*** Abstract ***

Background: The development of V2 word order has been investigated at length in historical data as well as in various acquisition contexts. It has also to some extent been studied in Norwegian heritage language (HL) spoken in the US, often referred to as American Norwegian (Eide & Hjelde 2015, Johannessen 2015a, Alexiadou & Lohndal 2017; Westergaard & Lohndal 2019), and it is found that, while this word order is generally intact in many speakers, there are numerous examples of non-V2; see (1), i.e., variable V2 word order.

(1) Og der dem lager vin. (American Norwegian; from Eide & Hjelde 2015: 89) and there they make wine

‘And there they make wine.’ Target: Og der lager de(m) vin.
Variable V2 is also attested both in historical and synchronic systems, e.g. Old/Middle English and many Norwegian dialects (e.g. Haeberli 2002, Vangsnes 2005). In these cases, the variation is not random, but based on syntactic and information structure patterns. It has also been shown that children acquire this word order variation from early on (e.g. Westergaard 2009).
Goals: This talk addresses the following research questions: 1) Is the choice of V2 vs. non-V2 in Norwegian HL dependent on syntactic/IS factors?, 2) Is the pragmatic structure of declaratives affected in this HL situation, similar to what was found in the history of English?, 3) When V2 is affected in Norwegian HL, is this due to attrition of grammatical structure or to temporary cross-linguistic influence from the dominant language in processing?
Our study: The current study is based on 50 speakers in a corpus of Norwegian heritage language spoken in the USA, the Corpus of American Nordic Speech (CANS; Johannessen 2015b). We have investigated all V2 and non-V2 declaratives for 50 speakers in the corpus. These data have been compared with the Nordic Dialect Corpus (Johannessen et al. 2009), which is a spoken corpus of 400 speakers from around 100 different areas in Norway. Results: The results show that V2 is relatively robust, attested more than 90% in non-subject- initial declaratives. There is no distinction between V2 and non-V2 with respect to subject and verb types chosen, as both word orders are found to the same extent with both NP and pronominal subjects and any verb type. The only statistically significant distinction between the two word orders is related to the initial element, in that non-V2 typically appears with longer elements, while short and very frequent initial elements clearly trigger V2, e.g. så, da ‘then’. Furthermore, while the proportion of non-subject-initial declaratives in NorDiaCorp is similar to what has previously been found for V2 languages, the context for V2 is produced significantly less in CANS, but the variation between speakers is huge. Some speakers have a proportion of non-subject initial declaratives similar to Norwegian speakers (20-40%), while others show a more English-like pattern (6-17%). Importantly, we find a statistically significant correlation between V2 violations and the proportion of non-subject-initial declaratives (p < 0.001): V2 violations are mainly produced by speakers who have a low proportion of non- subject initial declaratives.
Discussion: We interpret the findings in the following way: The syntax of V2 is intact in most of the heritage speakers’ I-language grammars; i.e. there is no syntactic attrition. However, the pragmatic structure of declaratives is affected by cross-linguistic influence from English, reducing the number of contexts for V2. This results in a severe reduction in the activation of this rule in the heritage language, in turn making the syntax vulnerable in processing. Thus, we argue that the non-V2 word order that is attested in American Norwegian is the result of temporary cross-linguistic influence from English in processing (except in cases where V2 is triggered by very frequent initial elements).