1 Introduction

We investigate the representational changes that may occur in late bilinguals’ production of passives, both as a function of L2 proficiency and of cross-linguistic overlap. Cross-linguistic priming tasks, where participants process prime sentences in one language and respond in another language, are increasingly used to investigate the nature, processing, and development of syntactic representations in late bilinguals (Hartsuiker et al., 2017). We ask whether a structure that exists only in the learner’s L1, such as the Italian si-causative passive (see Table 1), can prime a structure in the L2 that is not directly parallel. According to the Shared Syntax Account (Hartsuiker et al., 2017; Hwang et al., 2018; Son, 2020), priming may still occur in such circumstances. Whether L1- and L2-specific syntactic variations and L2 proficiency modulate priming across languages remains controversial (Son, 2020, p. 1244). We examined cross-linguistic priming of target passives in English when primed by Italian “venire passives” (similar to English passives) and by Italian “si-causative passives” (which English does not have). Previous within-language studies in Italian have found that venire passives prime si-causative passives in children (Manetti & Belletti, 2015). Whether the converse holds in adults is unknown.

2 Method

2.1 Participants

84 late bilinguals (Italian L1 English L2) were recruited and tested remotely. Participants were asked to fill in a questionnaire to assess their linguistic background and English proficiency. We asked participants to self-rate their writing, speaking, reading and comprehension abilities on a 1-5 Likert scale and to assess their English level according to the CEFR. For the analysis, we sum the scores of writing, speaking, reading and comprehension and the score corresponding to the CEFR level they indicated, then the mean was computed and put in the model as a continuous fixed effect.

2.2 Materials

Participants read and copied 24 primes and 24 fillers in Italian and subsequently described target pictures in English. For the prime items, we used Italian (i) active, (ii) venire-passive (overlapping condition) and (iii) si-causative passive sentences (L1-specific condition). The target pictures illustrated black-and-white actors performing transitive actions and appeared on the screen with an English verb underneath. Participants were explicitly instructed to copy out the prime items in Italian, and then to describe the target picture in English. Importantly, there was no lexical overlap between prime items and target stimuli.

Tab.1. Material examples
PrimeType	Example
active prime	Il papà veste il figlio (Eng. The dad dresses his son)
venire-passive prime	Il figlio viene vestito dal papà (Eng. The child gets dressed by his dad)
si-causative passive prime	Il figlio si fa vestire dal papà (Eng. The child gets his dad to dress him)

3 Results

Data were analyzed in R using the glmer() formula from the lme4 package (Bates et al., 2015). We predicted the log-odds of English passive production (binary coded) as an effect of prime condition, participants’ proficiency and item order. We allowed for random slopes for items and participants.

	Lax Pass
Predictors	Odds Ratios	CI	p
(Intercept)	0.01	0.00 – 0.03	<0.001
prof z	1.67	0.91 – 3.07	0.100
condition e1	1.28	0.97 – 1.68	0.085
condition e2	0.92	0.69 – 1.22	0.549
ItemOrder	1.03	1.00 – 1.06	0.056
prof z * condition e1	0.96	0.73 – 1.25	0.744
prof z * condition e2	0.90	0.68 – 1.19	0.455
Random Effects
σ²	3.29
τ₀₀ _ID	5.17
τ₀₀ _ItemNumb	0.51
ICC	0.63
N _ItemNumb	24
N _ID	84
Observations	1899
Marginal R² / Conditional R²	0.037 / 0.647

3.0.1 English Passives predicted by Prime Condition

We used contrast coding to compare venire passives to si-causative passives and active primes. We found no difference between si-causative passives (L1-specific condition) and active primes (β = -0.08, SE = 0.14, p = 0.54), whereas the effect of venire passives (overlapping condition) was only marginally significant (β = 0.2, SE = 0.14, p = 0.08), increasing the log-odds of English passive production irrespective of proficiency (see Figure 1).

3.0.2 English Passives predicted by Prime Condition and Proficiency

We found a marginally significant effect of proficiency (β = 0.5, SE = 0.4, p = 0.09), yet not in its interaction with prime condition. As Figure 2 illustrates, participants with higher proficiency were more likely to produce passives when asked to describe pictures in English.

4 Discussion

Our aim was to determine the extent of shared cross-linguistic representations in a case of L1-specific syntactic variation.

Can we observe cross-linguistic priming when L2 learners are primed with L1-specific syntactic conditions?

We did not find a strong priming effect in the L1-specific condition (si-causative passive), nor the priming effect was substantial in the overlapping condition (venire passive). One possibility is that si-causatives were not processed semantically as passives, perhaps because the subject has agentive and volitional properties associated with the initiator role, therefore overlapping with actives. In our data, venire passives show an increasing effect of cross-linguistic priming as a function of item order, as opposed to actives and si-causative passives that show an identical effect in this respect (see Figure 3).

This result further suggests that actives and si-causative passives might be processed similarly. If this hypothesis is confirmed, then English passive sentences were not primed equally to actives, resulting in a weak priming effect following venire passive primes. Additionally, it is also plausible to assume that what gets primed in cases of language-specific syntactic conditions is the event encoded in the sentence, rather than the syntactic structure per se.

The second goal of the present study was to investigate the role of participants’ L2 proficiency.

Our results are in line with the Shared Syntax Account (Hartsuiker et al., 2017). This model predicts different developmental stages of abstract representations as a function of increasing L2 proficiency. Our results are in line with Hartsuiker’s et al.’s (2017) predictions.

4.1 Follow-up study: Semantic interpretation task

To explore the hypothesis that si-causative passives might be processed more similarly to active than to passive sentences, we conducted a semantic interpretation task with si-causative and regular transitive passives, asking a different group of participants (N=72) to identify the agent and patient role. We found that Italian native speakers are less accurate and slower in assigning thematic roles in si-causative passives relative to transitive passives. This suggests that when processing si-causative passives, participants might use the Agent-first strategy (Ferreira, 2003) hence assigning the agent role to the first NP of the sentence and processing these structures as actives rather than passives.

	Accuracy
Predictors	Odds Ratios	CI	p
(Intercept)	31.73	14.16 – 71.11	<0.001
SyntacticStructure [TransitivePassive]	4.05	2.05 – 8.02	<0.001
MetaLing	0.67	0.36 – 1.23	0.198
Decision [Patient]	1.95	1.04 – 3.67	0.038
SyntacticStructure [TransitivePassive] * MetaLing	1.86	1.16 – 2.99	0.010
SyntacticStructure [TransitivePassive] * Decision [Patient]	0.43	0.17 – 1.13	0.086
Random Effects
σ²	3.29
τ₀₀ _ID	3.49
τ₀₀ _Item	0.29
ICC	0.53
N _ID	66
N _Item	32
Observations	1779
Marginal R² / Conditional R²	0.054 / 0.560

5 References

Hartsuiker, R., Bernolet, S. (2017). The development of shared syntax in second language learning. Bilingualism: Language and Cognition, 20(2), 219-234. doi:10.1017/S1366728915000164;
Belletti, A. (2020). (Reflexive) Si as a route to passive in Italian. In L. Franco, & P. Lorusso (a cura di), Linguistic Variation: Structure and Interpretation (pp. 73-86);
Hwang, H., Shin, J.‐A. and Hartsuiker, R.J. (2018). Late Bilinguals Share Syntax Unsparingly Between L1 and L2: Evidence From Cross Linguistically Similar and Different Constructions. Language Learning, 68: 177-205. doi:10.1111/lang.12272;
Manetti C., and Belletti A. (2015). Causatives and the acquisition of the Italian passive, Proceedings of Generative Approaches to Language Acquisition 2013, Oldenburg;
Son, M. (2020). Cross-linguistic syntactic priming in Korean learners of English. Applied Psycholinguistics, 41(5), 1223-1247. doi:10.1017/S0142716420000545

Can Italian si-causative and venire passives prime English passives in late learners?

Ilaria Venagli (Universität Konstanz), Michaela M. Vann (Ca Foscari University of Venice), Giulia Bencini (Ca Foscari University of Venice), Virginia Valian (Hunter College NYC & CUNY Graduate Center, NYC)