Elsevier

Cognition

Volume 89, Issue 1, August 2003, Pages B27-B41
Cognition

Brief article
Processing correlates of lexical semantic complexity

https://doi.org/10.1016/S0010-0277(03)00069-6Get rights and content

Abstract

This paper explores how verb meanings that differ in semantic complexity are processed and represented. In particular, we compare eventive verbs, which denote causally structured events, with stative verbs, which denote facts without causal structure. We predicted that the conceptually more complex eventive verbs should take longer to process than stative verbs. Two experiments, a lexical decision task and a self-paced reading study, confirmed this prediction. The findings suggest that (a) semantic complexity is reflected in processing time, (b) processing verb meanings involves activating properties of the event structure beyond participants' roles, and (c) more generally, lexical event structures, which subsume thematic roles, may mediate between syntactic knowledge and semantic interpretation.

Introduction

How are verb meanings processed and represented? Several studies have provided processing evidence showing that lexical semantic properties such as thematic roles and argument structure are quickly accessed by the processor when the verb is recognized (e.g. Altmann, 1999, MacDonald et al., 1994, McRae et al., 1997, Tanenhaus and Carlson, 1989, Tanenhaus et al., 1989, Trueswell et al., 1994). For example, Trueswell et al. (1994) found that the semantic fit between the subject NP and the verb's thematic structure plays a critical role in facilitating the reading time of ambiguous garden path constructions. Similarly, several studies have shown that verbs are typically associated with a preferred (most frequent) argument structure, even in contexts biasing the less frequent structure (MacDonald, 1994, McElree, 1993, Shapiro et al., 1993, Shapiro et al., 1987, Trueswell et al., 1993).

Other studies investigating word recognition in restricted contexts or in isolation have also suggested that argument and thematic structure have processing correlates. Ferretti, McRae, and Hatherell (2001) have shown that verbs prime their typical agents and patients (e.g. praying primes nun) when presented in isolation and in consistent syntactic contexts, but not in inconsistent contexts. They argue that verb meanings are represented in structured situation schemas containing participant slots plus sets of typical thematic features that are computed on-line. Moreover, McKoon and Macfarland, 2000, McKoon and Macfarland, 2002 have found that recognition times of verbs denoting externally caused events (break) are longer than those of internally caused verbs (grow). These verb classes differ in that externally caused verbs, but not internally caused ones, conceptually require two participants. Their findings thus show that semantic arguments give rise to complexity effects. Taken together, these findings suggest that thematic and argument structures are processed when the verb meaning is activated to construct a schematic representation of the event referred to. Such lexical properties, which are projected onto sentential arguments, are thus taken to mediate between parsing mechanisms and sentential semantic interpretation (e.g. Tanenhaus et al., 1989).

It is commonly assumed, however, in both linguistics and psycholinguistics that thematic roles and syntactic arguments do not exhaust verb meanings. Rather, these properties may be derived from other properties of the event structure (MacDonald et al., 1994, Tanenhaus and Carlson, 1989, Tanenhaus et al., 1989). Lexical semanticists in particular argue that thematic roles and argument structure are subsumed by the event structure defined in terms of causal properties (Croft, 1998, Dowty, 1991, Grimshaw, 1990, Jackendoff, 1987, Ladusaw and Dowty, 1988, Van Valin and Wilkins, 1996). For example, the agent in a breaking event is the initiator of a causal chain affecting the patient. This sort of information is required to semantically distinguish verbs such as love and discover, which are both associated with the 〈experiencer, theme〉 thematic structure. The critical property distinguishing these verbs is whether they denote a change of states.

Linguists and philosophers argue that several verb types can be distinguished depending on their causal structure. One general distinction typically made is that between states and events (Davidson, 1971, Dowty, 1979, Jackendoff, 1990, Jackendoff, 1991, Moens and Steedman, 1988, Parsons, 1990, Pustejovsky, 1991, Rappaport Hovav and Levin, 1998, Smith, 1991, Taylor, 1977, Vendler, 1967, Verkuyl, 1993): eventive verbs entail one or several changes from an initial situation to a resulting one (e.g. destroy, discover, play), while stative verbs entail a single stable situation (e.g. love, belong, contain). These verb types thus differ in their ability to entail sub-situations and changes (the change-of-state entailment, cf. Dowty, 1979). This fundamental conceptual distinction between eventive and stative verbs led lexical semanticists to propose that their lexical meanings differ in complexity: eventive verbs entail simpler conceptual units such as CAUSE, BECOME or CHANGE and resulting STATE, corresponding to the event's internal dynamics they denote, while stative verbs lack any such causal entailments. Table 1 exemplifies this sort of logical analysis and various event substructures classified as eventive verbs: accomplishment verbs (e.g. break), achievement verbs (e.g. discover), and activity verbs (e.g. carry), each defined by the presence/absence of the relevant sub-component.

The question now arises whether there are empirical correlates of such event representations, and, in particular, correlates of their differences in semantic complexity. If semantic properties are immediately processed when the verb is activated, as in previous findings, and if the number of arguments/roles is kept constant, differences in semantic complexity may emerge: eventive verbs, which have more complex meanings (more entailed properties), should take longer to process than stative verbs. The rationale for this hypothesis is that processing the meaning of a stative verb requires the activation of a single non-dynamic situation. In contrast, processing the meaning of a dynamic event entails activating more substructure or entailed sub-situations, such as an initial state, a change and a final state. In processing terms, more semantic structure (however represented) is accessed or activated with eventive verbs. More time would then be required for the relevant activation to spread (cf. Seidenberg & McClelland, 1989), and therefore, for the reader to fully process the meaning.

In this paper, we ask whether eventive and stative verbs, which by hypothesis differ in their semantic-conceptual complexity, are processed in ways consistent with this complexity difference. Although the processing and semantic literature suggests that semantic complexity effects are likely to arise, no such effects have been demonstrated. Early studies on verb meanings failed to find evidence supporting semantic complexity effects (e.g. Fodor et al., 1980, Rayner and Duffy, 1986). These findings, together with Fodor's and colleagues' theoretical arguments (Fodor, 1998, Fodor and Lepore, 1998), were taken to support the view that verb meanings lack internal causal structure.

Section snippets

Studies

To test whether the semantic complexity of eventive and stative verbs involves differential processing costs, we conducted two psycholinguistic experiments. The first experiment was a self-paced reading study, in which the reading time of verbs occurring in sentences was measured. To exclude an explanation of complexity effects in which reading times are due to the cost of integrating the verb with prior information, the second experiment was a visual lexical decision task. This task has been

General discussion

The results of these experiments show that events take longer to process than states by about 25 ms. These findings cannot be attributed to argument structure differences among verbs, to the assignment of thematic roles to sentential arguments, or to plausibility relations between subject and verb. These alternative interpretations are excluded by the factors controlled for and by the lexical decision results. Instead, they must be attributed to the manner in which semantic or conceptual

Acknowledgements

We thank Amy Weinberg, Juan Uriagereka, Gerry Altmann, Colin Phillips and Maryellen MacDonald for insightful discussions, support and encouragement of this work.

References (50)

  • L. Shapiro et al.

    Sentence processing and the mental representation of verbs

    Cognition

    (1987)
  • J. Trueswell et al.

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

    Journal of Memory and Language

    (1994)
  • D. Balota

    Visual word recognition: the journey from features to meaning

  • D. Balota et al.

    On the early influence of meaning in word recognition: a review of the literature

  • F. Blesdale

    Concreteness-dependent associative priming: separate lexical organization for concrete and abstract words

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (1987)
  • C. Chiarello et al.

    Imageability and distributional typicality measures of nouns and verbs in contemporary English

    Behavioral Research Methods, Instruments, and Computers

    (1999)
  • J.D. Cohen et al.

    PsyScope: a new graphic interactive environment for designing psychology experiments

    Behavioral Research Methods, Instruments, and Computers

    (1993)
  • M. Coltheart et al.

    Access to the internal lexicon

  • W. Croft

    Event structure in argument linking

  • D. Davidson

    Eternal vs. ephemeral events

    Nous

    (1971)
  • D. Dowty

    Word meaning and Montague grammar

    (1979)
  • D. Dowty

    Thematic proto-roles and argument selection

    Language

    (1991)
  • J. Fodor

    Concepts

    (1998)
  • J.A. Fodor et al.

    The emptiness of the lexicon: Reflections on James Pustejovsky's The Generative Lexicon

    Linguistic Inquiry

    (1998)
  • J. Grimshaw

    Argument structure

    (1990)
  • Cited by (68)

    • How language and event recall can shape memory for time

      2019, Cognitive Psychology
      Citation Excerpt :

      Linguistic phrases such as building a house have the power to cue prior event knowledge stored in semantic memory, often referred to as event concepts or schemas. Event schemas may contain information about typical actors, event structure and typical event features (Coll-Florit & Gennari, 2011; Ferretti, McRae, & Hatherell, 2001; Gennari, 2004; Gennari & Poeppel, 2003), and are thought to emerge from regularities extracted from experience over time. One enduring question in memory and language research is whether and how linguistic concepts shape the memory representation of an observed event.

    • He Gave My Nose a Kick or He Kicked My Nose? Argument Structure Alternations and Event Construal

      2018, Psychology of Learning and Motivation - Advances in Research and Theory
    • Neurolinguistic Processing of Psychological Verbs

      2015, International Encyclopedia of the Social & Behavioral Sciences: Second Edition
    View all citing articles on Scopus
    View full text