Discourse behavior of possessives reflects the importance of interpersonal relationships
Jesse Storbeck & Elsi KaiserUniversity of Southern California
Experiments in Linguistic Meaning, Sept. 17, 2020
Who we are
2
Jesse [email protected]
www-scf.usc.edu/~jstorbec/
Elsi [email protected]
elsikaiser.com/
Possessives in discourse● As people comprehend language, they keep track of relevant entities in a
mental model of the discourse
● Most of the work on how entities are represented on the discourse level has focused on simple nominal phrases (i.e. one NP → one discourse referent)
● What about complex NPs like possessive constructions?○ Sam’s car refers to two entities → two discourse referents
● How are possessives represented on the discourse level?○ Does the extremely broad range of semantic relations possessives can
express affect the way the two referents are represented in discourse?
3
Roadmap1. Background on discourse, possessives, and interpersonal relationships
● Animate possession → interpersonal relationship (e.g. her brother)
2. Three competing hypotheses about discourse representations of possessions
3. A sentence-continuation experiment to assess our hypotheses
● We tested how animacy and being possessed affect a referent’s likelihood of re-mention in subsequent discourse
4
Discourse-level representation and prominence● Entities in discourse vary in prominence (i.e. salience)
● Many factors can influence referent prominence, including:
○ Grammatical role■ Subjects usually more prominent than objects (e.g. Chafe, 1976;
Crawley et al., 1990)
○ Animacy■ Animates usually more prominent than inanimates ■ Cross-linguistic effects on word order, choice of nominal form, etc.
(e.g. Bock et al., 1992; Dahl & Fraurud, 1996)
5
Possessives in discourse● the car → one entity explicitly mentioned simple nominal
● Sam’s car → two entities explicitly mentioned possessivepossessor possession
1. How are the two entities represented in discourse?○ Simply two independent discourse referents?○ Linked referents?
2. If linked, does the link differ across various types of possession relations?○ e.g. Sam’s car [ownership], Sam’s arm [part-whole], Sam’s dad [kinship]○ Same morphosyntactic realization in English, different semantics
6
● Little work on the discourse-level behavior of possessives● Conflicting claims by accounts within Centering Theory (Grosz et al., 1995):
○ Possessors more prominent than possessions (Chae, 2003)■ Sam’s car → Sam > car
○ Possessors more prominent than inanimate possessions but not animate possessions (Di Eugenio, 1998)
■ Sam’s car → Sam > car■ Sam’s doctor → doctor > Sam
7
Prior work: conflicting claims about prominence of possessors vs. possessions
Prior work: the need for further research● Centering Theory addresses the relative prominence of possessors and
possessions (i.e. it ranks all referenced entities according to prominence)
○ Less clear how being possessed affects a referent’s discourse representation (i.e. what’s the effect of being possessed per se?)
● Possessive constructions are commonplace
○ Yet current discourse theories struggle with how to represent them○ Thus, the lack of research on their behavior in discourse represents a
significant theoretical gap
8
Interpersonal relationships● With animate possessions, possessives denote interpersonal relationships
(e.g. her father, his cousin, their boss)
● Interpersonal relationships are critical for human health and well-being (e.g. Cacioppo & Hawkley, 2009; Eisenberger & Cole, 2012)
○ Longevity (e.g. Holt-Lunstad et al., 2010)○ Disease resistance (e.g. Miller et al., 2009)○ Loneliness and cognitive decline (e.g. Tilvis et al., 2004)
● Interpersonal relationships are cognitively privileged○ What about in discourse?
9
Three competing hypotheses about the discourse prominence of possessed referents:
1. The animacy hypothesis
2. The possessive hypothesis
3. The interaction hypothesis
10
1. The animacy hypothesis● In general, animate referents are more prominent in discourse and memory
(e.g. Bock et al., 1992; Dahl, 2008; Gelin, 2019)
● Therefore, animate possessions are more prominent than inanimate ones
○ Sam’s car → Sam > car○ Sam’s doctor → doctor > Sam
● The animacy hypothesis predicts that this effect should be comparable in magnitude to animacy effects in simple nominals ([a] doctor > [a] car)
11
2. The possessive hypothesis● Possessive NPs are referentially more complex than simple NPs
○ They refer to two entities, instead of just one
● Representational complexity promotes retrieval from memory (e.g. Hofmeister, 2011; Karimi & Ferreira, 2016; Troyer et al., 2016)
● Therefore, possessions are more prominent than simpler nominals
○ [Sam’s] car > [a] car
■ Being possessed → increased complexity → increased prominence
12
3. The interaction hypothesis● If animacy hypothesis and possessive hypothesis hold →
animate possessions should be most prominent○ Maybe another contribution to animate possessions’ prominence?
■ Interpersonal relationships
● Interpersonal relationships support human health → cognitive privilege○ Adaptive memory theories: better memory for animates arose from
evolutionary importance of identifying threats, mates, and social groups (i.e. interpersonal relationships) (e.g. Nairne et al., 2013)
● Due to the physiological, cognitive, and adaptive importance of interpersonal relationships, animate possessions are exceptionally prominent in discourse
13
Research questions● How are possessives represented on the discourse level?
● Do different semantic possession relations affect representation in discourse?○ Sam’s car [ownership] vs. Sam’s doctor [interpersonal relationship]
● Are different types of possessed entities more or less likely to be mentioned subsequently in the discourse? With what linguistic form?
● Can we find support for our animacy hypothesis, possessive hypothesis, or interaction hypothesis?
14
Method: sentence continuation paradigm● Sentence continuation task measures the discourse prominence of competing
referents (e.g. Arnold, 2001; Stevenson et al., 1994)
● Participants (n=40 MTurkers) get a prompt sentence, produce an additional sentence which naturally continues the discourse
15
Analysis of sentence continuations● Dependent variable(s): which entities from the prompt do people mention in
their continuations? In what grammatical position? With what form?○ Higher likelihood of mention → more prominent○ Focus today on likelihood of mention (ask me about pronominalization data in Q&A 😃)
● Subject position is most prominent (e.g. Chafe, 1976; Crawley et al., 1990)○ Thus the most prominent referent from the prompt is likely to be realized
as the subject of the continuation (e.g. Stevenson et al., 1994)
● We understand mentions across the entire continuation as a correlated, perhaps more holistic measure
16
Experiment design: prompt structure
Melissa lupted a/her chauffeur/backpack.● 2×2 design manipulating properties of direct object:
○ Animacy (human role nouns vs. alienable inanimates)○ Possessed vs. indefinite
● Nonce verbs to minimize potential effects of verb semantics○ Implicit causality (e.g. Hartshorne & Snedeker, 2013)○ Nonce verbs can be held constant within items○ Post-hoc analysis: no consistent biases for nonce verbs
● Role nouns’ stereotypical gender contrasted with name (Misersky et al., 2013)○ This is only in order to mitigate potential pronoun ambiguity
17
Example targets (2 of 24)
18
Example responses
● Jennifer pranned her surgeon.○ She asked him how long her surgery would last.○ The surgeon did a good job.○ It wasn't very sanitary for an OR.
19
Coding responses● Jennifer pranned her surgeon.
○ She asked him how long her surgery would last.
1. What is mentioned in the subject position of the continuation?She → preceding subject
2. What is mentioned anywhere in the continuation?She, her → preceding subject, him → preceding object
● Collapsing across all forms of mention (ask me about pronominalization data in Q&A 😃)
● 97% agreement in double-coded 10% subset of the responses
20
Statistical analysis● Generalized linear mixed-effects (glmer) models in lme4 (Bates et al., 2015)
● Models fit to binomial outcome of mentioning a given entity from the prompt in a given position of the continuation → combinations yield four models○ Preceding subject as continuation subject? Yes → 1, No → 0○ Preceding object as continuation subject?○ Preceding subject mentioned anywhere?○ Preceding object as mentioned anywhere?
● Maximal convergent models○ All have random intercepts for items and participants○ Random slopes added when they did not result in non-convergence
21
Predictions● The animacy hypothesis, possessive hypothesis, and interaction hypothesis
make different predictions of the results
● Predictions broadly relevant for subject position and across the continuation
○ Subject position is a “winner take all” view of prominence
○ Mentions across the continuation will be correlated but more inclusive
● These predictions relate mainly to mentions of the preceding object, since that is where the experimental manipulation takes place
22
Predictions: the animacy hypothesis● The animacy hypothesis
○ Animate possessions are more prominent than inanimate ones
■ Mentions of animate objects > inanimate objects■ Animacy effect in possessives parallel to its effect in indefinites■ We expect a main effect of animacy in the models for mentions of
the preceding object
23
Predictions: the possessive hypothesis● The possessive hypothesis
○ Possessions are more prominent than simple NPs (here, indefinites)
■ Mentions of possessions > indefinites■ Possessive effect in animates parallel to its effect in inanimates■ We expect a main effect of being possessed in the models for
mentions of the preceding object
● Note that differences in givenness and specificity between possessions and indefinites (e.g. Barker, 2000; von Heusinger, 2011) could affect this prediction, but more on this in the results/discussion
24
Predictions: the interaction hypothesis● The interaction hypothesis
○ Animate possessions are exceptionally prominent in discourse
■ Superadditive effect of animacy and possession together
■ We expect an interaction of animacy and being possessed in the models for mentions of the preceding object
25
Results: two sets of figures● Jennifer pranned her surgeon.
○ She asked him how long her surgery would last.
1. What is mentioned in the subject position of the continuation?○ Subject position is mutually exclusive, so proportions within a
condition will sum to 1* (*except for a very small number of they responses)
2. What is mentioned anywhere in the continuation?○ Both entities can be mentioned across the continuation, so
proportions within a condition do not sum to 1
26
Results: subject position of continuations (Fig. 1)
“Melissa lupted a/her chauffeur/backpack.”
27
Results: subject position of continuations (Fig. 1)
“Melissa lupted a/her chauffeur/backpack.”
28
Results: subject position of continuations (Fig. 1)
“Melissa lupted a/her chauffeur/backpack.”
29
● Animacy effect: preceding subjects mentioned more often as continuation subject when preceding object is inanimate (p<0.01)
Results: subject position of continuations (Fig. 1)
“Melissa lupted a/her chauffeur/backpack.”
30
● Animacy effect: animate preceding objects more likely than inanimates to be mentioned in subject position of continuations (p<0.001)
Results: subject position of continuations (Fig. 1)
“Melissa lupted a/her chauffeur/backpack.”
31
● Interaction: animate possessed preceding objects are extra likely as continuation subjects (i.e. animacy:possession interaction, p=.05; simple effect, p=.03)
Results: all positions of continuations (Fig. 2)
“Melissa lupted a/her chauffeur/backpack.”
32
Results: all positions of continuations (Fig. 2)
“Melissa lupted a/her chauffeur/backpack.”
33
Results: all positions of continuations (Fig. 2)
“Melissa lupted a/her chauffeur/backpack.”
34
● No significant effects: preceding subjects are equally likely across conditions to be mentioned anywhere in the continuation
Results: all positions of continuations (Fig. 2)
“Melissa lupted a/her chauffeur/backpack.”
35
● Interaction: animate possessed preceding objects are extra likely to be mentioned anywhere in the continuation (i.e. animacy:possession interaction, p<.01)
Results: summary● Animacy effects in subject position for
preceding subjects and objects, but not across the entire continuation
● Interactions: animate possessions particularly likely to be mentioned○ Effects in both subject position and
across the entire continuation○ These effects not accounted for by
either animacy or possession alone○ Cognitive importance of
interpersonal relations influencing language?
What is mentioned in subject position of continuation?
What is mentioned anywhere in continuation?
Predictions revisited1. Animacy Hypothesis
○ Mentions of animate objects > inanimate objects○ Animacy effect in possessives parallel to its effect in indefinites
2. Possessive Hypothesis○ Mentions of possessed objects > indefinite objects○ Being possessed affects animates and inanimates similarly
3. Interaction Hypothesis○ Possessed animate objects mentioned most often○ Superadditive effect of animacy + possession
37
Predictions revisited
1. ✗ Animacy Hypothesis○ Mentions of animate objects > inanimate objects○ Animacy effect in possessives parallel to its effect in indefinites
2. ✗ Possessive Hypothesis○ Mentions of possessed objects > indefinite objects○ Being possessed affects animates and inanimates similarly
3. ✓ Interaction Hypothesis○ Possessed animate objects mentioned most often○ Superadditive effect of animacy + possession
38
Research questions reviewed● Are different types of possessed entities more or less likely to be mentioned
subsequently in the discourse?
✓ YES
● Do different possession relations affect representation in discourse?
✓ YES results compatible with such a theory
● Can we find support for our animacy hypothesis, possessive hypothesis, or interaction hypothesis?
39
Research questions reviewed● Are different types of possessed entities more or less likely to be mentioned
subsequently in the discourse?
✓ YES
● Do different possession relations affect representation in discourse?
✓ YES results compatible with such a theory
● Can we find support for our animacy hypothesis, possessive hypothesis, or interaction hypothesis?
40
Research questions reviewed● Are different types of possessed entities more or less likely to be mentioned
subsequently in the discourse?
✓ YES
● Do different possession relations affect representation in discourse?
✓ YES, results compatible with such a theory
● Can we find support for our animacy hypothesis, possessive hypothesis, or interaction hypothesis?
41
Research questions reviewed● Are different types of possessed entities more or less likely to be mentioned
subsequently in the discourse?
✓ YES
● Do different possession relations affect representation in discourse?
✓ YES, results compatible with such a theory
● Can we find support for our animacy hypothesis, possessive hypothesis, or
✓ interaction hypothesis?
42
Discussion● Possessed animates are especially prominent in discourse, as measured by
their likelihood of re-mention○ In general, animates are more prominent in discourse and memory, but
possessed animates seem to get a special boost in prominence■ Possessed animates explicitly denote interpersonal relationships■ May relate to non-linguistic theories on importance of interpersonal
relationships (social, physiological, and adaptive benefits)
● Compatible with Di Eugenio’s (1998) Centering Theory account○ Furthermore, we show that this boost is due to more than a simple
animacy effect or the additive effects of animacy and possession
43
Discussion● We observe expected animacy effects (e.g. competition for prominence
between animate referents) but what about an overall possessive effect?○ Possessions and indefinites may differ in givenness and specificity, which
affect prominence (e.g. Barker, 2000; Kaiser, 2011; von Heusinger, 2011)■ Potential to mask a main effect of possession■ Maybe a different result with definites?
● Crucially, givenness/specificity differences do not explain interactions○ Givenness and specificity should not vary with animacy○ Choosing a different class of nominals to compare with possessions
should modulate possessive effect, but interactions should remain
44
46
Appendix: pronominalization results● Mentions of the preceding subject were much more likely to use a pronoun
47
Appendix: pronominalization results● This holds across conditions, even in subject position with preceding animate
objects, where the preceding object was more likely as continuation subject○ Likelihood of mention ≠ likelihood of pronominalization (Kehler & Rohde, 2013)
48
Appendix: pronominalization results
Appendix: item details
49
Name Nonce Verb
Animate Object
Inanimate Object Name Nonce
VerbAnimate Object
Inanimate Object
Carol debbed butler violin Megan gerped chiropractor clarinet
Emily pletched stockbroker saxophone Melissa lupted chauffeur backpack
Helen swudged tattooist unicycle Rebecca gweeshed bodyguard umbrella
Jennifer pranned surgeon blanket Sarah chooped plumber scarf
Jessica rulked electrician chandelier Stephanie shupped mechanic flashlight
Kimberly chabbed chef fork Veronica brilted priest towel
Appendix: item details
50
Name Nonce Verb
Animate Object
Inanimate Object Name Nonce
Verb Animate Object Inanimate Object
Anthony jolfed florist toaster Jason joiped congresswoman wheelbarrow
Brad churbed receptionist screwdriver Joseph tammed babysitter microwave
Brian sibbed waitress basket Kevin blorned stewardess dictionary
Daniel zatted nurse jacket Mark meared housekeeper stereo
Fred risped nanny broom Robert dasped secretary television
James fammed hairdresser thermometer William dreezed psychic helmet
Appendix: glmer formulaesubject_model <- glmer(glmer_subj ~ animacy * det + (1 + animacy | subjectID) + (1 | itemID)
object_model <- glmer(glmer_obj ~ animacy * det + (1 + animacy + det | subjectID) + (1 | itemID)
subject_model_anywhere <- glmer(itemSubjMention ~ animacy * det + (1 + animacy | subjectID) + (1 | itemID)
object_model_anywhere <- glmer(itemObjMention ~ animacy * det + (1 + animacy | subjectID) + (1 | itemID)
51
ReferencesArnold, J. E. (2001). The effect of thematic roles on pronoun use and frequency of reference continuation. Discourse Processes, 31(2), 137-162.
doi:10.1207/S15326950DP3102_02
Arnold, J., & Griffin, Z. M. (2007). The effect of additional characters on choice of referring expression: Everyone counts. Journal of Memory and Language, 56(4), 521-536. doi:10.1016/j.jml.2006.09.007
Barker, C. (2000). Definite possessives and discourse novelty. Theoretical Linguistics, 26(3), 211. doi:10.1515/thli.2000.26.3.211
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1-48. doi:10.18637/jss.v067.i01
Bock, K., Loebell, H., & Morey, R. (1992). From conceptual roles to structural relations. Psychological Review, 99(1), 150-171. doi:10.1037/0033-295X.99.1.150
Cacioppo, J. T., & Hawkley, L. C. (2009). Perceived social isolation and cognition. Trends in cognitive sciences, 13(10), 447-454.
Chae, S. (2003). Possessives in naturally occurring discourse: A centering approach. University of Pennsylvania Working Papers in Linguistics, 9(1). Retrieved from https://repository.upenn.edu/pwpl/vol9/iss1/6
Chafe, W. L. (1976). Giveness, contrastiveness, definiteness, subjects, topics, and point of view. In C. N. Li (Ed.), Subject and topic. New York: Academic Press.
52
ReferencesDahl, Ö. (2008). Animacy and egophoricity: Grammar, ontology and phylogeny. Lingua, 118(2), 141-150. doi:10.1016/j.lingua.2007.02.008
Dahl, Ö., & Fraurud, K. (1996). Animacy in grammar and discourse. In T. Fretheim, & J. K. Gundel (Eds.), Reference and Referent Accessibility. Pragmatics & Beyond New Series, (pp. 47-64). Amsterdam: John Benjamins.
Di Eugenio, B. (1998). Centering in Italian. In M. Walker, A. Joshi & E. Prince (Eds.), Centering theory in discourse (pp. 115-137)
Eisenberger, N. I., & Cole, S. W. (2012). Social neuroscience and health: neurophysiological mechanisms linking social ties with physical health. Nature neuroscience, 15(5), 669-674.
Fukumura, K., & van Gompel, R. P. G. (2010). Choosing anaphoric expressions: Do people take into account likelihood of reference? Journal of Memory and Language, 62(1), 52-66. doi:10.1016/j.jml.2009.09.001
Gelin, M., Bonin, P., Méot, A., & Bugaiska, A. (2018). Do animacy effects persist in memory for context? Quarterly Journal of Experimental Psychology, 71(4), 965-974. doi:10.1080/17470218.2017.1307866
Grosz, B. J., Joshi, A. K., & Weinstein, S. (1995). Centering: A framework for modeling the local coherence of discourse. Computational Linguistics, 21(2), 203.
Hartshorne, J. K., & Snedeker, J. (2013). Verb argument structure predicts implicit causality: The advantages of finer-grained semantics. Language and Cognitive Processes, 28(10), 1474-1508.
53
ReferencesHofmeister, P. (2011). Representational complexity and memory retrieval in language comprehension. Language and Cognitive Processes, 26(3),
376-405. doi:10.1080/01690965.2010.492642
Holt-Lunstad, J., Smith, T. B., & Layton, J. B. (2010). Social relationships and mortality risk: a meta-analytic review. PLoS medicine, 7(7), e1000316.
Kaiser, E. (2011). Focusing on pronouns: Consequences of subjecthood, pronominalisation, and contrastive focus. Language and Cognitive Processes, 26(10), 1625-1666. doi:10.1080/01690965.2010.523082
Karimi, H., & Ferreira, F. (2016). Informativity renders a referent more accessible: Evidence from eyetracking. Psychonomic bulletin & review, 23(2), 507-525.
Kehler, A., & Rohde, H. (2013). A probabilistic reconciliation of coherence-driven and centering-driven theories of pronoun interpretation. Theoretical Linguistics, 39(1), 1-37. doi:10.1515/tl-2013-0001
Miller, G., Chen, E., & Cole, S. W. (2009). Health psychology: Developing biologically plausible models linking the social world and physical health. Annual review of psychology, 60, 501-524.
Misersky, J., Gygax, P. M., Canal, P., Gabriel, U., Garnham, A., Braun, F., ... Sczesny, S. (2014). Norms on the gender perception of role nouns in Czech, English, French, German, Italian, Norwegian, and Slovak. Behavior Research Methods, 46(3), 841-871. doi:10.3758/s13428-013-0409-z
Nairne, J. S., VanArsdall, J. E., Pandeirada, J. N., Cogdill, M., & LeBreton, J. M. (2013). Adaptive memory: The mnemonic value of animacy. Psychological Science, 24(10), 2099-2105.
54
ReferencesStevenson, R. J., Crawley, R. A., & Kleinman, D. (1994). Thematic roles, focus and the representation of events. Language and Cognitive Processes,
9(4), 519-548. doi:10.1080/01690969408402130
Tilvis, R. S., Kähönen-Väre, M. H., Jolkkonen, J., Valvanne, J., Pitkala, K. H., & Strandberg, T. E. (2004). Predictors of cognitive decline and mortality of aged people over a 10-year period. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 59(3), M268-M274.
Troyer, M., Hofmeister, P., & Kutas, M. (2016). Elaboration over a discourse facilitates retrieval in sentence processing. Frontiers in psychology, 7, 374.
von Heusinger, K. 2011. Specificity. In K. von Heusinger, C. Maienborn, & P. Portner (eds.), Semantics. An international handbook of natural language meaning, vol. 2, 1025–1058. Berlin: de Gruyter Mouton.
55
Top Related