Session 3C

Anaphora resolution in null subject L2s has received considerable attention, relating to the syntax/discourse interface (Sorace & Filiaci 2006), processing (Sorace 2011) and prosody (White et al. 2017). In contrast, pronoun interpretation in L2s like English has been less investigated. We examine the effects of stress on interpretation of English pronouns by Italian speakers.
In Italian sentences like (1a), the antecedent of a null pronoun is typically the subject of a higher clause (Monica), whereas unstressed overt pronouns prefer object antecedents (Claudia) (Carminati 2002). In contrast to Italian, unstressed pronouns in English favour subject antecedents; see (1b). Pronoun stress results in an 'overturning' effect in both languages: a reduction in object antecedents in Italian, an increase in object antecedents (away from subjects) in English (see 1a/b and Table 1).
(1a) Monicai ha telefonato a Claudiaj quando Øi/leij/LEIi/j era in ufficio.
(1b) Monicai phoned Claudiaj when shei/SHEj was in the office.
As Table 1 indicates, Italian speakers must establish a correspondence between English unstressed pronouns and Italian null pronouns and between English stressed pronouns and Italian unstressed pronouns. Following previous research demonstrating effects of stress on Italian pronoun interpretation (Gargiulo 2020; White et al. 2017), we hypothesize that Italian-English L2ers will come to realize that stressed pronouns in English shift the interpretation away from subject antecedents, like unstressed pronouns in Italian. 
We report on two experiments, administered online. 21 Italian-speakers (intermediate/advanced English proficiency) participated in an experiment on English; 18 Italians participated in an experiment on Italian. Each experiment involved auditorily-presented biclausal sentences (as in (1)) (24 in English, 36 in Italian), manipulating pronoun type and presence/absence of stress. Stimuli were preceded by written contexts introducing potential referents (main clause subject, object or external). After listening to a sentence, participants indicated their preferred referent for the pronoun. See (2).
Results for L2 English (Figure 1) show that subjects were the preferred antecedents for unstressed pronouns; stress led to a significant increase in object antecedent choices. For L1 Italian (Figure 2), participants showed a significant difference between null and overt pronouns, subject antecedents being preferred for null pronouns and object antecedents for overt pronouns. Stress was also significant, reducing object preferences and increasing the proportion of other antecedents (subjects or external). 
The results suggest that, as hypothesized, L2ers are sensitive to differences in pronoun interpretation between the two languages. Notably, the differential role of stress is well understood: stress on L2 English pronouns resulted in an increase in object choices (compared to unstressed pronouns), whereas stress on pronouns in L1 Italian resulted in a decrease.


References
Carminati, M. N. (2002). The processing of Italian subject pronouns. PhD dissertation, UMass Amherst.
Gargiulo, C. 2020. On L1 attrition and prosody in pronominal anaphora resolution. PhD dissertation, Lund University.
Sorace, A. 2011. Pinning down the concept of "interface" in bilingualism. LAB 1: 1-33.
Sorace, A. & F. Filiaci. 2006. Anaphora resolution in near-native speakers of Italian. SLR 22: 339-368.
White, L. et al. 2017. Prosodic effects on pronoun interpretation in Italian. BUCLD 41: 744-752.

Prosodic boundary is marked by the presence of several acoustic cues, such as pitch, final lengthening and pause. These acoustic cues are used differently in different languages. A core question in speech acquisition research is how learners exploit L2 prosodic cues to constrain syntactic ambiguity.
In this study, 40 French noun phrases belonging to two different syntactic categories (direct object or subject) were inserted within locally ambiguous sentences that differed in late / early closure (e.g., Whenever the snake {was eating the rat, the rabbit would hide. (LC) / was eating, the rat would hide. (EC)}). These sentences were produced by a French male speaker. Acoustic analyses showed that he produced reliable French prosodic cues to differentiate the ambiguous meanings. Forty intermediate (n=20, ma: 22, sd: 2.7) and advanced (n=20, ma: 24.2, sd: 2.4) Chinese L2 learners were tested whether they could correctly assign the ambiguous items (here "the rat") to their syntactic categories based on available prosodic cues. The sentences were cut after the target nouns and divided into 2 blocks: each member of a given pair appeared in a different block. In each block, we had 20 experimental stimuli (10 LC and 10 EC) plus 5 filler sentences. Participants were asked to listen to each stimulus (e.g., "Whenever the snake was eating (,) the rat") and to complete it by writing the rest. The completed sentences were coded as to whether ambiguous items were interpreted as subjects or as direct objects.
Our results showed that Chinese learners had difficulties in solving the syntactic ambiguity: they gave more LC responses (i.e., ambiguous items interpreted as direct objects) both in LC (9.85 correct responses in advanced learners vs. 9.80 in intermediate learners) and in EC (6.15 vs. 5.65 in advanced vs. intermediate learners) conditions. Two ANOVAs were conducted on the number of correct responses given in LC and EC, with proficiency level (intermediate vs. advanced) as between-participants factor and blocks as within-subject factor (block 1 vs. 2). The analyses revealed that neither proficiency level (F(1,36)=0.17, p=0.68 for LC; F(1,36)=0.19, p=0.67 for EC) nor block (F(1, 36)=1.53, p=0.22 for LC; F(1, 36)=0.27, p=0.61 for EC) had significant effect on the scores obtained.
Our results could be interpreted within the Informative Boundary Hypothesis (IBH) (Carlson et al., 2001, 2009; Clifton et al., 2002) and the Late Closure Preference (Frazier, 1979). According to the IBH, the effectiveness of a prosodic boundary is determined by its size relative to relevant earlier and global prosodic boundaries in the utterance. Our stimuli were rather short (about 3-4s) and incomplete, which did not give enough prosodic information. The lack of prosodic context made learners insensitive to the prosodic boundary cues present in the signal. Our results provide additional evidence for the Late Closure strategy being favored in syntactic parsing (Frazier, 1979). For ambiguous sentences, parsers attach the new information to the clause being processed. The noun phrases in LC and EC conditions would therefore more likely be attached to the preceding verb and interpreted as direct objects.

The encoding of motion events is known to be challenging for second language (L2) users, particularly if the lexicalization patterns of their first language (L1) diverge from those of the L2 (e.g., Treffers-Daller & Tidball 2016). A major distinction is typically made between verbframed (V) and satellite-framed (S) languages (Talmy 1985). In V-languages, path, is expressed in the main verb (e.g., to cross, to enter). Manner is only expressed, e.g., in gerunds (e.g., enter running), if particularly salient. By contrast, S-languages are manner-salient (Slobin 2004), the root of the finite verb typically encoding manner information (e.g., to march, to run, to scuttle), while path is expressed in satellites (e.g., directional adverbs, verbal prefixes and particles, or prepositional phrases).
German is assumed to be a rather typical S-language with dense information packaging (Author1 et al. 2017). Given the broad range of path encoding options in German (Autor1 & Author2 under review), L2 users of German experience substantial levels of variation within the S-framed basic pattern. The challenge of learning to encode path in motion events in L2 German might therefore go way beyond the path-manner dichotomy as typically discussed (e.g., Cadierno 2008). So far, only a few studies have empirically investigated L2 German path encoding in motion event descriptions in detail (e.g. Scheirs 2015). We therefore ask the following research questions: How do L2 users of German with different L1s deal with the large array of path encoding options? What can error patterns tell us about their ability to differentiate between satellite types, to acquire their respective form-meaning mappings, usage restrictions, and combinatorial potential?
This presentation analyzes oral and written motion event descriptions produced by advanced L2 users of German (L1 English: n=6; L1 Danish: n=16; L1 French: n=6; L1 Spanish: n=14). We show that advanced L2 users, independently of having a V- (French, Spanish) or S-framed L1 (Danish, English), particularly struggle with the "smaller" devices of path encoding, such as particles, locative and directional adverbs, their formal and functional discrimination (and their differentiation from prepositions), and the identification of their combinatorial potential and restrictions. We discuss characteristics of the target language input that might contribute to these challenges as well as implications and options for L2 teaching.
References
Author2 et al. 2017
Author1 & Author2, under review
Talmy, L. (1985). Lexicalization patterns. Semantic structure in lexical forms, in T. Shopen (ed.), Language typology and syntactic description: Grammatical categories and the lexicon, Cambridge University Press, 57-149.
Slobin, D. I. (2004). The many ways to search for a frog. Linguistic typology and the expression of motion events, in S. Strömqvist and L. Verhoeven (eds.), Relating Events in Narrative. Volume 2: Typological and Contextual Perspectives, Lawrence Erlbaum, 219-257.
Treffers-Daller, J., and Tidball, F. (2016). Can L2 learners learn new ways to conceptualize events? A new approach to restructuring in motion event construal, in P. Guijarro-Fuentes, K. Schmitz & N. Müller (eds.), The Acquisition of French in Multilingual Contexts. Multilingual Matters, 145-184.

Word lists grounded in principles of frequency and/or functionality have been advocated in instructed SLA since the early 20th century (González-Fernández & Schmitt 2017), though they have at times proved controversial (Coste 2006). Policy-makers in England are currently adopting a frequency-led approach for foreign languages in schools; a recent specification (DFE 2022) requires that learners aged 16 have a vocabulary of 1700 items, and that 85% of lexis to be taught be drawn from the "2000 most frequent words" in the language.
Vocabulary theorists concur that instructed learners require a mix of intensive focused instruction and extensive exposure to build vocabulary (Webb 2020). But little is known about the lexis to which young instructed learners are actually exposed, and how far authentic classroom discourse reflects the frequency distributions in reference corpora.
This paper explores relationships between lexical frequency in a reference corpus (RC) for French developed for educational purposes (Lonsdale & Le Bras 2009), and an input corpus (IC) of authentic classroom L2 French. IC comprises a 33-hour sequence of French lessons recorded with a single class of 7-8 year old beginners (AUTHOR 2012, 2019). All speech was transcribed and analysed with CHAT/CLAN (MacWhinney 2000); analysis showed that the learners were exposed over time to just under 700 word types (with token frequencies ranging from many hundreds, to single occurrences). Using the tool MultiLingProfiler (Finlayson 2021), frequency band positions according to RC were calculated for the lexis of IC. This analysis showed that 61.0% of input items fell within the 2000-word band in RC; detailed analysis showed these items were well distributed across word classes, and the commonest RC function words were well represented. However, IC also included many less frequent words, e.g. 16.5% of word types had frequencies below 1 in 5000 words. These "rare" words were mainly nouns and adjectives, relating to particular topics which the teacher used to motivate language use (e.g. animals, body parts, foods). Further analysis of learner achievement data and of instructional strategies showed that such words were just as learnable as more frequent words, given reasonable input frequency, and multimodal support.
Conclusions are drawn regarding the relevance of reference corpora for early instructed SLA, and the design features which may maximise their usefulness.


References
AUTHOR 2012
AUTHOR 2019
Coste, D. (2006). Français élémentaire, débats publics et représentations de la langue. Documents pour l'histoire du français langue étrangère ou seconde, 36. 
Department for Education. (2022). French, German and Spanish: GCSE subject content. https://www.gov.uk/government/publications/gcse-french-german-and-spanish-subject-content
Finlayson, N., Marsden, E., Anthony, L., Bovolenta, G., & Hawkes, R. (2021). MultiLingProfiler (Version 2) [Computer software]. University of York. https://www.multilingprofiler.net/
González-Fernández, B., & Schmitt, N. (2017). Vocabulary acquisition. In S. Loewen & M. Sato (Eds.), The Routledge handbook of instructed second language acquisition (pp. 280-298). Routledge.
Lonsdale, D. & Le Bras. Y. (2009). A frequency dictionary of French: Core vocabulary for learners. Routledge.
MacWhinney, B. (2000). The CHILDES project: Tools for analyzing talk. 3rd Edition. Lawrence Erlbaum Associates.
Webb, S. (Ed.) (2020).  Routledge handbook of vocabulary studies. Routledge.