West Coast Conference on Formal Linguistics (WCCFL) (38th : 2020)

How to value gender : lexicon, agree and feature transmission under ellipsis Kučerová, Ivona; Chapman, Cassandra; Moulton, Keir 2020

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
73804-Kucerova_I_et_al_Gender_WCCFL38_2020.pdf [ 788.66kB ]
Metadata
JSON: 73804-1.0389860.json
JSON-LD: 73804-1.0389860-ld.json
RDF/XML (Pretty): 73804-1.0389860-rdf.xml
RDF/JSON: 73804-1.0389860-rdf.json
Turtle: 73804-1.0389860-turtle.txt
N-Triples: 73804-1.0389860-rdf-ntriples.txt
Original Record: 73804-1.0389860-source.json
Full Text
73804-1.0389860-fulltext.txt
Citation
73804-1.0389860.ris

Full Text

How to value gender: lexicon, agree and feature transmission under ellipsisIvona Kučerová (McMaster)kucerov@mcmaster.caCassandra Chapman (U of Toronto)cassandra.chapman@utoronto.caKeir Moulton (U of Toronto)keir.moulton@utoronto.ca 1gender in the grammar• useful test-case for understanding the interactions of lexical semantics, narrow syntax and its interfaces 2this talk• an experimental study testing:  • whether the gender interpretation associated with nominal roots (~bio-social gender*) maps onto gender as a syntactic feature, and its morphological realization 3*Ackerman 2019the studies• gender under NP ellipsis (lexical semantics, agree, morphology) • follow up: gender in NP arguments (lexical semantics versus syntactic agree only) 4Gender under NP ellipsis 5gender-marking pairs of nouns in English*• symmetric nouns (prince ~ princess) • asymmetric nouns (actor ~ actress) 6*Percus 2001, Bobaljik and Zocca 2011, among otherssymmetric nouns(1) Harry is a prince/ *princess (2) Meghan is a *prince/ princess • => both lexically specified for conceptual gender • note: sociological construals of gender identities are rather complex; here, we assume that the speaker knows that Harry self-identifies as a male and Meghan as a female 7further evidence: NP ellipsis*(1) *Harry is a prince, and Meghan is too. (2) *Meghan is a princess, and Harry is too. 8*Bobaljik and Zocca 2011asymmetric nouns(1) Brad is an actor/ *actress (2) Angelina is an actor/ actress • => masculine appears unmarked 9further evidence: NP ellipsis*(1)  Brad is an actor, and Angelina is too. (2) *Angelina is an actress, and Brad is too. 10*Bobaljik and Zocca 2011Bobaljik & Zocca (2011)• masculine form: • unmarked in asymmetric nouns (actor) • marked in symmetric nouns (prince) • feminine form always marked (actress, princess) 11Sprouse, Messick & Bobaljik (2017)• only a partial confirmation of Bobaljik & Zocca’s proposal • feminine gender in actress much easier to strip under ellipsis than feminine gender in princess• unexpected if feminine gender is marked 12alternative hypothesis• structurally two different -ess • a grammatical shift  • from -ess denoting a female* => lexical semantics • to -ess being a morphological realization of n with a valued gender feature as a narrow syntax feature, i.e., emptied of its lexical meaning 13*akin to >zh< in Pesetsky (2013)motivation: agree in nominal predicates• copular clauses exhibit an effect of agree between the subject and a nominal predicate*  14*Béjar and Kahnemuyipour 2017number agreement(1)  Aya is the tallest student in the class (2)  Aya and Janine are the tallest students/*student in their respective classes. 15agree hypothesis 16• NP-NP copular clauses can exhibit agreement in gender as well • => emergence of grammatical gender agreement in a class of nominals that have undergone sociologically driven re-grammaticalizationa grammatical shift• lexicon representation undergoing a change because of changes in social attitudes • reanalysis of -ess as a morphological reflex of agree 17stage 1: actress (lexical) 18JessicaFisan nPn√actor+√essJessicaFisan nPn√actorJessicaFisan nPnuGender⇒ F-ess√actorPPPtoDPDaNPAPnewNPNwifePPPofDPDPtheyD’D’sNPAPcommonNPfriend1lexical semantics: femaleno syntactic agreestage 2: actor+ess (agree) 19emergence of unvalued gender on nJessicaFisan nPnuGender√actorDPD nPn√actorDPDuGender⇒MnPn√actorJessicaFisan nPn√actor+√essJessicaFisan nPn√actorJessicaFisan nPnuGender⇒ F-ess√actor1stage 2: actor+ess (agree) 20JessicaFisan nPn√actor+√essJessicaFisan nPn√actorJessicaFisan nPnuGender⇒ F-ess√actorPPPtoDPDaNPAPnewNPNwifePPPofDPDPtheyD’D’sNPAPcommonNPfriend1-ess only a morphological reflex of agreeprediction• 2 distinct gender representations of feminine NP predicates: • lexically specified • specified syntactically by agree • => should be testable under ellipsis  21design considerations• lexical semantics versus syntactic agree • adding gender on bound pronouns as a control • => gender on bound pronouns can be easily stripped under ellipsis 22what needs to be tested• symmetric nouns, with lexically specified gender,  • asymmetric nouns, with gender on n valued by agree in syntax, and • pronouns bound by (a)symmetric nouns, with gender being solely a morphological reflex (feature transmission) (sloppy identity) 23Experiment 1 24the study• an online 7-point acceptability judgement study, comparing the acceptability of gender mismatch under NP ellipsis and pronoun ellipsis • with Gender (whether the Male or Female noun appeared first) and Ellipsis type (NP or Pronoun) as within subject variables, and • Noun type (Asymmetric or Symmetric) as between item variables 25task• participants were asked to rate a target sentence with respect to a preceding context, which in the case of the Pronoun conditions ensured sloppy identity readings 26 27pronoun ellipsis examplePronoun ellipsis (Asymmetric/Symmetric)• Male: (1) The actor practiced his lines and the actress did too. (2) The prince practiced his lines and the princess did too. • Female: (3) The actress practiced her lines and the actor did too. (4) The princess practiced her lines and the prince did too. 28 29NP ellipsis exampleNP ellipsis (Asymmetric/Symmetric)• Male: (1) John is an actor and Mary is too. (2) John is a prince and Mary is too. • Female: (3) Mary is an actress and John is too. (4) Mary is a princess and John is too. 30predictions• ellipsis licensing requires syntax & semantics parallelism • => we expect an effect of gender mismatches in the semantic (lexical semantics) and syntactic representation (agree) 31predictions cont’d• ellipsis licensing does not require parallelism at the level of morphological realization • => we do not expect an effect of gender mismatch in cases where morphological realizations are not identical 32Prediction I• lexical semantics cannot be ignored: • => gender of symmetric nouns (prince~princess) cannot be stripped under ellipsis because the gender information is part of lexical semantics • => any gender mismatch under ellipsis of symmetric nouns is expected to be rated as ungrammatical 33Prediction II• we expect masculine asymmetric nouns to be acceptable under NP ellipsis because they are fully acceptable in their unelided form 1. Angelina is an actor. 34Prediction II cont’d• if -ess in feminine asymmetric nouns (actress) is established by Agree, the agree chain & the corresponding n will violate the syntactic part of the parallelism requirement • but since agree in narrow syntax does not have semantic consequences, the semantic part of the parallelism should be satisfied• => we expect that stripping gender (feminine) in asymmetric nouns under NP ellipsis should be significantly better (higher ratings) than stripping gender in symmetric nouns 35Prediction III• nevertheless, stripping feminine agree-based gender shouldn’t be as good as stripping gender on bound pronouns  • arises solely by morphological feature transmission,* thus it does not matter whether the antecedent is a symmetric or an asymmetric noun • parallelism will always be satisfied (no syntactic  agree, or lexical semantics at play) • => we expect the Pronoun ellipsis condition to be  always acceptable (high ratings), irrespective of the antecedent (asymmetric or symmetric) 36*no agree; Kratzer 2009, Heim 2008results• in our z-scored rating data, we observed a significant three-way interaction (p = 0.01) between Gender, Ellipsis type and Noun type 37results 38NP PronounFemale Male Female Male−1.0−0.50.00.5GenderRating (z−score)SymmetryAsymmetricSymmetricn = 64*plot shows predicted 
values from a linear mixed 
effects regression modelno significant differences 
in pronoun ellipsis conditions —> all rated high*asymmetric > symmetric 
(p < 0.01)* male > female 
(p < 0.01)prediction I: lexical semantics ✓prediction II: syntactic parallelism ✓prediction III: 
feature transmission ✓interim summary• -ess in actress is no longer lexically specified as female • instead, -ess is a morphological realization of n that syntactically agrees in gender with the subject 39−1.0−0.50.00.5Female MaleGenderRating (z−score)SymmetryAsymmetricSymmetricresultsquestion• if -ess can be a morphological reflex of agree, what is the gender status of the asymmetric nominal without -ess, i.e., actor? 40Hypothesis I• nouns like ‘actor’ are truly underspecified for gender (stage 3 of the grammatical re-analysis) • Prediction: if such a noun appears in an argument position, for example, as an agentive subject, it should be compatible with female antecedents 41stage 3: actor 42no unvalued gender on n;no lexical genderDPDthenPn√actorentered. . .TPDPDTheNPclaim CPCthatTPDPHomerT’Tdidn’tVPVthinkPPabout DPitT’TpastVPVoccurredPPPtoDPanyoneJessicaFisan nPnuGender√actor1Hypothesis II• arguments and NP predicates are different • for example, because of an implicated presupposition of lexical -ess (e.g., Heim 2008, Sudo 2011; cf. Sauerland 2003) that applies only to referential uses of NP • => if actor and actress in an argument position are in a direct competition, we expect actor to be interpreted as a male 43Hypothesis II cont’d• if argument and predicate NPs are different, then under Hypothesis II, actor should be interpreted as male in argument position 44previous literature?• Merchant 2014 argues that at least in Greek there is a difference between NP ellipsis and argument position of asymmetric nouns (~Hypothesis II) • however, Spathas & Sudo 2019 provide empirical evidence that questions the validity of the data reported in Merchant; instead, their study (judgment elicitation from a small number of speakers) supports Hypothesis I 45Experiment 2 46the study• an online 7-point acceptability judgement task, comparing the acceptability of gender mismatch with NP subjects and NP predicates • with Gender (whether the Male or Female noun appeared first) and Syntactic Position (Subject or Predicate) as within subject variables, and • Noun type (Asymmetric or Symmetric) as between item variables 47task• participants were asked to rate a target sentence in response to a context  • context contained an antecedent clearly gendered as female• contrasting NP subjects (arguments) and NP predicates in the target sentences 48 49example: NP subjectNP subjects (Asymmetric/Symmetric)• Asymmetric: (1) … a womani…. The actori has a wonderful stage presence. (2) … a womani…. The actressi has a wonderful stage presence. • Symmetric: (3) … a womani…. The princei is very regal. (4) … a womani…. The princessi is very regal. 50 51example: NP predicateNP predicates (Asymmetric/Symmetric)• Asymmetric: (1) … a womani…. Shei is a wonderful actor. (2) … a womani…. Shei is a wonderful actress. • Symmetric: (3) … a womani…. Shei is a very regal prince. (4) … a womani…. Shei is a very regal princess. 52Predictions: asymmetric subjects 53• Context: John is attending the opening performance of a play. He is watching a woman in a gold dress playing Cleopatra.  (1) The actori has a wonderful stage presence. (2) The actressi has a wonderful stage presence. • Hypothesis I (genderless actor): no rating differences between (1) and (2)  • Hypothesis II (arguments are different): actress > actor Predictions: symmetric subjects 54• Context: Harry is attending a ball at the palace. He notices a woman enter with a beautiful tiara. He says: (1) The princei is very regal. (2) The princessi is very regal. • asymmetric nouns are lexically specified for gender  • prince should be rated poorlyPredictions: asymmetric predicates 55• Context: John is attending the opening performance of a play. He is watching a woman in a gold dress playing Cleopatra.  (1) Shei is a wonderful actor. (2) Shei is a wonderful actress. • Hypothesis I (genderless actor): no rating differences between (1) and (2)  • Hypothesis II (arguments are different): no rating differences between (1) and (2) Predictions: symmetric predicates 56• Context: Harry is attending a ball at the palace. He notices a woman enter with a beautiful tiara. He says: (1) Shei is very regal prince. (2) Shei is very regal princess. • asymmetric nouns are lexically specified for gender  • prince should be rated poorlyresults• in our z-scored rating data, we did not observe a significant three-way interaction between Gender, Position, and Symmetry  57results 58Predicate SubjectFemale Male Female Male−1.0−0.50.00.5GenderRating (z−score)SymmetryAsymmetricSymmetricn = 63*plot shows predicted 
values from a linear mixed 
effects regression modelThree-way interaction 
not significant (p > 0.05)*female > male in symmetric nouns 
(p < 0.01)asymmetric nouns: 
Gender*Position 2-way 
interaction (p < 0.05)actor in subject position is rated lower than actress in subject position 

BUT actress in subject position is also better than actress in predicate position hypothesis II: arguments are differenthypothesis I:genderless actor✓ 
(at least partially)interim summary• actor-type (male asymmetric) nouns on the way to being truly underspecified for gender even in argument positions • but not quite there yet 59gender typology• lexically specified gender (lexical semantics) • gender as morphological reflex of agree in syntax (syntax & morphology) • gender as a morphological feature transmission (only morphology) • => reveals itself under NP ellipsis 60consequences for gender under ellipsis• gender on bound pronouns ignored • gender as a morphological reflex of syntactic agree can be only partially ignored (contra Merchant 2014) • lexically specified gender cannot be ignored • => ellipsis sensitive to both syntax & semantics but not morphology 61thank you to• SSHRC Insight Grant #435-2016-1034 (Grammatical vs semantic features: the semantics-morphology mapping, and its consequences for syntax; PI: Kučerová)  • SSHRC Insight Grant #435-2018-1012 (Processing Pronouns and Quantifiers; PI: Moulton) 62

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.73804.1-0389860/manifest

Comment

Related Items