This paper explores cross-linguistic influences on object marking in L2 Finnish by contrasting advanced L2-learners from various L1 backgrounds. Possible analogical systems in learners’ L1s expose the alternation to cross-linguistic influences. The results suggest that such influences are not only construction-specific but also sensitive to constructional variation of neighboring phenomena.
We explore cross-linguistic influences (CLI) on object case marking in L2 Finnish by contrasting advanced L2-learners from various L1 backgrounds. We explore whether the object case marking align with the strategies available in the learners’ respective L1s. Such correlations would constitute a comparison-based argument for CLI (Jarvis 2010). The total–partitive alternation of Finnish object is a persistent issue in L2 acquisition: the partitive is licenced by negative polarity, aspectual unboundedness, or unbounded quantity. Possible analogical systems in learners’ L1s expose the alternation to CLI (Spoelman 2013). The data include only texts by advanced learners and come from two corpora: International Corpus of Learner Finnish and Corpus of Advanced Learner Finnish, and L1 reference data. L1s included group in terms of partitivity-related alternation in object marking: 1) highly similar (Estonian); 2) similarities and differences (Polish; Russian); 3) no partitivity-related alternation (English, German). We parsed the data with a Universal Dependencies parser, took a maximum-size balanced random sample of nominal objects (N=298 / L1), and annotated manually the prototypical aspect and the quantifiability. Distributions suggest that learners with partitivity-related object case alternation in their L1 use more partitives than others. A logistic mixed model – (R2: 0.85) with the object case as the response variable; L1, polarity, quantifiability, aspect, wordorder and grammatical voice as fixed predictors; and the informant as a random predictor – shows that while polarity, quantifiability and boundedness generally predict well the alternation, there are L1-specific differences interacting with other variables: mass nouns trigger partitivity less among L1-en learners, the OV wordorder triggers total objects in L1-pl and L1-ru irrespective of the grammatical voice, whereas passive voice triggers total objects in L1-de irrespective of the wordorder. These results suggest that CLI are not only construction-specific but also sensitive to constructional variation of neighboring phenomena.
