ten data points, and the non-CD variant which is found is bare phrases, as exemplified in 10:

(10)	yngst-e	søster	hans	far	min
	youngest-def	sister	his	father	my
	‘my father’s youngest sister’
	Conservative:(den) yngste søstera hans (far)
	(westby_WI_24gm, CANS-1942, van Baal, 2022)

Summing up the research on CD referenced above, Anderssen et al. (2018) and van Baal (2020, 2022) show that CD in post-2010 NAmNo is clearly different not only from European Norwegian, but also from preceding NAmNo periods. van Baal (2022) finds some evidence for innovation in 1942. She comments that these innovations, especially the omission of the prenominal determiner, may be the beginnings of the situation we see in present-day NAmNo. We would like to add to van Baal’s argument that the trends in CANS-1942 might provide indications for the present state of NAmNo, but they do not serve as a definitive prediction. In CANS-1942, the omission of the suffixed article appears as frequently as the omission of the prenominal determiner, but in present-day NAmNo, the lack of prenominal determiners prevails markedly more (refer to Table 2 above). Even if the CANS-1942 data indicate some changes in early NAmNo, these changes do not follow the same trajectory observed by Anderssen et al. (2018) and van Baal (2020) in post-2010 NAmNo. van Baal (2022) suggests that language shift plausibly underlies these differing trends.

3.1.4. Possessive Placement

Staying with NP syntax, we move on to discuss the diachronic development in the placement of possessives in NAmNo. In Norwegian NPs, possessives (POSS) can be preposed, placed before the noun (N), or postposed, placed after the noun as exemplified in 11a-b (Lykke & Hjelde, 2022, p. 68).

(11)				Preposed possessive (POSS-N)
		a.		Min	bil
				my	car
				‘My car.’

			Postposed possessive (N-POSS)
	b.		bilen			min
			car.def			my
			‘My car.’

There is some variation in Norwegian regarding the distribution of preposed and postposed possessives. In European Norwegian spoken language, N-POSS is the unmarked word order and makes up approximately 75% of all possessive constructions (Westergaard & Anderssen, 2015, pp. 25–26). Preposed possessive (POSS-N) yields a contrastive reading, is usually associated with a stressed possessive and is less frequent (c. 25%) (Westergaard & Anderssen, 2015; Eide & Hjelde, 2023, pp. 18–19). These frequency statements, however, only hold for spoken language. The most widely used Norwegian written language (Bokmål) is markedly different. In Bokmål texts, postposed possessives vary between 22% (in newspaper and magazine texts) and 47% (in fiction) (Lødrup, 2012, p. 191). This makes possessive placement a domain where written language input is qualitatively different from spoken language.

Westergaard and Anderssen (2015) have studied possessive placement in present-day NAmNo, and Eide and Hjelde (2023) have performed a diachronic study of the variable. We repeat the data of the latter study in

Table 3. Diachronic development of possessive placement in NAmNo (Eide & Hjelde, 2023, p. 19).

Period/Recordings	POSS-N %	POSS-N #
1930s and 1940s (Haugen)	23%	91/404
1942 Coon Valley/Westby (Haugen)	13%	16/115
1980s and 1990s (Hjelde)	12%	15/126
2010s (CANS-project)	5%	40/739

, because of its diachronic orientation. Both studies find the same difference in the distribution of NAmNo preposed and postposed possessives, when compared to European Norwegian. The overarching picture is that one of the two word orders is generalized more strongly in post-2010 NAmNo and that a rise in postposed possessives occurs with the most speakers (Westergaard & Anderssen, 2015; Eide & Hjelde, 2023). Eide and Hjelde’s (2023) data in Table 3 show that the usage frequency of the word order POSS-N in early NAmNo is the same as in present-day European Norwegian.

Eide and Hjelde (2023) point out a possible caveat to the early NAmNo stability. Their data in Table 3 show that the occurrence of POSS-N word order specifically from the Coon Valley/Westby area in 1942 is about 10% lower than other early NAmNo and European Norwegian. Eide and Hjelde (2023, p. 19) argue that there is dialectal variation in European Norwegian with regard to the occurrence of different word orders and that this variation persists in NAmNo.29 Thus, they claim that the Coon Valley/Westby numbers are not indicative of change in early NAmNo. It is unclear whether they view the later data from the 1980s–1990s as evidence of change. However, they comment that the occurrence of word orders remains constant in data from Coon Valley/Westby but say nothing specific about the two other locations, Zumbrota and Wanamingo. We will add that the data in Table 3 imply that the use of prenominal possessives has gone somewhat down at these other locations, when compared to early NAmNo and European Norwegian.

Most importantly for our discussion, the trend notably differs between post-2010 NAmNo and early NAmNo. With post-2010 speakers, the use of the prenominal possessive has gone down, whatever the baseline. Additionally, however, there is substantial interindividual variation in the present-day data (Westergaard & Anderssen, 2015). Marked interindividual variation is also reported by Eide and Hjelde (2023), who find one outlying speaker predominantly using prenominal possessives (89%), in stark contrast to the overall trend (5%). The overall picture is that change occurs in the period after language shift, and the most pronounced trends are found with the present-day speakers.

3.1.5. Verb Second (or Verb Placement)

Typically, finite verbs are placed in second position in Norwegian, but the syntax of NAmNo shows some tendencies of change. Here, we discuss only main clause declaratives with fronting (topicalization) of non-subjects.30 As a rule, Norwegian requires verb-second (V2) with topicalization, as exemplified in 12:31

(12)	Da	hør-er	gutt-ene	noe	leven	ifra	andre	sid-a	av	låv-en.
	Then	hear-pres	boy-pl.def	some	noise	from	other	side-sg.def	of	barn-sg.def
	‘Then the boys hear some noise from the other side of the barn.’ (Lykke, 2018, p. 76)

Increased use of non-V2 word order has been found in a study of post-2010 NAmNo. The increase is shown to correlate with decreased use of topicalization (Westergaard & Lohndal, 2019; Westergaard et al., 2021).32 In European Norwegian spoken language, approximately 28% of all declaratives have topicalized non-subjects (Eide & Sollid, 2011, p. 347), and non-V2 in obligatory contexts is practically non-existent.33 In an analysis of 50 post-2010 speakers of NAmNo, Westergaard et al. (2021, pp. 13–15) find that the overall frequency of topicalization is at 18.5% (1961/10,609 clauses) (see also Larsson & Kinn, 2022, p. 249) and that non-V2 is at 10% (188/1961 non-subject initial clauses). There is great intra- and interindividual variation, however, with regards to both variables. All speakers still produce V2 structures, but to a variable extent. The frequency with individual speakers of non-subject initial clauses, i.e., with topicalization, ranges from 0% to 42%, and non-V2 ranges from 0% to 82%. There is a clear correlation between the frequency of topicalization structures and non-V2 produced by individual speakers.

Diachronic investigation of V2 in NAmNo reveals that the occurrence of topicalization structures gradually decreases after 1942. The emergence of non-V2 structures occurs, however, primarily in the post-2010 dataset. Eide’s (2019) examination of data from the 1940s finds topicalization at approximately 30–35%, with minimal instances of innovated non-V2 structures (see also Eide & Hjelde, 2015, 2018).34 Notably, she reports that the 1940s NAmNo has the same amount of topicalization as European Norwegian (Eide, 2019; see also Larsson and Kinn’s (2022, p. 249) data on subject placement). However, in the data from 1980–1990, Eide (2019) finds a decline in topicalization to 15–17% and, concurrently, a certain occurrence of innovative non-V2 word order. We can add that Eide’s 1980–1990 data present topicalization seemingly akin to the post-2010 NAmNo of Westergaard et al. (2021), but with fewer non-V2 structures.

In different work, a corpus of 2179 declaratives produced by eight speakers shows a correlation between year of birth, the amount of topicalization, and non-V2 production (Lykke & Hjelde, 2022; see also Eide & Hjelde, 2023). These data are displayed in

Table 4. Diachronic development of V2 in NAmNo (Lykke & Hjelde, 2022, p. 67; Eide & Hjelde, 2023, p. 21) *.

Speaker (Birth Year)	1A 1908	2A 1915	4A 1922	3A 1932	1B 1939	2B 1943	3B 1957	4B 1961
Declaratives	235	87	254	779	274	164	68	318
Topicalization %	23	28	27	29	13	25	10	10
Non-V2 %	5	6	0	1	11	0	14	20

* Eide and Hjelde (2023, pp. 13–15, 21) provide this information about the times when these recordings were made: 1942 (1A and 2A), 1987/1992 (3A and 4A), 2010 (1B and 2B), and 2010–2018 (3B and 4B) (Eide & Hjelde, 2023, pp. 13–15, 21). These times correspond quite closely to the three NAmNo periods we use otherwise, CANS-1942, CANS-1987–1992, and post-2010 NAmNo. Speaker 4b: this speaker is the same as the one called «Ole» by Eide (2019, p. 50).

. Neither Lykke and Hjelde (2022) nor Eide and Hjelde (2023) analyze the data in light of date of birth, but rather focus on these speakers being parent−child pairs. The speaker designations in Table 4 are motivated by this: 1A is the parent of 1B, 2A is the parent of 2B, and so forth. Firstly, the cited studies both point out that the parents (A) have more topicalization and less non-V2 than their children, with one exceptional pair, 2A and 2B. Secondly, we will remark that the overall trend is that speakers born earlier have more topicalization and less non-V2, albeit with individual variation. The four earliest born speakers (b. 1908–1932) show an average of 27% topicalization and 3% non-V2, whereas the latter born speakers (b. 1939–1961) produce an average of 14.5% topicalization and 11% non-V2. Apart from the outlier 2B, speakers born in 1939 or later show a different pattern from the four speakers born in 1932 or earlier.35 The correlation between date of birth and non-V2 syntax in NAmNo is parallel to the findings from Iowa Low German (see below). We furthermore believe the year of recording may be relevant to the trends of the two groupings (A vs. B), because a later year of recording removes a speaker from the time of language shift. The earlier born speakers 1A and 2A were recorded in 1942 and 3A and 4A were recorded in “1987/1992”, while 1B–4B were recorded after 2010 (Eide & Hjelde, 2023, pp. 13–15, 21). The time of recording is one possible explanation for the difference between e.g., 3A and 1B, whose years of birth are close but who were recorded about twenty years apart.36

To sum up the trends in the diachronic study of V2 word order, we see an overall trend that both the frequency of topicalization and V2 decrease over time. Early NAmNo shows definitely less change than post-2010 NAmNo which shows innovation and, particularly, high interspeaker variation. Topicalization decreases first (according to Eide, 2019), followed by an increase in non-V2 usage in the very last stage of the language (Eide, 2019; Westergaard et al., 2021). A supporting trend is seen when data are analyzed by year of birth (with data from Lykke & Hjelde, 2022; Eide & Hjelde, 2023). Speakers born and/or recorded earlier have more topicalization and more V2 in their language use.

3.1.6. Argument Placement

Diachronic work on argument placement in NAmNo shows some of the more stable trends among the phenomena discussed in this section. Larsson and Kinn (2022) examine the placement of subjects, objects, and particles in older dialectal Norwegian (LIA), early NAmNo, and late NAmNo. A main finding is that the word order varies across all datasets and that this variation predominantly remains stable across several generations in NAmNo. Norwegian allows variable word order with regards to the placement of subjects, objects, and particles, and Larsson and Kinn (2022) argue that this underlying grammar has not changed across the history of NAmNo. According to Larsson and Kinn (2022, pp. 253–260), the placement of objects and particles in post-2010 NAmNo differs the least from early NAmNo and older dialectal Norwegian, presenting examples of high stability and little change between the periods.

When it comes to subject placement, however, Larsson and Kinn (2022, pp. 248–253, 259) comment on a certain effect of the heritage language context, which may be relevant to the relation between language shift and change. Norwegian allows placement of subjects (S) in three positions: clause-initially 13a; after the finite verb, but before a sentence adverbial 13b; or after a sentence adverbial 13c. Larsson and Kinn (2022) study clauses with negation (Neg) as the sentence adverbial. The word order in 13b, S-Neg, is commonly called subject shift (SS).

(13)	a.	Lisa	likte	ikke	boka.		(Clause-initial subject)
		Lisa	liked	not	book-def
	b.	Derfor	likte	Lisa	ikke	boka.	(S-Neg, Shifted subject)
		Therefore	liked	Lisa	not	book-def
	c.	Derfor	likte	ikke	Lisa	boka.	(Neg-S, Unshifted subject)
		Therefore	liked	not	Lisa	book-def
		(Larsson & Kinn, 2022, p. 242)

As in the previously discussed studies of V2, Larsson and Kinn (2022) find that subject-initial clauses become more frequent in post-2010 NAmNo (i.e., that topicalization of non-subjects decrease). What is more, however, they find a frequency change with regards to subject shift. Early NAmNo has somewhat less subject shift than older European Norwegian, and post-2010 NAmNo has even less (see

Table 5. Subject-shift occurrence over time (Larsson & Kinn, 2022, p. 250).

	S-Neg	Neg-S	N
LIA	91%	9%	190
Early NAmNo	75%	25%	79
Post-2010 NAmNo	52%	48%	141
Total	75%	25%	410

).

Larsson and Kinn (2022) do not see grammatical change in this frequency shift, but a continuation of pre-existing grammatical optionality. They argue that the principal explanation for the decrease of S-Neg word order in post-2010 NAmNo is the concurrent increase of subject-initial clauses. A main reason it that subject-initial clauses and S-Neg clauses contain the same kinds of subjects, namely topical subjects. This renders the decrease in S-Neg word order an epiphenomenon of the increase of subject-initial clauses, according to Larsson and Kinn (2022, p. 251). As touched upon before, early NAmNo differs from post-2010 NAmNo because it does not see an increase in subject-initial clauses. Hence, the increase in subject-initial clauses cannot be evoked as explanation of the decrease in S-Neg word order in that period.

Thus, there are some differences between early and post-2010 NAmNo, although they are more subtle than with e.g., CD or possessive placement. Although it is not the main point of Larsson and Kinn’s paper, it is relevant to our discussion that they report some differences between early and post-2010 NAmNo. As mentioned, the post-2010 decrease in subject shift can mostly be explained by the increase in subject-initial clauses. Larsson and Kinn also report a rise in unshifted subjects in interrogative clauses, which cannot be explained the same way. The LIA (European Norwegian) data show 94% (n = 18) shift in interrogatives, early NAmNo has 100% (n = 6), but late NAmNo has only 24% (n = 46). Like Anderssen and Westergaard (2020), Larsson and Kinn (2022, p. 252) commonly find such unshifted subjects in tag questions (14).

(14)	du	kann	klare	det	kann	ikke	du?
	you	can	make	it	can	not	you?
	‘You can make it, can’t you?’
	(Post-2010 NAmNo, coon_valley_WI_07gk, Larsson & Kinn, 2022, p. 252)

Unshifted subjects like in (14) are not found in early NAmNo. As also suggested by Anderssen and Westergaard (2020), Larsson and Kinn (2022, p. 252) see the increase in such unshifted subjects as an effect of cross-linguistic influence (CLI) from English, because the English syntax does not have subject shift in such contexts. With regard to interrogatives, early NAmNo (although with few data points) patterns like LIA, i.e., older European Norwegian.

Overall, Larsson and Kinn attribute the relatively minor frequency shifts between the LIA data, early NAmNo, and post-2010 NAmNo to factors such as grammatical complexity and syntactic economy. The distinction between the two NAmNo periods, however, is that post-2010 NAmNo has an effect of CLI which is not seen in 1942-NAmNo.

3.1.7. Variation and Stability of /r/

Natvig (2022b) studies the phonetic and phonological patterns of /r/ in NAmNo, and finds a certain degree of change over time. In the Eastern Norwegian dialects studied, /r/ is protopically realized as an alveolar tap (Kristoffersen, 2000, p. 24; Natvig, 2022b, p. 820). This realization occurs in onsets and variably in codas. In coda positions, however, /r/ undergoes different alternations which occur variably. Either /r/ is deleted, or it partakes in retroflexion when followed by a coronal consonant. In retroflexion, /r/ and the following coronal merge into a retroflex or postalveolar (Kristoffersen, 2000, pp. 96–97, 311–315; Natvig, 2022b, p. 821). Retroflexion of the kind known from pre-emigration dialects is realized in Natvig’s acoustic data as a short decrease in the third formant (F3) in the preceding vowel. Importantly, in the retroflexion, realization with an approximant [ɹ] is unknown in the relevant Norwegian pre-immigration varieties (Ross, 1907, pp. 37–73; Haugen, 1953, p. 433; Natvig, 2022b, p. 820). Natvig does find such realization in the NAmNo data, however. Acoustically, realization with an approximant [ɹ] is also characterized by an F3-decrease, but in such cases, the decreased F3 lasts for more than half of the duration of the preceding vowel. Thus, the innovative realization is acoustically differentiated from the non-innovative by the duration of the F3 drop.

The main trend of change reported by Natvig (2022b, pp. 828–832) is that approximantic realization of /r/ in retroflexion coda positions increase with the year of birth of the speaker, with a correlating decrease in /r/-deletion in the same position, i.e., speakers born later have more approximantic /r/ realization, which is an innovation of NAmNo. Realization in other positions do not change significantly. Natvig (2022b, pp. 826–827) studied 20 speakers from CANS from Westby/Coon Valley, recorded in 1942 (n = 5), 1990–1992 (n = 3), and 2010–2012 (n = 12), i.e., the three periods covered by CANS. As opposed to other studies referred here, however, Natvig does not analyze explicitly by year of recording, making comparison more difficult. The speakers range in year of birth from 1879 to 1957.37 All of the speakers recorded in 2010–2012 are typical post-2010 NAmNo heritage speakers, born in 1922 or later.

3.1.8. Trends in Diachronic Data of American Norwegian

The studies introduced here show that late (post-2010) NAmNo exhibits trends distinct from early NAmNo. While some change can be subtle, like with argument placement, the trends differ clearly in many cases. In other words, the variables which have been subject to diachronic study show trends of change which occur after the communities shifted away from Norwegian in the 1930s or 1940s. Importantly, we do not claim that all studied variables in post-2010 NAmNo show change, as varying degrees of grammatical stability are found (e.g., Kinn, 2020; Larsson & Kinn, 2022; Natvig et al., 2023). What we do see is that most change is found in the post-2010 recordings or, relatedly, that later born speakers exhibit more change.

3.2. Heritage German Varieties in North America

As described before, countless German-speaking settlements were established across the United States throughout the 19th century. Because these communities often retained distinct dialects or developed koinés based on the heritage varieties, it is important to distinguish these different groups in linguistic comparisons. While diachronic data are available for some communities based in Wisconsin, Texas, and Iowa, unfortunately, studies comparing different data sets in a diachronic perspective are still rare (typically because of a lack of older audio recordings). Nonetheless, the examples in this overview can paint a distinct picture of the communal language shift and different trends of language change across geographically dispersed and linguistically distinct German-speaking communities.

3.2.1. Front Rounded Vowels in Texas German

One well-studied phenomenon is the gradual decline of front rounded vowels [y] and [ø] in New Braunfels German (NBG, a Texas German dialect). The earliest study by Eikel (1954) suggests a generational shift in the way rounded vowels are pronounced. In the oldest speaker group, born between 1855 and 1875, two out of six informants consistently round the target vowels, two informants use rounded and unrounded variants, and two informants consistently use unrounded vowels, meaning that /y:/ is consistently produced as /i:/. In the middle generation (born between 1880 and 1910), one informant consistently rounds the vowels, while eleven informants fluctuate between rounded and unrounded variants. In the youngest generation (born between 1910 and 1930), all informants use only unrounded vowels. Similarly, in a study by Clardy (1954), the oldest informant shows rounded vowels, two informants in the middle generation show unrounded vowels, and two informants use both variants, while the youngest speaker consistently uses unrounded vowels (as cited in Boas, 2009, pp. 107–109). In Gilbert’s (1972) Linguistic Atlas of Texas German, all 13 NBG speakers born between 1880 and 1910 as well as those two speakers born after 1910 only produced unrounded vowels. Here, the data include the unrounding of the high round front vowel [y:] to [i:] and the front rounded vowel [ø] to [e:]. These findings are summarized and quantified by Boas (2009, p. 109) in

Table 6. Age-graded distribution of rounded and unrounded front vowels (in Texas German) (Boas, 2009, p. 109) *.

Generation	Eikel (1954)	Clardy (1954)	Gilbert (1972)
1855–1875
Rounded	33.33% (2 of 6)	100% (1 of 1)	n.a.
Unrounded	33.33% (2 of 6)	0%	n.a.
Mixed	33.33% (2 of 6)	0%	n.a.
1880–1910
Rounded	8.3% (1 of 12)	0%	0%
Unrounded	n.a.	50% (2 of 4)	100% (13 of 13)
Mixed	91.6% (11 of 12) *	50% (2 of 4)	0%
1910–1930
Rounded	0%	0%	0%
Unrounded	100% (X)	100% (1 of 1)	100% (2 of 2)
Mixed	0%	0%	0%

* We added information of the number of informants in brackets per age group. Note that Boas (2009, pp. 107–109) is cautious in some of his estimations given the lack of quantifiable data in the original sources. * In Boas’ original table, the percentage is given as 100% for mixed realization (Boas, 2009, p. 109); based on the citation of Eikel’s description (Boas, 2009, p. 107), however, we have changed this percentage to 91.6%, based on the number of speakers.

.

In the Texas dialect project (Boas, 2002), a resampling of Gilbert’s stimuli die Tür (“the door”; target vowel [y]) and zwei Töchter (“two daughters”; target vowel [ø]) was included. For die Tür, 49 of 52 informants (98%) produced the unrounded vowel [i:] and one informant used [y:], while two speakers did not give any answer. For Töchter, 27 out of 52 informants (55%) produced the unrounded vowel [e], 19 informants (39%) produced [o], and only three informants (6%) produced the target rounded vowel [ø], while three did not provide the expected lexical item. Thus, while the larger sample size in Boas’ informant pool shows that front rounded vowels are not entirely extinct from New Braunfels German, their unrounded counterparts seem to be the preferred or sole variant for most speakers. Pierce et al. (2015, pp. 123–129) suggest a number of possible factors that may have influenced the decrease in front rounded vowels in New Braunfels German, including the cross-linguistic markedness of front rounded vowels [1], the impact of the original immigrant varieties that shaped NBG, the influence of Standard German and English, as well as individual attrition. For reasons of scope, we will focus on the last four suggestions here.

Many of the early settlers in Texas originated from “Nassau, Darmstadt, Hessen, Hannover, Württemberg and Bavaria” (Boas, 2009, p. 44), so it is possible that the original emigrant varieties lacked front rounded vowels, which may have resulted in NBG not having this linguistic feature (Pierce et al., 2015, p. 123). However, note that identifying the precise origin of many settlers who arrived between 1840 and 1860 is challenging due to missing demographic information (Boas, 2009, p. 299, fn. 9). If we assume that many of the emigrant dialects did not have front rounded vowels, the fact that diachronic data indicate their existence in earlier NBG stages (Eikel, 1954, 1966; Clardy, 1954; Boas, 2009, p. 109; Pierce et al., 2015, pp. 123–124) highlights the role of some standardized variety.

While Salmons and Lucht (2006) argue that Texas German speakers historically had substantial exposure to Standard German (Pierce et al., 2015, p. 126), Boas (2009, pp. 52–53), in contrast, emphasizes that such exposure was limited for many speakers due to restricted schooling opportunities. Although Boas argues that Standard German knowledge may not have been the norm for all German-speakers in Texas (Boas 2009, pp. 52–53), at least the informants in Eikel’s (1954) study are reported to have had formal instruction in German (as reported in Gilbert, 1965, p. 19). Thus, while those German speakers who were more isolated geographically were less likely to have had regular exposure to Standard German, they were probably also less likely to be included in the early studies cited above due to being more geographically removed (e.g., Eikel, 1954; Clardy, 1954; Gilbert, 1972). Importantly, while institutions using Standard German existed in the preshift stage (including schools, churches, and written media), the shift to English in these domains means that early shift bilinguals typically did not have any exposure to the standard variety.

Although the shift from Standard German to English in the public domains since the 1940s likely contributed to the loss of front rounded vowels (Pierce et al., 2015, p. 128), the retention of front rounded vowels in Texas Alsatian (Roesch, 2012) suggests that English contact is not the primary driver of this change but rather a reinforcing factor (Pierce et al., 2015, pp. 125–126). Following Bousquette’s (2020) interpretation of changes in Dative case marking in Wisconsin German, early use of front rounded vowels in NBG may have stemmed from speakers’ knowledge of Standard German, whereas the later usage of unrounded vowels aligns with a broader return to the spoken heritage variety. Therefore, the shift from Standard German to English in the public domains may have affected the decline of front rounded vowels indirectly, as speakers received less input in the Standard variety exhibiting front rounded vowels and only learned a spoken heritage variety that may not have had front rounded vowels to begin with. Even if the spoken vernacular showed front rounded vowels, we can assume that decreased exposure to German and increased usage of English in a wide range of domains impacted the usage of front rounded vowels, resulting in a steady decline of this feature.

3.2.2. Case Marking

Multiple studies provide compelling evidence for changes in the case marking systems within several German heritage varieties, most prominently in early shift bilinguals. This section introduces studies on Wisconsin German and Texas German.

For Wisconsin German varieties, the earliest audio-recording were done by Lester W.J. Seifert in 1946–1949 (Seifert, 1946), including 62 interviews from Southeast Wisconsin which are now part of the North American German Dialect Archive (NAGDA) at the University of Wisconsin–Madison. In a subset of this corpus, six speakers were asked to translate English sentences into High German, which is the data source for Bousquette’s (2020) analysis of dative case marking. Speakers were born between 1864 and 1901 in Wisconsin and reported exposure to different German home varieties, including Low German, Pomeranian, Oderbrüchisch, and Swiss German (Bousquette, 2020, p. 494), some of which lack a formal distinction of accusative and dative case marking. The analysis of environments where the dative case is expected shows that 69% of these forms are realized as dative, but with considerable individual variation (ranging from 52% to 93% dative case marking). When considering the non-dative forms, speakers seem to systematically replace dative with accusative case marking (59% of non-dative cases are marked accusative). This feature is common in the speakers’ home varieties, most of which have only one oblique case to mark accusative and dative. Thus, Bousquette concludes that the high rate of standard-like dative case marking is due to the speakers’ exposure to Standard German and their level of formal education through schools, churches, and access to German-language media (Bousquette, 2020, p. 489), which was still the status quo for speakers in the multilingual preshift generation.

A similar study by Lewis (1992) examines the dative marking after prepositions and in personal pronouns in the Swiss German of speakers from New Glarus, Wisconsin. Lewis reports that language use was stable until the end of the 19th century, with Swiss German as the main community language and Standard German used in church, school, and the newspaper. Language shift to English gradually replaced Standard German in schools by 1914, the newspaper by 1921, and the church by 1950 (Lewis, 1992, p. 2). By the late 1960s and early 1970s, knowledge of Standard German was rare, and Swiss German was limited to older generations using it within close family circles. Between 1966 and 1972, a total of 28 informants translated (parts of) Lester Seifert’s Wisconsin German Questionnaire (Seifert, 1946). The questionnaire includes around 180 contexts in which dative forms are canonically expected. Ten out of eleven speakers born between 1885 and 1920 consistently employed dative forms in the expected grammatical environments, using “four or fewer non-dative forms” (Lewis, 1992, p. 7). However, speakers born in 1920 or after exhibited a higher frequency of non-dative variants (Lewis, 1992, p. 7), often using accusative forms instead:

(15)	Mir	häid	e	Ränter	ùff	üüseri (expected:üüserer)	Farem
	we	have	a	renter	on	our	farm
	‘We have a renter on our farm’ (Lewis, 1992, p. 7)

Unfortunately, Lewis does not quantify the results for the younger speakers in more detail, but since Lewis’ refers to a “loss of the dative” (Lewis, 1992, p. 7), we can assume that the innovative use of non-dative forms is much higher among speakers born after 1920 from his qualitative statements. The loss of dative case marking in New Glarus likely reflects both inherited patterns from the original dialect and the effects of language shift (Lewis, 1992, p. 10). In the multilingual preshift stage, speakers received input in Standard German in the school, church, and newspaper, meaning that preshift speakers may have been more aware of a standardized case system and thus implemented the dative case more consistently in Swiss German as well. As reported, the local institutions gradually shifted to English, so speakers born in the early shift bilingualism stage would have received very little, if any, input in Standard German (and probably no formal instruction in the language). The shift to English in Standard German domains directly impacted the early shift bilinguals in that they did not receive input in a standardized variety, which could have strengthened their awareness and use of the dative case. Interestingly, Lewis mentions that dialect maps from Switzerland show some loss of dative endings in Glarus, suggesting that this feature may have been an inherent feature in the immigrant dialect (Hotzenköcherle, 1975, pp. 172, 181, 190, as cited in Lewis, 1992, p. 10). Therefore, the perceived decline in dative case marking may also be indicative of a lack of Standard German knowledge and a return to the heritage grammar, as argued by Bousquette (2020) for Wisconsin German. However, since Glarus Swiss German at least has some dative case marking as part of its grammatical system, the shift to English in the public domains probably also resulted in fewer opportunities to speak Swiss German, meaning fewer chances to practice dative case marking during language acquisition. It should be noted that Auer and Derungs (2018, p. 7) reanalyzed the data collected by Lewis and compared it to homeland data to investigate the preservation of Glarus dialect features. They observe that both phonology and morphology of New Glarus Swiss German was influenced by “dialect contact with the Bernese Swiss German as well schooling in Standard German”.

For Texas German, Salmons (1994) analyzed dative case marking based on Gilbert’s (1972) Texas German Atlas. A notable decline in dative case marking was observed among Texas German speakers based on year of birth. Speakers born before 1900 already showed considerable variation in the use of dative case marking (66.1% dative), which is even more noticeable in speakers born between 1900 and 1912 (55.1% dative). This development culminates in individuals born after 1912, who typically used the accusative case instead of dative case (28.5% dative). Salmons suggests that the introduction of English in local schools may have contributed to the decline of the dative case. Boas’ (2002, p. 395) study, which draws upon the works of Gilbert (1965), Eikel (1966; but see also Eikel, 1949), and Guion (1996), provides additional evidence for the decreasing use of the dative case and concludes its complete replacement by the accusative case in Texas German in data collected after 2002. In addition, Boas (2002, p. 396) points out that some instances of nominative usage, instead of accusative (in previously dative environments), can be observed.

3.2.3. Verb Second (or Verb Placement)

Although Germanic languages (with the exception of English) typically place the finite verb in second position (V2) in non-subject initial declarative main clauses, many heritage or contact varieties show verb placement variation (see Section 3.1.5 for NamNo)., In some varieties, the finite verb might also occur in third position (V3), after a sentence-initial adverb and the subject, as shown in this Nebraska Low German example (16):

(16)	in	2001	wi	sünd	in	Arizona	ween.
	in	two thousand one	we	be-aux	in	Arizona	be-part
	‘In 2001, we were in Arizona.’
	(adapted from Wirrer, 2009, p. 141)

For varieties of German, similar examples have been found in Middle High German (Speyer & Weiß, 2018), Middle Low German (Petrova, 2012), spoken High German (Breitbarth, 2022; Bunk, 2020; Schalowski, 2017), the urban vernacular Kiezdeutsch (te Velde, 2017; Wiese, 2011; Wiese et al., 2009, inter alia), contact varieties such as Cité Duits (Pecht, 2019), Russlanddeutsch (C. Andersen, 2016), Wisconsin German (Sewell, 2015), and heritage Low German in the US (Bender, 1980; Rocker, 2022; Wirrer, 2009).38 Although these structures seem to be well described in terms of their occurrence, the only study to date which has analyzed V3-structures in heritage languages from a diachronic perspective is Rocker (2022). For the analysis, main clauses with sentence-initial adverbials were extracted from 52 interviews with 40 heritage Low German speakers. The recordings stem from two data sets, recorded in 1998 (Group A) and 2018/19 (Group B). A total of 664 tokens with a sentence-initial adverbial were found, 180 (27%) of which showed verb third placement instead of canonical verb second placement. The distribution between the two groups is balanced, with 305 tokens obtained from 29 transcripts in Group A (11 women, 18 men) and 359 tokens from 23 transcripts in Group B (7 women, 16 men) (Rocker, 2022, p. 153). The tokens were coded by nine different factor groups and analyzed using a generalized linear mixed model (Rocker, 2022, p. 161), which showed four linguistic and one social factor to be statistically significant in predicting the occurrence of V3-structures: prosodic weight, prosodic marking, tense, verb complexity, and gender. This supports the idea that the use of V3-structures is not arbitrary but highly predictable based on linguistic factors.

Although the social factors (other than gender) were not statistically significant, descriptive comparisons indicate a gradual generational language change between two data sets from 1998 and 2018/19. Most notably, Group A (recorded in 1998) exhibits a V3-usage of only 19%, whereas an increase to 34% is observed in 2018/19. When the data are analyzed based on speakers’ year of birth, a trend emerges (see Figure 1). Despite individual variation, average V3 usage is higher among speakers born after later, pointing toward a gradual language change. Speakers born between 1905 and 1919 typically exhibit V3 rates below 25%, indicating that V3 was already in use before domain-specific language shift but not yet widespread.

In contrast, while many speakers born after 1925 still use little or no V3, a substantial number of speakers show usage rates ranging from 30% to 100%. The dotted trendline shows a slight incline, suggesting a slow increase in V3 usage across successive cohorts.

Interestingly, similar trends were found for NamNo (see Section 3.1.5): speakers born earlier use less V3, though there is some individual variation. The four oldest speakers (born in 1908–1932) average only 3% V3, while the younger group (born in 1939–1961) averages 11% V3. The fact that similar linguistic patterns are found in NamNo and Iowa Low German likely reflects the differences in language acquisition conditions brought on by the language shift itself, as speakers who grew up as early shift bilinguals (i.e., when the heritage language was no longer dominant) appear to be developing and using V3-structures more frequently. Younger speakers grew up during or after a period of language shift, when the heritage language was no longer used as widely across different social domains. As a result, early shift bilinguals typically had less exposure to the language, encountered it in fewer contexts, and may have acquired it under more restricted conditions. These differences in acquisition and input likely shaped their grammatical representations and usage patterns. Moreover, ongoing sociocultural changes have weakened the density of the communal network (once close-knit, it has become more dispersed) and have also contributed to a declining speaker base. Both of these factors are known to accelerate language change, potentially facilitating the stabilization of V3-structures as a characteristic feature of this speech community.

3.2.4. Trends in Diachronic Data of Heritage German Varieties

The studies discussed above demonstrate a variety of features undergoing change within different German heritage varieties. Unfortunately, older audio recordings are often unavailable to current researchers and diachronic studies are still rare. Although no single variety has been comprehensively studied across all the changes discussed, the available data from various variables and varieties suggest that language shift and related sociolinguistic processes impacted language acquisition for early shift bilinguals. Sociolinguistic communal language shift, and particularly the effects of verticalization, occurred in immigrant communities at similar times and had comparable effects. The role of domain-specific language use and the structure of social networks can be identified as complex and overlapping factors influencing heritage speakers’ language acquisition. Rather than marking a clear endpoint, the shift reflects a gradual reduction in both exposure to the heritage language and opportunities to use it in daily life. This prolonged period of English/heritage language bilingualism offers crucial context for understanding ongoing language change. It underscores the importance of considering not just the timing of the shift, but also the changing conditions under which speakers acquired and used the heritage language over time.

4. Discussion

In the previous sections, we discussed how multilingual preshift and early shift speakers had different experiences with regard to the contexts in which their heritage language was used, due to language shift in many (public) domains. Before verticalization and sociocultural changes in the communities happened, the heritage language was used in several domains outside the home, including church, school, and media. With the onset of gradual but widespread language shift in these domains, the heritage language was relegated to use as the home language. This happened roughly around the time of the present-day speakers’ births or happened during their lifespans (see Section 2). It follows from these sociolinguistic differences that the language acquisition and language use across the lifespan of preshift speakers and the present-day, early shift speakers were affected by changes in communal language use. Therefore, early shift bilinguals can be more clearly defined as heritage speakers in the narrow sense, who show differences from majority language speakers in language acquisition, use, and grammars (e.g., Montrul, 2008, 2016; Polinsky, 2018; Johannessen, 2018; Polinsky & Scontras, 2020). The fact that current studies on linguistic change using diachronic data shows higher rates of change in early shift bilinguals may therefore be traced back to the societal changes that affected these speakers in language acquisition and use (as previously suggested by Johannessen, 2018; Eide & Hjelde, 2023). In some instances, the shift to English in domains that were previously occupied by a standard heritage variety may lead to a loss of the standard variety and maintenance of only the vernacular heritage variety in the private domains.

The role of input in heritage language acquisition has been widely discussed with both its quantity and quality cited as key factors contributing to differences between heritage speakers and majority language speakers. The quantity of input is operationalized as the number of years a speaker is exposed to a language during childhood, as well as the amount of exposure a speaker receives during those years. A period of dominance in the heritage language corresponds to a greater quantity of input (Allen et al., 2006; Montrul, 2002; 2008, pp. 152, 193–202; 2016, p. 113), which may result in language acquisition patterns that are indistinguishable from those of majority language speakers (Thordardottir, 2015).39 Sequential bilinguals who begin acquiring second language later in childhood often experience a period of language dominance in their first language (L1), in this case the heritage language. In contrast, simultaneous bilinguals may never have been dominant in their heritage language. Furthermore, exposure to the majority language early in speakers’ lives decreases both the amount of input and opportunities for using the heritage language from early childhood onwards.

In the case of NAmNo and the German-speaking communities under discussion, the sociocultural changes that led to language shift in nearly all domains (see Section 2) must have affected the quantity of input. It is likely that simultaneous bilingualism was less common in the preshift stage where the heritage language was spoken in more domains of society. Certainly, children acquiring the heritage language in preshift communities had more opportunities to speak the heritage language in a larger number of domains and with more different speakers. While many of the recorded present-day speakers still self-report that they are sequential bilinguals, simultaneous bilingualism is not uncommon in speakers born from the 1940s onward (Rocker, 2022, pp. 104–106). As a result, preshift speakers would have received more input in the heritage language than those who grew up during the early stages of language shift (i.e., our present-day speakers).

An example of the effects of quantity of input in the studies discussed here is compositional definiteness of NAmNo. As described in Section 3.1.3, a prevalent trend in late NAmNo CD is omission of the prenominal determiner in modified definite NPs. Studies of monolingual acquisition indicate that the determiner is acquired late in childhood, around age six (van Baal, 2020, pp. 163–164). Acquisition may be even later, or at least different, with Norwegian−English bilinguals in Norway (Anderssen & Bentzen, 2013). The trends of NAmNo heritage speakers are clearly similar to what is seen in monolingual acquisition. van Baal (2020, pp. 162–189) then argues that a likely cause of the change in NAmNo is that input in Norwegian decreased markedly when the NAmNo speakers started going to English-language schools around age 6. We believe this generally marks increased participation of these child speakers in English-speaking domains outside the home. The trends found with early shift bilinguals by van Baal (2020) are not replicated for the earlier stages of NamNo (van Baal, 2022), indicating that quantity of input factors work differently in communities where NAmNo occupied more domains.

Quality of input is another important factor influencing heritage language acquisition, including the diversity, complexity, and richness of the input a speaker receives with regard to both vocabulary and grammatical structures. It is related to the contexts or domains of language use (Montrul, 2016, p. 118; Unsworth, 2016; Eide & Hjelde, 2023), affecting the topics and the type of discourse (e.g., familiar and presentational as opposed to argumentative). Since a standard variety of the heritage language was used in many public domains during the preshift stage, speakers were exposed (at least passively) to different registers and received input covering different topics. For church services, this included listening to services, singing hymns, and potentially interacting with the preacher in a more formal register. In addition, learning to read and write in the heritage language (either in local schools or through religious education) would have provided speakers with access to many forms of language input, including the bible, newspapers, and other literature. The exposure to written media (even without formal instruction in schools) may have created an awareness of standardized phenomena both in terms of pronunciation and grammar, which in turn may have slowed the spread of emerging variables. Norwegian does not have a standard language comparable to High German, but Danish written language and the standard-near spoken variety, or register, served a similar sociolinguistic function (see also Section 3.1.4).

The potential effect of having access to written input is possessive placement in NAmNo (see Section 3.1). Since the use of N-POSS is much more frequent in written Norwegian (and in English), when access to written Norwegian, and standardized spoken language disappears as part of language shift, it affects the quality of the input (see Eide & Hjelde, 2023). Thus, early shift NAmNo bilinguals who are (with few exceptions) illiterate in the heritage language and have also received less spoken standardized language input use POSS-N more frequently than preshift or European speakers of Norwegian. Similarly, it has been argued that the increased use of accusative case marking in traditionally dative case environments in both Texas German and Wisconsin German could stem from a lack of Standard German exposure. For Wisconsin German, Bousquette (2020) argues that preshift speakers showed knowledge of Standard German case marking even in their heritage variety, while the “lack” of dative case marking indicates a return to the heritage variety. For Texas German, the steady decline of dative case marking has been interpreted as language change, which may also have been influenced by the lack of Standard German exposure in early shift bilingual speakers.

In addition to factors influencing language acquisition during childhood, it is essential to consider language use across the lifespan and potential aging effects. Most present-day NamNo and the German heritage speakers are elderly (D’Alessandro et al., 2021; see Salmons, 2022) and may thus be subject to aging effects, which may affect lexical retrieval and grammatical processing (Goral, 2012; Goral et al., 2007; Kemper et al., 2001; note that these studies focus on monolingual majority language speakers, not on heritage speakers). This being said, the studies on heritage speakers reviewed here have found no measurable effect of ageing on linguistic outcomes (Lykke, 2020, p. 48; 2022; Rocker, 2022), but further studies may be necessary.

While ageing by itself has not been shown to affect language change in these varieties, the decreasing use of the heritage language across the lifespan seems to be a more relevant influence on early shift bilinguals. It has been argued that limitations on online processing resources can lead to changes in heritage grammars (Polinsky & Scontras, 2020, p. 14), as bilingual language processing generally incurs a higher cognitive cost. These effects may be intensified in unbalanced bilinguals, where the heritage language is the non-dominant one (Polinsky, 2018, pp. 66–69, 289–290). In such cases, constraints on online processing may lead to restructuring of the heritage grammars to reduce processing demands. Similarly, Paradis (2004, pp. 28–30; 2007) proposes that lack of activation of the heritage language increases the cognitive cost of accessing linguistic items, which may also result in language change (see also Montrul, 2016, pp. 112–113). Given that early shift bilinguals of moribund NAmNo and German have seen little use of their heritage language (and its usage often decreased throughout their lifespan), their language is likely to show the influence of online processing economy.40

Lykke (2020) argues exactly one such case. As outlined in Section 3.1.1, a trend of innovative syncretism is found in NAmNo tense morphology, despite the otherwise high stability of tense morphology in NAmNo and other heritage varieties (Polinsky, 2018, p. 174; Lykke, 2020; Natvig et al., 2023). The innovative items, present and past participle inflectional forms in preterit contexts, are not found in the input or in Norwegian child language, and thus, language acquisition is an unlikely cause of the innovations. Lykke (2020, pp. 222–224) rather argues that the cause is increased cost of processing because of low use of the heritage language across the lifespan. The present tense inflectional forms of verbs have higher token frequency than the preterit, which it variably supplants, and the higher frequency arguably makes the present tense less costly to access. With regard to the relation to language shift, this innovated tense syncretism in NAmNo is only found with the early shift bilinguals of late NAmNo, and not in early NAmNo (Lykke, 2022). These findings imply that language shift may lead to a situation with low language use across the lifespan which is a plausible cause of certain linguistic innovations (see Boas & Fingerhuth, 2017).

While decreased exposure to the heritage level may affect speakers’ grammars on an individual level, the lack of a large community of speakers and the changed structure of the social network may also be conducive to language change. In closely knit communities, individuals often maintain multiplex ties—for instance, a brother may also be a co-worker and a fellow member of the same local church congregation. According to social network theory (J. Milroy & Milroy, 1985; L. Milroy & Milroy, 1992; see also Bousquette, 2022, p. 83), such dense and multiplex social ties tend to reinforce existing linguistic norms, fostering linguistic conservatism and resisting innovation. In contrast, individuals with more uniplex ties, where social relationships are limited to single roles, are more susceptible to linguistic innovation and are more likely to transmit them throughout their looser social networks.

In the Norwegian and German-speaking communities under discussion, sociocultural changes caused by verticalization processes entailed exactly such a transition from multiplex to more uniplex social networks. This “great change” (Warren, 1963, p. 53) brought about structural transformations in rural immigrant communities, including increased mobility, the diversification of personal interests, technological advancements, and specialization of employment (see Bousquette, 2022 for a detailed account). These changes had far-reaching implications—not just for heritage speakers’ personal lives, but impacted the social networks of the community. As Hjelde (2015) notes for NAmNo, the transition from multiplex to more uniplex social networks entails the erosion of dense, norm-enforcing social circles such as family and close-knit peer groups (see also L. Milroy & Llamas, 2013, p. 421). Such social change allows for more intraindividual variation and faster spread of linguistic innovations between individuals. This transformation is a defining factor between the preshift and early shift stages of heritage language communities.

In addition to the makeup of social networks, community size is another potentially crucial relevant factor for innovation spread, as Rocker (2022, pp. 224–225) argues for the spread of V3-structures in Iowa Low German. In small communities, new linguistic forms may spread more quickly since the “critical mass” (Chambers, 2013, p. 312) of adopting new forms may be reached much sooner (see Beeksma et al., 2017 for a computational model of language change in small groups). In large communities, where speakers interact with hundreds of interlocutors, the individual variation of one particular interlocutor will affect the language of an individual to a lesser degree, since the overall proportional input drowns out any innovated forms (Beeksma et al., 2017, p. 35; Eide & Hjelde, 2023). In smaller communities, however, repeated interactions with a limited number of interlocutors may change the proportional input of a particular feature and eventually lead to the adoption of this feature into a speaker’s linguistic repertoire. This makes present-day speaker communities (where they still exist) more conducive to innovation spread than the larger, denser preshift communities. If Beeksma et al.’s (2017) model can be applied to heritage languages, then community size can be viewed as an additional plausible factor influencing the higher rate of innovations observed in contemporary heritage varieties.

The discussions in our paper are highly relevant to questions about the structural linguistic differences between first- and second-generation immigrants, otherwise discussed in this special issue on transgenerational language shifts and changes. In view of the data we present, we see that differences between the “multilingual preshift stage” and the “early shift bilingualism stage” (following Bousquette, 2020) are central. In the linguistic variables discussed here, the crucial factor seems to be the state of language shift across domains in the community during speakers’ language acquisition and lives, somewhat irrespective of generational depth. In many cases, migrant speakers and heritage speakers pattern similarly as long as an active speech community with heritage language use in a variety of domains supports the heritage language of the speakers, which is the case in the “multilingual preshift stage” (e.g., NAmNo tense morphology or V2 syntax).

5. Conclusions and Outlook

Structurally oriented work on North American heritage Norwegian (NAmNo) has claimed that language change in the last generation of NAmNo speakers is caused by preceding language shift (van Baal, 2022; Lykke, 2022; Eide & Hjelde, 2023; compare also Johannessen, 2018). To evaluate this claim, we have presented the present state of diachronic research on language change in NAmNo, Texas German, Wisconsin German, and Iowa Low German and placed the findings in the context of the present historical and sociolinguistic understanding of these communities, most importantly of language shift in the perspective of verticalization analyses (see works in Brown, 2022b).

The present state of sociolinguistic research on NAmNo, Texas German, Wisconsin German, and Iowa Low German paints a clear picture of how and when different domains shifted from the heritage language(s) to English. Although these communities initially maintained their heritage languages for multiple generations, broader, societal changes eventually led to communal changes. Local institutions, once community-run and meeting local needs, became increasingly influenced by external entities like state authorities, larger church synods, or national governments around the turn of the 20th century. Concurrently, labor practices changed, and community members became more reliant on jobs outside the community. These transformations prompted a communal language shift as key public domains (education, church, media, and labor) ceased to support the heritage language and shifted to English. Depending on the community, the onset of communal language shift occurred sometime between 1910 and 1940 and often was a slow but steady shift within different domains.

In addition, we surveyed the state of diachronic linguistic research on the aforementioned Norwegian and German varieties. In our discussion, we highlight the variables which are subject to innovation and change, which is often more pronounced in the present-day generation of speakers.41 The existing body of work indicates clear differences between the different generations of NAmNo speakers studied (e.g., between 1942 and 2010–2016, e.g., van Baal, 2022), or an effect of year of birth (e.g., Natvig, 2022b). Although changes were already found in previous generations, the trends found in Wisconsin German (Lewis, 1992; Bousquette, 2020), Texas German (Boas, 2009; Pierce et al., 2015), and Iowa Low German (Rocker, 2022) indicate an acceleration of linguistic change in the present-day generation.

To a high degree, the work we have discussed supports the hypothesis that language innovation and change is more pronounced in speakers who grew up after key domains had already shifted to English or shifted early on in their lives. Thus, we believe that differing sociolinguistic contexts of the early and late periods of the heritage varieties can be related to commonly invoked causes of change in heritage grammars (compare Johannessen, 2018; van Baal, 2020, 2022; Lykke, 2022; Eide & Hjelde, 2023). The (often gradual) shift away from the heritage language in all societal domains except the home before or during the lives of early shift speakers plausibly affects both their quantity and quality of input and language use across the lifespan. Put differently, the last generation of speakers are heritage speakers in a narrow sense, i.e., majority language-dominant bilinguals (Polinsky, 2018, pp. 4–9), to a much larger degree than preshift speakers. This main difference in speaker profiles between early shift speakers and preshift speakers seems a likely underlying cause for much of the linguistic differences between the groups. In addition, the sociocultural changes that lead to gradual language shift in many public domains also affected the social networks of the heritage language communities. Dense, multiplex networks gave way to looser, more uniplex networks. Such changes to social networks are likely to foster linguistic change, according to social network theory (Bousquette, 2022; J. Milroy & Milroy, 1985; L. Milroy & Milroy, 1992). What is more, innovations may spread more rapidly in small communities, where speakers have fewer interlocutors (Beeksma et al., 2017), which may have affected present-day NAmNo and German heritage varieties as well.

The fact that similar patterns are found in both Norwegian and German varieties in terms of language shift and change supports the idea that larger societal changes impacted immigrant communities at similar times and to a similar extent. Other studies using the verticalization approach have found similar trends for multigenerational immigrant communities and their gradual shift to English. These studies include Wisconsin West Frisian (Bousquette & Ehresmann, 2010; Bousquette, 2022), American Finnish (Johnson, 2018, 2022), Pennsylvania Dutch (Brown, 2022a), among others. Importantly, all of these studies show that language shift does not occur after three generations but rather after larger sociocultural changes affecting the usage of the heritage language in a variety of domains. Similarly, Petersen et al. (2021) show for Danish heritage speakers in North America and Argentina that language proficiency is not directly related to the “immigrant generation” but rather to the sociolinguistic setting.

We hope that this review inspires further studies using diachronic data to analyze linguistic change in heritage languages and approaches that connect sociolinguistic theories with studies on language change. To shed more light on the effect of language shift on language change, further studies are needed on other heritage varieties in the English-dominated context of North America (including, but not limited to American Finnish, Swedish, Dutch, or Danish) that present similar historical developments. Perhaps ideally, however, future work will explore different contact languages in other societies to open for broader contrastive analysis.