Tes$ng Mixed Influences on Reference Resolu$on in Persian Dennis Ryan Storoshenko & Elias Abdollahnejad School of Languages, Linguis=cs, Literatures, and Cultures Comparison case: Korean caki Issue • Persian bi-morphemic anaphor xod-eš ‘self-3SG’ can be locally or long- distance bound • The 3SG pronoun un is equally acceptable and ambiguous in the same long-distance context as xod-eš (but is locally subject to Principle B): Sara i be David j goQ [ke man xod-eš i/j /un i/j =ro dust dar-am] S. to D. said that I self-3SG/3SG =OM like have-1SG ‘Sara said to David that I like self i/j /them i/j .’ • Unexpected in the context of Reinhart and Reuland’s (1993) typology: xod-eš has the form of a SELF anaphor, but the usage of a SE anaphor Eye Tracking Results • Subject Advantage (SA) = • SA from anaphor onset to 1800ms (30ms =me slices): [email protected] [email protected] • 31 adult na=ve speakers of Persian par=cipated in Calgary, Alberta • Par=cipants see sta=c images of two people on either side of an inanimate object in the middle of the screen • Two short sentences introduce the named characters and the se‘ng • Eye-movements are tracked (500 Hz) during the audio playback of ambiguous target sentences • Followed by a forced-choice reference resolu=on task Methodology References and Acknowledgements • Bates, Douglas, Mar=n Maechler, Ben Bolker and Steve Walker. 2015. Fi‘ng Linear Mixed- Effects Models Using lme4. Journal of Sta7s7cal So9ware, 67(1): 1-48. • Han, Chung-hye, Dennis Ryan Storoshenko, Beky Leung and Kyeong-min Kim. 2015. The =me course of long-distance anaphor processing in Korean. Korean Linguis7cs, 17(1): 1-32. • Kaiser, Elsi, Jeffrey T. Runner, Rachel S. Sussman and Michael K. Tanenhaus. 2009. Structural and seman=c constraints on the resolu=on of pronouns and reflexives. Cogni7on 112: 55-80. • Kuznetsova, Alexandra, Per B. Brockhoff and Rune H.B. Christensen. 2017. lmerTest Package: Tests in Linear Mixed Effects Models. Journal of Sta7s7cal So9ware, 82(13): 1-26. • Reinhart, Tanya and Eric Reuland. 1993. Reflexivity. Linguis7c Inquiry 24(4): 657-720. This research is funded by a Social Sciences and Humani$es Research Council of Canada (SSHRC) Insight Development Grant 430-2016-00128 to Storoshenko. • Reference resolu=on for forms with many poten=al antecedents within a sentence can be governed by an interac=on of mul=ple influences (Kaiser et al, 2009) • Korean caki is described as subject oriented (a typical SE anaphor trait) • Han et al. (2015) tests whether this is moderated by a matrix predicate in long distance cases: Jinswu i -ka Yenghuy j -hanthey nongwutay yep-eyse [caki i/j -ka J. -NOM Y. -to hoop beside-at self-NOM syus-ul te manhi sengkong-siki-lke-lako] malha-yess-ta shoot-ACC more much success-CAUS-FUT-COMP say-PST-DECL • Anaphor type (caki vs. null) tested against three predicates: malha ‘say’, tut ‘hear’, myenglyengha ‘order’ • In offline forced choice tes=ng, the anaphor caki is significantly more subject-oriented than the null pronoun, and the say trials were shown to be significantly more subject oriented, regardless of the anaphor • In eye-tracking data, caki shows an overall subject bias compared to the null pronoun, not emerging un=l over one second aQer the anaphor (a significant interac=on between anaphor and =me emerges in the 1200-1800ms aQer ukerance =mespan) • So subject orienta=on is confirmed for caki, before encountering the verb, but moderated by downstream sentence content • Persian has two addi=onal key differences: 1. S – IO – V– Clause or S – IO – Clause – V (non-canonical) 2. No predicate-neutral preposi=on Anaphor Predicate Reflexive (xod-eš) Pronoun (un) Say Condi=on 1 Condi=on 3 Hear Condi=on 2 Condi=on 4 Discussion • Xod-eš clearly shows a greater degree of subject orienta=on than un, in line with predic=ons for long-distance SE anaphors • During processing, the preposi=ons appear to be signalling a contrast between predicates, but no main effect, only interac=ons • There is no transla=on of this online predicate effect into the offline forced choice task (opposite of Korean) • Persian speakers may be less disposed to adjust final reference resolu=on based on downstream informa=on (normally the verb would precede the embedded clause) Sara i be/az David j , kenār-e māshin,[inke man mi-xā-m barāye xod-eš i/j /un i/j S. to/from D. beside-EZ car that I DUR-want-1SG for self-3SG/3SG ye mekanik-e māher be-ferest-am] =ro goQ/šenid. one mechanic-EZ skilled SUB-send-1SG =OM said/heard ‘Sara i said to/heard from David j beside the car that I want to send a skilled mechanic for self i/j /them i/j .’ ?? • Data analyzed in R with lme4 (Bates et al. 2015) and lmerTest (Kuznetsova et al. 2017) • AQer comparisons, the best model does not include a predicate effect reference~Anaphor+(1|Part)+(1|Item) • Overall significant preference for subjects with xod-eš over un, regardless of predicate (p<0.001) Overarching Ques$on Does Persian xod-eš show other proper=es of known SE anaphors? 0 0.05 0.1 0.15 0.2 0.25 0 300 600 900 1200 1500 1800 Pronoun Say Hear 0 0.05 0.1 0.15 0.2 0.25 0 300 600 900 1200 1500 1800 Reflexive Say Hear Refined Research Ques$ons for Persian: 1. Does xod-eš show subject orienta=on during sentence processing? 2. Does the choice of predicate in non-canonical embedded clause order have the same impact as in (canonical) Korean word order? Forced Choice Results Subject Fixa=ons Subj+Obj Fixa=ons Object Fixa=ons Subj+Obj Fixa=ons - • Sta=s=cal analysis uses six 300ms =me slices • AQer model comparisons, a full interac=on of anaphor, predicate, and =me slice is best, including a random effect for items: SA ~ Predicate * Anaphor * TimeSlice + (1 | Item) • Significant effect of Anaphor (p=0.05) and early TimeSlices (p<0.001) reflects an unexpected difference at the anaphor onset • Interac=ons between Anaphor and TimeSlice (p<0.001) reflect subject orienta=on for xod-eš, most pronounced at 600-1200ms • Interac=ons with Predicate and TimeSlice, and three-way interac=ons reflect the opposing trends by anaphor 85.88 78.62 65.72 64.11 14.11 21.37 34.27 35.88 say hear say hear REFLEXIVE PRONOUN % of subject and object choices subject object