Prof. Cristiano Chesi | A competence-based (Top-Down) model for parsing (complex) non-local dependencies

Guest Lecture

  • Date: Jul 18, 2017
  • Time: 02:00 PM - 03:00 PM (Local Time Germany)
  • Speaker: Prof. Cristiano Chesi
  • University Institute for Advanced Studies (IUSS), Pavia, Italy
  • Location: MPI for Human Cognitive and Brain Sciences
  • Room: Wilhelm Wundt Room (A400)
Minimalism (Chomsky 1995-2008) introduced a simplification in the derivational machinery necessary for building phrase structures. The transparency of the assumed derivation with respect to on-line processing of phrase structure is still under discussion both from a psycholinguistic/computational (Phillips, 2013; Chesi, 2015) and a neurolinguistic (Nelson et al., 2017; Zaccarella et al., 2017) perspective.

Here I will focus on well know performance asymmetries elicited in (object) clefts processing: interpreting the head of an Object Cleft (e.g. “the banker”, (1.a)) as the correct argument of the lexical verb (“avoided”, (1.a)) is a critical step for comprehending this construction. This dependency has been deeply investigated both from the theoretical/competence perspective (Friedmann et al., 2009; Belletti & Rizzi, 2012 a.o.) and from the psycholinguistic/performance one (Gordon et al., 2004 a.o.), especially when an intervening “similar” DP (i.e. the subject of the cleft, e.g. “the lawyer” in (1.a) vs. “Dan” or “we”, (1.b)) is processed between the head and the lexical verb (Warren & Gibson, 2005).

(1) a. it was [the banker]i that [the lawyer] avoided _i at the party

b. it was [the banker/Pat/you]i that [the lawyer/Dan/we] avoided _i at the party

c. it was [the/you banker]i that [the/you lawyer] avoided _i at the party

The goal of this talk is twofold: on the one hand, I want to discuss an extra piece of evidence by presenting the results of some (on-line and off-line) studies on Italian, focusing on person feature mismatch (3rd (default) vs. 2nd person, (1.c)) when the “lexical restriction” is present; on the other, I want to discuss a feature-based “complexity metric” that predicts parsing asymmetries in more precise and graded way compared to alternative approaches, eventually attempting at reconciling a formal theory of phrase structure building (competence) and the psycholinguistic performance.

Here I adopt a left-right (top-down) derivational minimalist model (Chesi, 2015) where the non-local dependency (Internal Merge) is computed using a memory buffer. Coherently with featural relativized minimality (Friedmann et al., 2009), a Feature Retrieval Cost (FRC) function is presented, expressing the cost of retrieving specific items from memory: as in cue-based memory retrieval (Van Dyke & McElree, 2006) effects on the verb are predicted (this is not evident using a standard bottom-to-top minimalist derivation!). In case of mismatch (higher accuracy in comprehension questions), the processing cost is (correctly) predicted to be lower. On matching conditions, pro-pro 2nd person (vs. default 3rd person) matching condition pays the highest cost (two distinct “logophoric agents” should be instantiated in the left periphery, Sigurðsson, 2004), while art-pro (vs. pro-art) condition is easier to be interpreted (the salience of the 2nd person triggering verb agreement is rewarded) though it requires longer time to encode person feature (longer fixation times on the subject cleft segment).

Selected references

Belletti & Rizzi (2012) in Berwick & Piattelli Palmarini OUP. Chesi (2015) J. of Psych. Res. 44(1), 65-89. Friedmann et al (2009) Lingua. 119:67–88. Gibson (1998) Cognition 68, 1-76. Gordon et al (2004) J. of Mem. and Lang. 51:97-114. Nelson et al (2017). PNAS 201701590. Phillips (2013) Parser-grammar relations: We don’t understand everything twice. 294-315. Sigurðsson (2004) Italian J. of Ling. 16.1. Van Dyke & McElree (2006) J. of Mem. and Lang, 55(2), 157-166. Warren & Gibson (2005) Lang. and Cog. Proc. 20: 751-767. Zaccarella et al (2017) Cerebral Cortex. 27:411-421.

Poster
Go to Editor View