Elsevier

Cognition

Volume 101, Issue 1, August 2006, Pages 1-42
Cognition

Thematic role properties of subjects and objects

https://doi.org/10.1016/j.cognition.2005.08.002Get rights and content

Abstract

This paper tests two claims about the thematic roles Agent and Patient: first, that they can be decomposed into more primitive features, as laid out in Dowty's (1991) Proto-Roles Hypothesis; and second, that these properties can be inferred directly from the grammatical roles subject and object. In Experiments 1 and 2, participants rated the Proto-Roles properties of the subject and object of transitive sentences with real verbs. Subjects were rated as more Agent-like than objects, while objects were rated as more Patient-like than subjects. Experiment 3 used transitive sentences with nonsense content words (e.g. “The rom mecked the zarg”). Even in the absence of a known verb, the results were the same. Experiment 4 examined the corollary prediction that subjects and objects should not differ on grammatically irrelevant properties (e.g. “being liquid”); this prediction was confirmed. In Experiment 5, verbs that do not permit objects (e.g. “fall,” “erupt”) were placed in the transitive frame. Results were the same as in Experiments 1 through 3. Moreover, the semantics of the verb were altered by the frame, indicating that participants tried to fuse the semantics of the verb with those of the frame. Possible sources of these inferences are considered.

Section snippets

The thematic role properties of subjects and objects

A critical function of human language is to specify who did what to whom. Sentences serve this function by acting as miniature plays, with actors (typically labeled by noun phrases) playing roles in the event labeled by the verb. In the parlance of linguistic theory, these actors are arguments; the role that an argument plays in an event is its thematic role. Because of their central role in the semantics of sentences, thematic roles have attracted considerable attention from both linguists (

Experiment 1: The proto-roles properties of subject and object in the presence of real transitive verbs

To date, no attempt has been made to assess the psychological validity of Dowty's Proto-Roles hypothesis—that is, to determine whether ordinary speakers will be more likely to attribute Agent properties to subjects than to objects, and more likely to attribute Patient properties to objects than to subjects. The aim of Experiment 1 is to do just that.

Experiment 2: The proto-roles properties of subject and object in the presence of real transitive verbs of different semantic types

Because it used a set of randomly selected verbs, Experiment 1 represented a strong test of Dowty's Proto-Roles hypothesis, which was meant to explain argument selection for transitive verbs in general. Despite its advantages, this method has some obvious drawbacks. Because there were few or no verbs involving causation, there was little chance that the property CAUSED DO would be rated more highly for the subject than for the object. Likewise for the property CHANGED. To verify that each

Experiment 3: The proto-roles properties of subject and object in the absence of a known verb

When the verb in a sentence is real, the subject is judged as more Agent-like than the object, and the object is judged as more Patient-like than the subject. Do these patterns hold in the absence of a real verb? Prior work on the semantics of syntactic structures (reviewed in Kako & Wagner, 2001) suggests that the answer should be yes. Experiment 3 tests this prediction by asking participants to judge the Proto-Roles properties of the subject and object of a transitive sentence with a nonsense

Experiment 4: Comparing subjects and objects on grammatically irrelevant properties

The first three experiments suggest that subjects and objects differ in their Proto-Role properties. As noted in the Introduction, Gleitman et al. (1996) have shown that subjects and objects also differ in properties related to Figure and Ground. Their findings raise an important question: Will subjects and objects differ on any set of properties about which speakers are asked? Perhaps participants in these experiments are merely trying to be cooperative, or perhaps (relatedly) they are driven

Experiment 5: Where do thematic role inferences come from?

Experiments 1 through 4 demonstrate that subjects are seen as more Agent-like than objects, and objects are seen as more Patient-like than subjects, in the presence of both known and invented verbs. This confirms the prediction that thematic role-related inferences can be drawn from syntactic structure—in this case, grammatical roles—alone, even in the absence of a verb to supply such information. When the verb is real, it is clear where these inferences come from. But where do they come from

General discussion

The aim of this paper was to test two hypotheses: (1) that the thematic role properties proposed by Dowty (1991) in his Proto-Roles hypothesis have psychological validity, and (2) that speakers can make inferences about these properties from grammatical roles alone, even in the absence of a real verb that would otherwise assign those properties to the noun phrases in the sentence. All five experiments asked participants to judge the likelihood that nonsense subjects and objects possessed the

Conclusions

The research presented in this paper verifies the psychological validity of Dowty's (1991) Proto-Roles Hypothesis, and further confirms the claim that subjects tend to be characterized by Proto-Agent properties, while objects tend to be characterized by Proto-Patient properties. The latter finding supports the claim that thematic role information can be inferred from grammatical roles alone, likely due to differences in prominence between the two arguments of the verb, as suggested by Fisher's

Acknowledgements

I thank Laura Wagner and three anonymous reviewers for their helpful comments on an earlier draft of this paper. I also thank David January and Arpiar Saunders for their assistance in data collection.

References (83)

  • J. Kegl

    Levels of representation and units of access relevant to agrammatism

    Brain and Language

    (1995)
  • K. McRae et al.

    Modeling the influence of thematic fit (and other constraints) in on-line sentence comprehension

    Journal of Memory and Language

    (1998)
  • S. Pinker

    How could a child use syntax to learn verb semantics?

    Lingua

    (1994)
  • E. Rosch et al.

    Family resemblances: Studies in the internal structure of categories

    Cognitive Psychology

    (1975)
  • C.K. Thompson

    Unaccusative verb production in agrammatic aphasia: The argument structure complexity hypothesis

    Journal of Neurolinguistics

    (2003)
  • J.C. Trueswell et al.

    Semantic influences on parsing: Use of thematic role information in syntactic ambiguity resolution

    Journal of Memory and Language

    (1994)
  • H.K.J. van der Lely

    Canonical linking rules: Forward versus reverse linking in normally developing and specifically language-impaired children

    Cognition

    (1994)
  • R.P. Abelson

    Statistics as principled argument

    (1995)
  • C.J. Angiolillo et al.

    Experimental evidence for agent–patient categories in child language

    Journal of Child Language

    (1982)
  • M.C. Baker

    Thematic conditions on syntactic structures: Evidence from locative applicatives

  • E. Bates et al.

    Functionalist approaches to grammar

  • M. Bowerman

    Mapping thematic roles onto syntactic functions: Are children helped by innate linking rules?

    Linguistics

    (1990)
  • R. Brown

    Linguistic determinism and the part of speech

    The Journal of Abnormal and Social Psychology

    (1957)
  • L. Burzio

    Italian syntax: A government-binding approach

    (1986)
  • G. Carlson

    Thematic roles and their role in semantic interpretation

    Linguistics

    (1984)
  • N. Chomsky

    Lectures on government and binding

    (1981)
  • H.H. Clark et al.

    The semantics of sentence subjects

    Language and Speech

    (1971)
  • J. de Villiers

    The process of rule learning: A new look

  • R.M.W. Dixon

    Ergativity

    (1994)
  • D. Dowty

    Thematic proto-roles and argument selection

    Language

    (1991)
  • J.E. Emonds

    Subcategorization and syntax-based theta-role assignment

    Natural Language and Linguistic Theory

    (1991)
  • C.J. Fillmore

    The case for case

  • C.J. Fillmore

    The case for case reopened

  • C. Fisher

    Structure and meaning in the verb lexicon: Input for a syntax-aided verb learning procedure

    Language and Cognitive Processes

    (1994)
  • C. Fisher

    From form to meaning: A role for structural analogy in the acquisition of language

    Advances in Child Development and Behavior

    (2000)
  • C. Fisher

    Structural limits on verb mapping: The role of abstract structure in 2.5-year-olds' interpretations of novel verbs

    Developmental Science

    (2002)
  • W. Foley et al.

    Functional syntax and universal grammar

    (1984)
  • W.N. Francis et al.

    Frequency analysis of English usage: Lexicon and grammar

    (1982)
  • D. Gentner et al.

    Structural alignment facilitates the noticing of differences

    Memory and Cognition

    (2001)
  • D. Gentner et al.

    Structure mapping in analogy and similarity

    American Psychologist

    (1997)

    • Cited by (52)

    • Grammatical cues to subjecthood are redundant in a majority of simple clauses across languages

      2023, Cognition

    • Encoding of event roles from visual scenes is rapid, spontaneous, and interacts with higher-level visual processing

      2018, Cognition
      Citation Excerpt :

      In particular, some theories of event roles hypothesize that certain components of events (e.g., contact, causation, and change of state or motion) are conceptual primitives, posited as such because they are relevant for grammar (i.e., they constrain the sentence frame in which a verb can be used; Levin, 1993; Levin & Rappaport-Hovav, 2005; Pinker, 1989; Talmy, 2000) or because they are available early on in development (Strickland, 2016). Notably, these event components are similar to features proposed in cluster-concept notions of event roles (Dowty, 1991; Kako, 2006; White et al., 2017). Although the consistency we observed in the role cost across events is broadly suggestive of generality (see Fig. 5, and Section 5.3.3), we do not believe we have a convincing way to address the precise characteristics of this generality with the current data, for the following reasons.

    • A Method for Studying Semantic Construal in Grammatical Constructions with Interpretable Contextual Embedding Spaces

      2023, arXiv

    • Electrophysiological hallmarks for event relations and event roles in working memory

      2023, Frontiers in Neuroscience

    • Agency effects on the binding of event elements in episodic memory

      2023, Quarterly Journal of Experimental Psychology

    • A Method for Studying Semantic Construal in Grammatical Constructions with Interpretable Contextual Embedding Spaces

      2023, Proceedings of the Annual Meeting of the Association for Computational Linguistics
    View all citing articles on Scopus
    View full text
    Copyright © 2005 Elsevier B.V. All rights reserved.