Subject/object asymmetry in questions with quantifiers: Syntax or discourse? Asya Achimova 1 , eprez 2 , Julien Musolino 2 Viviane D´ 1 University of Leipzig 2 Rutgers University October 16, 2015
Outline Introduction Phenomenon Motivation Theoretical background Experimental data Discussion
Structural ambiguity in questions with quantifiers Subject-quantifier question (1) Which assignment did every student complete? a. The semantics assigment. Single answer b. Jane completed the semantics assignment, Tom completed the syntax assignment, and Mary completed the phonology assigment. Pair-list answer c. Their hardest assignment. Functional answer
Structural ambiguity in questions with quantifiers Object-quantifier question (2) Which student completed every assignment? a. Mary. Single answer b. * Jane completed the semantics assignment, Tom completed the syntax assignment, and Mary completed the phonology assigment. Pair-list answer
Structural ambiguity in questions with quantifiers Object-quantifier question (3) Which student completed each assignment? a. Mary. Single answer b. Jane completed the semantics assignment, Tom completed the syntax assignment, and Mary completed the phonology assignment. Pair-list answer
Problem ◮ Williams (1988) observed that each does not exhibit the asymmetry in questions ◮ Beghelli (1997), Szabolcsi (1997) proposed a distributivity-based account for the lack of asymmetry ◮ I will show today that pair-list answers (PLA) are more readily available for subject-quantifier questions than for object-quantifier questions even when the quantifier is each ◮ That is, we observe the subject/object asymmetry with both quantifier types ◮ Our results suggest that the asymmetry is a consequence of discursive rather than purely structural constraints
Theoretical background May (1985, 1988): Subject-quantifier questions CP DP IP D N DP IP Which assignment 1 D N I VP every t 2 student 2 did vP V DP complete t 1
Theoretical background May (1985, 1988): Object-quantifier questions CP DP IP D N DP IP Which student 1 D N I VP every t 1 assignment 2 did vP V DP complete t 2
Chierchia, 1993 ◮ Chierchia (1993) argues that wh -phrase leaves a functional trace indexed to the wh and its referential argument ◮ The argument trace acts as a pronominal element ◮ Quantifiers can only bind pronouns they c-command. Crossing over a pronoun is subject to WCO ◮ An object quantifier cannot bind a subject wh -trace, and no PLA is possible ◮ These examples are parallel to classic WCO ◮ *His mother 1 loves every boy 1
Ag¨ uero-Bautista, 2001 ◮ Ag¨ uero-Bautista (2001) makes use of the notion of reconstruction ◮ A PLA is available when the wh -phrase can reconstruct below the quantifier ◮ Presuppositional wh -phrases, such as which -N cannot reconstruct into a θ -position ◮ No PLA is possible for which -questions with object- every
Universal quantifiers are not all the same ◮ Williams (1988) observed that each escapes the restrictions on PLA availability ◮ PLAs are possible both for subject-quantifier and object-quantifier questions (4) Which student completed e ach assignment? a. Jane completed the semantics assignment, Tom completed the syntax assignment, and Mary completed the phonology assignment.
Distributivity and scope ◮ Beghelli (1997) proposed that strongly-distributive quantifiers, such as each , can target a position higher than IP - Distributive phrase ◮ Therefore, each can take scope over the subject- wh -phrase ◮ We expect no asymmetry for each
Beghelli, 1997 Object-quantifier question: strongly distributive quantifiers CP DP DistP D N DP ShareP Which student 1 D N AgrOP t 1 each assignment 2 AgrO vP V DP complete t 2
Beghelli, 1997 Object-quantifier question: pseudo-distributive quantifiers CP DP DistP D N DP ShareP t 1 Which student 1 AgrOP AgrO vP every assignment 2 V DP complete t 2
Motivation ◮ Are PLAs equally available for subject- and object-quantifier questions? ◮ Can we confirm the asymmetry for every and the lack of asymmetry for each ? ◮ In certain cases, judgments on the availability of PLAs appea to be contradictory
Syntax-semantics interface ◮ Can semantic properties of the quantifiers override the syntactic restrictions? ◮ Is the subject-object asymmetry a narrow phenomenon only applying to every ?
Experimental design ◮ Acceptability judgment task with na¨ ıve participants ◮ Judge the acceptability of a PLA on a 7-point scale ◮ Is that a possible answer? ◮ Definitely no (1) to Definitely yes (7)
Sample question Which toy did every child pick? John picked the car, Jane picked the truck, and Helen picked the toy tiger. Is that a possible answer? (Definitely no) 1 2 3 4 5 6 7 (Definitely yes)
Structure of the experiment 2x2x2 ◮ Quantifier position ◮ Subject quantifier ◮ Object quantifier ◮ Quantifier type ◮ Every ◮ Each ◮ Answer type ◮ Single answer ◮ Pair-list answer
Structure of the experiment Experimental questions ◮ 3 Practice questions ◮ 32 Critical questions ◮ 60 Control and filler questions
Structure of the experiment Participants 29 native speakers of English (undergraduate students)
Predictions ◮ We expect to see the asymmetry for every : PLAs should be less available for object-quantifier questions ◮ If each is not subject to stuctural constraints due to its semantics, PLAs should be equally available for both subject- and object-quantifier questions
Results ◮ PLAs for questions with each are more available ( β = − 1 . 385, SD = 0.334, p < 0.01) ◮ We observe a subject-object asymmetry for every β =1.573, SD = 0.455, p < 0.01) ◮ We also observe an asymmetry for each ( β =2.014, SD = 0.319, p < 0.01)
Results
Bayesian t-test ◮ The difference is scores for PLAs between subject- and object-quantifier questions is the same for every and each ◮ The t-test yields a Bayes Factor of 5, which corresponds to substantial evidence for H 0 on Jeffreys (1961) scale
Are PLAs with object- every ever possible?
Variability in the data: questions with every
Discussion Nature of the asymmetry: every ◮ Variability in the data is hard to explain within a structural account of the asymmetry ◮ We need a flexible constraint on the availability of PLAs
Discussion Nature of the asymmetry: each ◮ For the quantifier each , both subject- and object-quantifier PLAs are lequally licensed ◮ Yet, even with each we observe the asymmetry: a PLAs is more available for questions with subject quantifier ◮ We cannot use the grammatical/ungrammatical distinction to account for this difference ◮ It must come from another source ◮ Information structure is a possible candidate
Proposal ◮ Following Krifka (2001), Endriss (2009), and Eilam (2011) we suggest that the ability to give rise to PLAs depends on the ability of the quantifier phrase to be viewed as a topic ◮ We define topichood as aboutness, after Reinhart (1981) and Endriss (2009) ◮ For questions, Jaeger (2003) defines the topic as “what the question primarily requests information about” (2003:187)
Topichood and quantifier scope ◮ Subject-quantifiers more easily take wide scope since subjects are often topics (Li & Thompson, 1976) ◮ The ability of an object-quantifier to be construed as a topic depends on the context ◮ each -phrases are D-linked, therefore there is a set introduced in discourse already. Therefore, it is easier to construct a discourse where they act as topics
Implications ◮ The information structure account predicts an asymmetry for both universal quantifiers considered ◮ It also predicts that each can take wide scope more easily overall ◮ It allows to explain cases where PLAs are available for questions with object every
Recommend
More recommend
