EUROSLA XXIV University of YORK CLLR Syntactic, semantic, and discourse constraints in L2 production: the case of English genitive constructions. Francesco Romano Rui Wang and Guo Xiao Sabanci University University College London
[email protected]1 Outline Background Method Results Discussion and Conclusion References 2 Theoretical Background Possessive structure processing during language production At the level of concepts/intentions, POSS might be a semantic function with two arguments, just like transitive verbs (like WARM) are semantic functions with two arguments at this level. The concept POSS activates an abstract lemma in the Lexicon, just like WARM does, except that the semantic roles of Poss’s two arguments are Possessor and Possessum (= thing possessed): Lexicon → ‘xAgent warm yPatient’ ‘xPr Poss yPm’ Pr = possessor, Pm = possessum (adapting Branigan et al., 2007, p.176) 3 Theoretical Background How do possessives occur in language production? Model of language production (Levelt, 1989; Bock & Levelt, 1994), two steps. Step 1: Functional processing Step 2: Positional processing Processing of grammatical information is: Incremental (i.e. one syntactic element at a time) Parallel (i.e. simultaneous processing of syntactic, 4 morphological, phonological information) Theoretical Background Two levels of processing (1) Functional Processing level. Co-indexation of noun lemmas with the arguments of Poss; Pr and Pm functions assigned to lemmas. E.g. a student who rents/owns/lives in a room, the co- indexing produces: ‘studentPr’ ‘xPr Poss yPm’ ‘roomPm’ (2) Positional processing. Poss activates its two associated syntactic frames in the syntax/morphology store: Def N ___ Pr ‘s N ___ Pm 5 Semantic accessibility Example (1)the student’s room > the room of the student Pr[+anim] Poss Pm[-anim] Pm[-anim] Poss Pr[+anim] (2) the cleaner of the hotel > the hotel’ s cleaner Pm[+anim] Poss Pr[-anim] Pr[-anim] Poss Pm[+anim] 6 Interface 1: Conceptual accessibility How is a frame chosen? ‘Interface factors’ can significantly increase the probability with which one syntactic frame is used over the other. Conceptual accessibility (Rosenbach, 2008; Bock & Warren, 1985; Branigan & Feleki, 1999; Branigan, Pickering, & Tanaka, 2008; McDonald, Bock, & Kelly, 1993) 1. Animacy ‘how easily a concept is retrieved from memory’; concept = lexical concept such as lemmas (Branigan 2. Topicality 3. Syntactic Weight et al. 2007, p.172) 7 Conceptual accessibility Conceptual Accessibility as an interface factor: Syntax-Semantics ‘Animate entities are conceptually highly accessible and are therefore retrieved more easily. Because language production is incremental, easily accessed information is processed first; animate entities therefore tend to be privileged during syntactic processes of production.’ (Branigan et al., 2007, p.172) 8 Models of Animacy Effects Grammatical function (GF) vs. Linear Order GF (Bock and Warren (1985), Bock, 1987; McDonald et al. (1993)) • At the functional level, lemmas are co-indexed with the grammatical function Pr or Pm • Functions are hierarchically organised (e.g. Keenan & Comrie, 1977). • Most accessible noun lemma is co-indexed to the highest function on the scale (e.g. Pr), followed by next, incrementally. • Animacy affects functional not positional level. • Animacy affects choice of GF regardless of WO 9 Models of Animacy Effects Linear Order (Rosenbach, 2002, 2005; Sridhar, 1988; Branigan & Feleki’s, 1999) • Only one level of processing, non-incremental • Lemmas are assigned a syntactic position before being assigned a GF (Pr or Pm) • Most accessible lemma claims early position • Animacy affects linear order regardless of GF Is conceptual Which model of accessibility active in animacy effects is advanced L2 more compatible with production? L2 data? 10 Models of Animacy Effects L2 production Advanced Chinese and Arabic L2 speakers of English do not differ substantially from NSs when syntactic frame and animacy of referents is manipulated (Hawkins, Althobaiti, and Ma, 2012) Structure: • Dative sentences with double objects (S V DO DO) • Dative sentences with a DO + PP object (S V DO PP) Findings: • Conceptual accessibility is active in similar ways in L2 production • L2 data on dative sentences are compatible with the linear order model only. 11 Interface 2: Accessibility as priming Syntactic Priming through Discourse Reuse of a structure is more probable if primed in previous production or comprehension (Bock, 1986; Bock & Loebell, 1990; Branigan et al., 2000; Buhler, 1934; Hartsuiker & Westenberg, 2000; Kempen, 1977; Lashley, 1951; Weiner & Labov, 1983) A: I spoke to the personal assistant of the professor this morning B: Did you ask her whether the room-mate of your friend will be allowed to continue his studies? This can be traced to the syntactic form of preceding discourse ‘priming’ subsequent choice of syntactic frame (interface: syntax ↔ discourse). 12 Accessibility as priming Interfaces: Syntax-Semantics-Discourse Interface factors can operate independently, or might interact. E.g. it might be predicted that the ‘s-construction would be more strongly primed in cases where both conceptual accessibility and discourse priming are involved than where discourse priming is not involved. The probability that B will use the ‘s construction might be greater in the first of the contexts below than in the second: A: Where did you stay when you visited the artist’s studio? B: I stayed at the professor’s house (> the house of the professor) A: Where did you stay when you visited the friend of the artist? B: I stayed at the professor’s house (> the house of the professor – weaker preference?) 13 Research Questions An L2 study would need to: (i) determine inherent preferences in utterances where both referents are animate or both are inanimate, comparing L2 speakers with NS; (ii) determine the effect of conceptual accessibility on selection of syntactic frame (syntax-conceptual/intentional interface) and compare with NS; (iii) determine the effect of syntactic priming (syntax-discourse interface); (iv) determine the effect of both conceptual accessibility and priming (a three-way interaction: conceptual/intentional-syntax-discourse interfaces). 14 Predictions Predictions Syntax-Semantics • Animacy effects GFs not linear order • Pr (or Pm) to come first • Higher inversion rates in [-anim] Poss [+anim] than [+anim] Poss [-anim] if animacy effects are active in L2 production Compare to Hawkins’ results in the discussion Say which model is best supported in the discussion 15 Method: Participants L1 L2 N L2 Prof Mean Mean L2 Mean range/me AOE Instr Exposur an (SD) (months) e (months) SP EN 16 33-90%/71 17 102 21.83 (.15) FR EN 14 47-93%/73 11.4 80 6.45 (.14) EN EN 12 99% NA SP IT 17 25-81%/52 33 20 24 (.16) EN IT 15 27-80%/54 33 23 24.5 (.20) IT IT 12 99% NA Note. AOE = age of first exposure. No significant difference between L1 groups for L2 English and L2 Italian (df = 28, t = -.165, p = .87; df = 30, t = .98, p = .92, two-tailed, independent sample t-test). 16 Method: Tests Rationale Exploits topic shift/continuity distinction to elicit overt/null forms and test L1 cross-linguistic influence. IT and SP EN and FR Null=TopC Overt=TopC and TopS Overt=TopS (1) Sentence Completion: Speeded, semi-controlled. (2) Guided Oral Production: semi-free Note: Both tests required use of a picture depicting a busy high street. 17 Method: Rationale Given Given the the number number of of test test items items for for each each Topic Topic type type differed, differed, Sentence Completion results results for for will will not not be be compared. compared. 14 items testing TopS, 4 testing TopC. mask property via (1) distracter items, (2) speededness, and (3) cotesting. (1) 6 distracter words which needed to be ignored were present in a wordbank. (2) speeded: strict time limit established against native controls’ to complete test (3) cotesting: each TopS items with inflection, TopC with WO and inflection. 18 Method: sample items Topic Shift [Item 8] Nella figura c’e’ un taxi[Top1] e un uomo[Top2] in una cabina telefonica. Il taxi sta fermo. ________________ al telefono. In the picture there’s a taxi[Top1] and a man [Top2] in a phone booth. The taxi is still. ____________________ on-the phone. “The taxi is still. The man is on the phone” 19 Method: sample items Topic Continuity [item 6] I trasporti in città non sono efficienti il weekend. Molte persone[Top] si lamentano perche’ ci sono pochi mezzi. ___________ scontente il Sabato e la Domenica. “Sono sempre scontente il Sabato e la Domenica” Transportation in the city is not efficient on the weekends. Many people{Top} complain because there are not enough buses and trains. _____________unhappy on Saturdays and Sundays. “They are always unhappy on Saturdays and Sunday” 20 Method: Rationale Guided Oral Production participants describe the same picture used for the sentence completion elicited both null and overt subjects only topic shift/continuity contexts were analysed in a manner consistent with the sentence completion Note. only a subset of the sample completed this test due to restrictions in language school and length of experiment (included another 2 tests not reported here) 21 Results: Sentence Completion Topic Shift (SP-IT) Null Overt Missing 0.81 0.75 0.71 0.71 0.65 0.58 0.55 0.55 0.51 Learners ordered by Proficiency => 0.51 0.48 0.41 0.38 0.38 0.35 0.35 0.25 0.00 0.20 0.40 0.60 0.80 1.00 % use 22 Results: Sentence Completion Topic Shift (EN-IT) At At lower lower levels levels of of proficiency, proficiency, Null Overt Missing EN-IT EN-IT overextend overextend null null forms forms 0.8 to to topic topic shift shift contexts, contexts, at at rates rates Learners 0.8 ordered by Proficiency => (65%) (65%) higher higher than than the the L1 L1 SP SP 0.8 (35%). (35%). This This is is in in spite spite of of L1-L2 L1-L2 0.8 match match for for interface interface constraints constraints 0.7 0.633333333333333 0.5667 0.566666666666667 0.466666666666667 0.4 0.4 0.3 0.3 0.266666666666667 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 % use 23 Results: Sentence Completion Topic Continuity (SP-IT) Null Overt Missing 0.81 0.75 0.71 0.71 0.71 0.65 0.58 0.55 0.55 Learners 0.51 ordered by Proficiency 0.48 0.41 0.38 0.38 0.35 0.35 0.25 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 % use 24 Results: Sentence Completion Both Both EN-IT EN-IT and and SP- SP- IT IT overuse overuse overt overt forms, Topic Continuity (EN-IT) forms, even even though though the the L1 L1 SP SP only only can can Null Overt Missing transfer transfer interface interface constraints. constraints. 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 25 Results: Sentence Completion Topic Shift (SP-EN) Null Overt Missing 0.9 0.833333333333333 0.8 0.8 0.8 0.8 0.8 0.766666666666667 Learners ordered by Proficiency 0.766666666666667 0.766666666666667 0.733333333333333 0.733333333333333 0.566666666666667 0.533333333333333 0.466666666666667 0.333333333333333 0.00 0.20 0.40 0.60 0.80 1.00 % use 26 Results: Sentence Completion Topic Shift (FR-EN) Null Overt At At low low proficiency Missing proficiency levels levels SP- SP- EN EN overextend overextend null 0.933333333333333 null forms forms at at the the rate rate of 0.9 of 43% 43% andand 57% 57% in in Topic Topic Shift Shift contexts 0.9 contexts compared compared 0.866666666666667 to to 7% 7% inin the the FR-EN FR-EN group. group. 0.766666666666667 This This occurs occurs in 0.733333333333333 in spite spite of of the the L1- L1- L2 L2 interface interface 0.7 constraints constraints Learners ordered by Proficiency matching for matching 0.7 for both both L1s. L1s. 0.666666666666667 0.666666666666667 0.666666666666667 0.6 0.466666666666667 0.00 0.20 0.40 0.60 0.80 1.00 % use 27 Results: Sentence Completion Topic Continuity (SP-EN) Null Overt Missing Learners ordered by Proficiency 0.9 0.833333333333333 0.8 0.8 0.8 0.8 0.8 0.766666666666667 0.766666666666667 0.766666666666667 0.733333333333333 0.733333333333333 0.566666666666667 0.533333333333333 0.466666666666667 0.333333333333333 0.00 0.20 0.40 0.60 0.80 1.00 % use 28 Results: Sentence Completion Topic Continuity (FR-EN) Null Overt Missing FR-EN FR-EN hardly hardly ever ever use use null null 0.933333333333333 forms forms in in either context. context. SP-EN either 0.9 SP-EN use use null null forms forms 0.9 in in Topic Topic Continuity Continuity consistent consistent with 0.866666666666667 with the the L1 L1 (not (not L2) L2) interface, 0.766666666666667 interface, especially especially at at intermediate intermediate 0.733333333333333 Learners ordered by Proficiency levels. 0.7 levels. 0.7 0.666666666666667 0.666666666666667 0.666666666666667 0.6 0.466666666666667 0.00 0.20 0.40 0.60 0.80 1.00 % use 29 Summary of Results (1) Sentence Completion Individual EN-IT participants of low proficiency overextend null forms to Topic Shift contexts at a higher rate than the L1 SP (65 vs. 35%) despite L1-L2 interface constraints match for both L1s; Both EN-IT and SP-IT overextend overt forms in Topic Continuity contexts, even though the L1 SP can transfer interface constraints; At low proficiency levels individual SP-EN overextend null forms to Topic Shift contexts more than the FR-EN despite the L1-L2 interface constraints match for both L1s ; FR-EN hardly ever use null forms in either context; SP-EN use null forms in Topic Continuity, especially at low/intermediate level, consistent with the L1 interface. Conclusion 1: There is cross-linguistic influence. When interfaces differ between L1 and L2, this is reflected by overextension of either form, especially at low and intermediate levels, and by the different ‘developmental’ patterns between L1 groups for each L2. 30 Summary of Results (2) Sentence Completion Both SP-IT and EN-IT use null forms when required (D in AGR) SP-EN and FR-EN use overt forms when required (no D in AGR) Conclusion 2: Syntactic features are acquired earlier (syntax- before-discourse) 31 Results: L2 IT Oral task Topic Shift EN-IT Topic Shift SP-IT 100 Null Overt Other NullLearners Overtuse Learners some some null use Other null forms forms 90 15 in in Topic Topic Shift Shift contexts contexts butbut 29 27 31 100 comparatively comparatively less less than than the the 80 24 other other test. test. 70 80 38 49 60 50 60 40 85 67 71 69 40 71 30 62 20 51 20 10 0 0 0 6 0 0 6 0 0 0.4 0.63 0.8 0.8 0.47 0.67 0.93 32 Results: L2 IT Oral task Topic Cont EN-IT Topic Cont SP-IT Null Overt Other Null Overt Other Again Again these these results results parallel parallel 100 100 those those in in the the other other test. test. 11 14 10 90 25 90 17 23 80 10 80 33 29 20 7 70 11 70 60 61 60 17 50 50 40 76 40 64 63 70 30 62 30 50 20 20 28 10 10 0 0 0.4 0.63 0.8 0.8 0.47 0.67 0.93 33 Results: L2 EN Oral task Table. Overall Use of null and overt forms L2 L2 English English use use of of subjects subjects is is mostly mostly target-like target-like in in the the oral oral L1 Prof % Null Overt data. Total data. The The rare rare null null forms forms Counts % Counts % produced Counts are produced % used are used in in Topic Topic Continuity Continuity contexts contexts oror in in place place 90 2 3 65 97of of expletives. 67 100 expletives. SP 86 0 0 50 100 50 100 83 0 0 58 100 58 100 76 0 0 43 100 43 100 FR 70 2 2 83 98 85 100 70 1 2 71 97 78 100 66 1 3 36 97 37 100 34 Results: Oral task Examples. (1) I think they[Top] are shouting at each other. Now Ø watching the train, maybe they are speaking about this or (?). (FR-EN Prof 66%) (2) a. The kid[Top] is riding a bike, Ø is on the right of the picture, (FR-EN Prof 70%) b. Uhm the two boys[Top] are here, Ø are pointing for something. (FR-EN Prof 70%) c. But Øexpl is this poor guy, sarkozy[Top] take all the power so he don’t use it. (FR-EN Prof 70%) (3) He[Top] is looking for a job. Øexpl looks to me like Ø[Top] looking for a job. (SP-EN Prof 90%) 35 Summary of results Semi-guided oral task Both L2 IT groups use some null forms in Topic Shift contexts though to a less extent than the other test. Both L2 IT groups overextend overt forms as found in previous research. One L1 EN participant produced up to 61% overuse in one case. Intermediate/advanced L2 EN learners choices are native-like, in contrast to advanced L2 IT learners who show optionality. Most participants to the oral task were intermediate/ advanced, thus more research comparing patterns with oral data of lower levels is needed. 36 Discussion and Conclusion 1. What is the role of cross-linguistic influence when interfaces are considered? The data suggests that when L1-L2 interfaces do not match (i.e. EN-IT or SP- EN), both null and overt forms are inappropriately used. As expected by Sorace: • In L2 IT, though, overt forms are overused regardless of L1 • The syntactic features tied to null subject use is acquired earlier than the conjoining pragmatics 37 Discussion and Conclusion 2. Is the null form really licensed by default (universally available)? (Liceras & Diaz, 1999) Maybe. Although L2 IT learners know null subjects are possible, they take “time” to adjust the distribution of subject forms in general if the L1-L2 do not match. The same can be said for L2 EN. On this basis, it may be that the presence of one/both forms in finite clauses in the input is sufficient to clue learners into the forms allowed. See White (1989) for claims that the default is the overt form for learnability reasons (Talked with Ayse). 38 Discussion and Conclusion 3. What happens beyond the endstate? (White, 2011) Lower level learners commit more overextension errors as they are still in the process of figuring out how interface constraints operate (recall EN- IT and SP-EN). Convergence on L2 interface is gradual except for overuse of overt forms in L2 IT. 39 END Thank you!
[email protected] [email protected]40 References Sorace, A. (2010). SLA as bilingualism: or, it’s time to see the forest for the trees. Paper presented at the EUROSLA 20. Liceras, J. M., and Diaz, L. (1999). Topic-drop versus pro-drop: null subjects and pronominal subjects in the Spanish L2 of Chinese, English, French, German and Japanese speakers. Second Language Research 15, 1, 1-40. Sorace, A., & Filiaci, F. (2006). Anaphora resolution in near-native speakers of Italian. Second Language Research, 22, 339–368. Tsimpli, T. Sorace, A., Heycock, C. & Filiaci, F. 2004. First Language attrition and syntactic subjects: a study of Greek and Italian near native speakers of English. International Journal of Bilingualism 8, 257-277. 41