The Subject Gap Advantage in Georgian relative clause processing - PowerPoint PPT Presentation

The Subject Gap Advantage in Georgian relative clause processing Steven Foley • srfoley@ucsc.edu North East South Caucasian Chalk Circle (NESCCC) May 9, 2017

Introduction Relative clauses with subject gaps ( SRC s) are generally easier to process than ones with object-gaps ( ORC s). (1) the painter [ RC who __ inspired the writer ] SRC (2) the painter [ RC whom the writer inspired __ ] ORC Evidence: acquisition, aphasia, ERPs, reading times, eye movements, comprehension, acceptability, disambiguation bias (see Gibson 1998, Kwon et al 2010 for reviews) . 1

Introduction What might explain this Subject Gap Advantage (SGA)? • Hierarchical structure? • Incremental cues from morphological case? • Linear/temporal distance? 2

Introduction RC-processing data has primarily come from N OM –A CC languages with postnominal RCs (cf. Anand et al 2011) . T ypological confound: the three hypotheses converge given a language with these properties. Enter Georgian  , a split-ergative language with pre- and post- nominal RCs. A perfect storm for disentangling SGA theories! 3

Present study T wo self-paced reading experiments • Experiment 1: Case alignment pattern N OM –A CC vs. E RG –A BS vs. D AT –A BS ⎛ ⎞ • Experiment 2: Relative clause position ⎜ ⎟ N [ RC … ] vs. [ RC … ] N ⎝ ⎠ 4

Present study Both experiments provide strong evidence for a Structural source of the SGA. • Reading times (RT s) slowed where an ORC parse becomes unambiguous, no matter the case alignment. Disambiguation effect • RT s slowed again at the RC-final Verb. Integration effect (cf. Staub 2010, Levy & Keller 2013). 5

Roadmap • Theories of the SGA • Morphosyntactic properties of Georgian • Experimental design • Predictions • Results & Discussion 6

Roadmap • Theories of the SGA • Hierarchical structure: Subjects are universally more accessible • Morphological cues: N OM & A BS DPs are least surprising/informative 7

The Subject Gap Advantage Structural Hypothesis (Keenan & Comrie 1977) • Universally, subjects are higher ( → more accessible) than objects. SRC 😈 (3) the painter [ RC who [ TP __ inspired the writer ]] ORC 💪 (4) the painter [ RC whom the writer [ VP inspired __ ]] 8

The Subject Gap Advantage Structural Hypothesis (Keenan & Comrie 1977) • Prediction: SGA, no matter a language’s case alignment (Ch’ol & Q’anjob’al: Clemens et al. 2015, Avar: Polinsky et al. 2012) . SRC 😈 (5) the painter [ RC who ERG [ TP __ inspired the writer ]] ORC 💪 (6) the painter [ RC who ABS the writer [ VP inspired __ ]] 9

The Subject Gap Advantage Case Cue Hypothesis (Polinsky et al. 2012; cf. Hale 2006) • Information from a dependent case ( ACC in English) causes a processing cost. SRC 💂 (7) the painter [ RC who NOM __ inspired the writer ] ORC 🤕 (8) the painter [ RC whom ACC the writer inspired __ ] Must be transitive! 10

The Subject Gap Advantage Case Cue Hypothesis (Polinsky et al. 2012; cf. Hale 2006) • Prediction: E RG –A BS languages will have an Object Gap Advantage (OGA)! (Basque: Carreiras et al. 2010; Avar: Polinsky et al. 2012) . SRC 🤕 (9) the painter [ RC who ERG __ inspired the writer ] Must be transitive! ORC 💂 (10) the painter [ RC who ABS the writer inspired __ ] 11

Summary Theories of the SGA are difficult to disentangle — unless you’re in an ergative language. N OM –A CC E RG –A BS Structure SGA SGA Case Cue OGA 12

Roadmap • Theories of the SGA • Morphosyntactic properties of Georgian • Split ergativity 13

Georgian 101: Split ergativity Case on subjects & objects depends on the tense–aspect–mood / TAM (Aronson 1995) . (11) ekim- i ḳ ar- s gaa ġ ebs doctor- NOM door- DAT open. TR . FUT ‘the doctor will open the door’ Alignment TR SUBJ INTR SUBJ TR OBJ (12) ḳ ar- i gai ġ eba N OM –A CC FUT NOM DAT door- NOM open. INTR . FUT PAST ‘the door will open’ PERF 14

Georgian 101: Split ergativity Case on subjects & objects depends on the tense–aspect–mood / TAM (Aronson 1995) . (13) ekim- ma ḳ ar- i gaa ġ o doctor- ERG door- NOM open. TR . PAST ‘the doctor opened the door’ Alignment TR SUBJ INTR SUBJ TR OBJ (14) ḳ ar- i gai ġ o N OM –A CC FUT NOM DAT door- NOM open. INTR . PAST E RG –A BS PAST ERG NOM ‘the door opened’ PERF 15

Georgian 101: Split ergativity Case on subjects & objects depends on the tense–aspect–mood / TAM (Aronson 1995) . (15) ekim- s ḳ ar- i gau ġ ia doctor- DAT door- NOM open. TR . PERF ‘the doctor has opened the door’ Alignment TR SUBJ INTR SUBJ TR OBJ (16) ḳ ar- i ga ġ ebula N OM –A CC FUT NOM DAT door- NOM open. INTR . PERF E RG –A BS PAST ERG NOM ‘the door has opened’ D AT –A BS PERF DAT NOM 16

Recap RC processing has been investigated in only a few (Split-)Ergative languages. But these are just the place to test theories of the SGA. For Georgian… • Dependent case may be on Subj ( ERG ) or Obj ( DAT ). Case Hypothesis: SGA if FUT (N OM –A CC ); OGA if PAST or PERF (E RG –A BS or D AT –A BS ) 17

Recap RC processing has been investigated in only a few (Split-)Ergative languages. But these are just the place to test theories of the SGA. For Georgian… • Dependent case may be on Subj ( ERG ) or Obj ( DAT ). Structural Hypothesis: SGA in all TAM s 18

Roadmap • Theories of the SGA • Morphosyntactic properties of Georgian • Experimental design • Overview of Experiment 1 • Item Sets 19

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 20

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 21

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 22

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 23

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 24

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 25

Design Overview T ask: Self-paced reading A technique for measuring incremental processing. Design: 3 ( TAM /alignment) × 2 (gap site) { FUT , PAST , PERF } × {SRC, ORC} 36 item sets, 64 fillers (including 24 items of Experiment 2) Each sentence followed by a Y–N comprehension Q 26

Design Overview Participants: 57 native Georgian speakers 46 ♀ / 11 ♂ , average age 23, paid 40 lari All in Tbilisi  , recruited via Facebook 4 excluded from analysis for low comprehension scores Conducted online via Ibex (Drummond 2007) Georgian script, Georgian instructions 27

Item Sets Item set: { FUT , PAST , PERF } × { SRC , ORC } HdN wh P XP1 XP2 Adj CoArg V W2 W3 W4 W5 W6 W7 W8 … gogo, ⎡ romel-ic bnel ṭ q ̇ e- š i ma ġ al- ∅ bi č̣ -s naxavs ⎤ … (17) … girl. NOM ⎣ RC which- NOM dark woods-in tall- DAT boy- DAT see. FUT ⎦ ‘…the girl [ RC who __ will see the tall boy in the dark woods ]…’ … gogo, ⎡ romel-sac bnel ṭ q ̇ e- š i ma ġ al-i bi č̣ -i naxavs ⎤ … (18) … girl. NOM ⎣ RC which- DAT dark woods-in tall- NOM boy- NOM see. FUT ⎦ ‘…the girl [ RC who the tall boy will see __ in the dark woods ]…’ 28

Item Sets Item set: { FUT , PAST , PERF } × { SRC , ORC } HdN wh P XP1 XP2 Adj CoArg V W2 W3 W4 W5 W6 W7 W8 … gogo, ⎡ romel-ic bnel ṭ q ̇ e- š i ma ġ al- ∅ bi č̣ -s naxavs ⎤ … (17) … girl. NOM ⎣ RC which- NOM dark woods-in tall- DAT boy- DAT see. FUT ⎦ ‘…the girl [ RC who __ will see the tall boy in the dark woods ]…’ … gogo, ⎡ romel-sac bnel ṭ q ̇ e- š i ma ġ al-i bi č̣ -i naxavs ⎤ … (18) … girl. NOM ⎣ RC which- DAT dark woods-in tall- NOM boy- NOM see. FUT ⎦ ‘…the girl [ RC who the tall boy will see __ in the dark woods ]…’ 29