Studies of adult second language (L2) acquisition often report on difficulties foremost with phonology (producing and processing non-native phonemes) and grammar (producing and processing non-native word order or grammatical rules). However, studies have also shown that L2 learners are challenged when word meanings in L2 differ from that of their first language (L1), resulting in crosslinguistic influence (CLI) in speech production (e.g., Viberg, 1998, 1999).

When investigating CLI in processing, previous neurocognitive studies utilizing event-related potentials (ERPs) have mainly focused on syntax and morphosyntax (e.g., Andersson, Sayehli, & Gullberg 2019; Carrasco-Ortíz et al. 2017; Gillon Dowens et al. 2011; Sabourin and Stowe 2008; Sabourin, Stowe, & de Haan 2006; von Grebmer zu Wolfsthurn, Pablos Robles, & Schiller 2021). These studies have shown positive effects of CLI such that for instance when processing gender agreement only learners with gender agreement in L1 process L2 gender agreement similarly to native speakers of the language. The current study 1) extended previous results on CLI in speech production concerning verb semantics to neurophysiological processing of the same, and 2) extended previous ERP studies of CLI to semantics.

We investigated how L2 learners rate and process fine-grained L2 verb semantics (placement), and specifically the impact of if their L1 have similar or non-similar semantic categories. More specifically, we examined English and German L2 learners of Swedish and native Swedish speakers and their online neurophysiological processing (ERPs) and offline appropriateness ratings of three Swedish placement verbs obligatory for placement supported from below: sätta ‘set’, ställa ‘stand’, and lägga ‘lay’. The learners’ L1s differed from Swedish in that their placement verbs either shared or did not share semantic characteristics with the target language. English has a general placement verb put, whereas German has specific verbs similar (but not identical) to Swedish, stellen ‘set/stand’, and legen ‘lay’.

We had previously developed visual stimuli (images of placement events, i.e., objects being placed on a table) and in this study we developed auditory stimuli that was combined with these images. The objects differed in shape; with or without a functional base (e.g., cube and orange), symmetric or asymmetric (e.g., cube and candle), and for the asymmetric objects—in orientation against the ground; vertical (resting on its functional base) or horizontal (resting off its functional base). The auditory descriptions of the placement events were recorded in an anechoic chamber at Lund University Humanities Lab with a trained speaker allowing for excellent quality of the sound files.

All sentences describing the placement were presented in the present tense (sätter, ställer, and lägger) to keep syllable structure and word length comparable. The target clause was preceded by a contextual clause: Hon tar X och X den/det på bordet ‘She takes X.DET [the object on the picture] and X:s [one of three placement verbs] it [den/det depending on the gender of the object] on table.DET’. The images were used both for ERP recordings and in a computerized offline appropriateness rating task (google.docs). For the ERP recordings we time locked the continuous electroencephalogram to the verb in the auditorily presented sentences such that we compared the processing to the exact same verb but in combination with different images of placement events (e.g., ställer kuben ‘stands the cube’ and *ställer apelsinen ‘stands the orange’). In the offline ratings participants had to judge how appropriate the verb in the written version of each sentence was for describing the placement event depicted on a 6-point Likert-scale with 1 indicating least appropriate and 6 most appropriate.

Due to Covid and other unforeseen reasons our participant group was limited in size thus not allowing for the fine-grained comparisons we had planned for. The final participant group consisted of a total of 57 participants of which 20 were English learners, 21 German learners, and 16 native Swedish speakers, all matched on age (F(2, 54) = 1.22, p = .303). In addition, the two learner groups were matched on age of acquisition of L2 Swedish (t < 1), length of exposure to the L2 (t < 1), and on L2 proficiency (t < 1).

In contrast to previous ERP studies that generally show no qualitative differences in L1 and L2 semantic processing (Ardal et al. 1990; Brown, Hagoort, and Chwilla 2000; Mills, Coffey-Corina, and Neville 1993), we found differences depending on L1. Importantly, previous studies have not examined CLI on semantics, but rather cases where the semantics are assumed to be equivalent across languages, such as in open class words appearing in a congruent or incongruent sentence context (e.g., ‘the pizza was too hot to cry’; Kutas & Hillyard, 1980). These studies have also typically only examined learners from one L1.

In the current study, ERPs were recorded while participants watched the still images of the objects being placed on a table and listened to the sentences that described the event with verbs that matched the image or not (e.g., lägger apelsinen ‘lays the orange, *ställer apelsinen ‘stands the orange’). Participants also performed the offline appropriateness rating task. Even though learners as well as native speakers showed great individual variability in the ratings and in the ERP responses both tasks revealed CLI. English learners’ appropriateness ratings of atypical verb use differed from those of both native Swedish speakers’ and German learners, with no difference in the latter pair. Similarly, German learners’ ERP effects were more similar to those of the native Swedish speakers (increased lateral negativity to atypical verb use) than to those of the English learners (increased positivity to atypical verb use). The results thus reveal CLI both offline and online with similarity between L1 and L2 leading to more similar processing and judgements, in line with the production findings reported in previous work. Importantly, the combination of measures (ERPs and appropriateness ratings) highlighted different aspects of CLI.

In conclusion, this study has two specific outcomes. First, even when fine-grained verb semantics differ between the L1 and the L2, learners are able to acquire such distinctions and apply the knowledge in offline judgements, especially in cases where scenarios represent typical verb use in the target language. Second, semantic similarity between the L1 and the L2 affects both offline and online performance. Learners with similar semantic distinctions in their L1 (German learners) both judge and process L2 verbs more similarly to native speakers (of Swedish) than learners whose L1 lack similar semantic distinctions (English learners). Importantly, in comparison to previous studies of semantic processing, we have shown the processing of placement verb semantics to be qualitatively different depending on language background. There is thus evidence of CLI in the dealing with fine-grained verb semantics in L2 both offline and online.

Results from the study has been presented at national and international conferences, and to teachers during conferences and invited talks with focus on L2 acquisition, articles in open access journals are in print, under revision, and under preparation. We are currently planning on an extension to our investigations into CLI in semantic processing to learners of different ages to explore age of acquisition in relation to CLI. In addition, if funded, we will explore the role of type of exposure that relates to procedural and declarative memory systems in learning the semantic distinctions and processing of these verbs.