In this chapter we set out the broad descriptive generalizations which emerge from investigations of argument realization and its lexical semantic underpinnings, as well as the methodological issues which a comprehensive theory of argument realization must address. These constitute the major challenges for a theory of lexical semantic representation and a theory of argument realization that dovetails with it.

Taking lexical semantic representations seriously

Since the 1980s, many theories of grammar have been built on the assumption that the syntactic realization of arguments – their category type and their grammatical function – is largely predictable from the meaning of their verbs. Such theories take many facets of the syntactic structure of a sentence to be projections of the lexical properties of its predicator – its verb or argument-taking lexical item; see Wasow (1985) for discussion. To ensure this, these theories incorporate conditions requiring that the arguments of the verb are appropriately represented in the syntactic representation of its clause. Such principles include the Principles and Parameters framework's Projection Principle (Chomsky 1981: 29, 38), Lexical-Functional Grammar's Completeness and Coherence Conditions (Kaplan and Bresnan 1982: 211–12), and Role and Reference Grammar's (RRG's) Completeness Constraint (Foley and Van Valin 1984: 183; Van Valin 1993b: 74–75; Van Valin and LaPolla 1997: 325–26). The successful implementation of the program of deriving the syntactic properties of verbs from facets of their meaning depends on the existence of both an articulated theory of the lexical semantic representation of verbs and a theory of the mapping between this representation and the relevant syntactic representation.

The discussion of the semantic determinants of argument realization in previous chapters centered on individual arguments of a verb, but interactions between arguments that affect argument realization cannot be ignored. These interactions suggest that there are precedence – or prominence – relations among arguments statable in terms of their semantic roles. This chapter examines the possible sources of these relations through an in-depth examination of the construct known as the thematic hierarchy – a ranking of semantic roles chosen because it figures in the statement of linguistic generalizations. Although this construct has been widely implicated in linguistic phenomena, particularly those involving argument realization, many conflicting formulations have emerged, leading some researchers to voice skepticism concerning its validity. It is for this reason that we devote an entire chapter to the thematic hierarchy. Three major questions are the focus of this chapter: which notion of prominence underlies the ranking that defines a particular formulation of the thematic hierarchy, which linguistic phenomena is it supposed to account for, and can its effects be derived from more basic components of a lexical semantic representation?

We show that two major conceptions of prominence find their way into analyses of linguistic phenomena that appeal to a thematic hierarchy. On the first, the ranking of semantic roles is determined by structural properties of a lexical semantic representation – properties defined over a predicate decomposition or event structure. On the second, the ranking is determined by entailments associated with arguments. Our examination of these two conceptualizations supports the conclusions of earlier chapters that the mapping from lexical semantics to syntax needs to make reference to arguments in both ways.

In the previous chapter, we pointed out that the term “event structure” is now widely used to refer to the lexical semantic representation which determines argument realization. This term reflects a consensus that such representations encode properties of events. Nonetheless, there are fundamental differences among the representations that have been proposed. Many of these stem from alternative hypotheses about which semantic properties of events influence argument realization and, thus, are central to the organization of event structure. Our goal in this chapter is to delineate theories concerning these semantic properties. Semantic properties of events are shown to be relevant for the organization of event structure to the extent that the subclasses of events which they define share identifiable grammatical properties. We present a discussion of these properties under the rubric of theories of event conceptualization since a hypothesis about what facets of an event are grammatically relevant is a hypothesis about how language users conceptualize happenings in the world for linguistic encoding. It is reasonable to assume that those properties of events that are grammatically relevant are also cognitively salient in some pretheoretically intuitive way (B. Levin and Pinker 1991) and that such properties should find their way into semantic representations.

As we review in this chapter, broadly speaking, three ways of conceptualizing events have been proposed to be grammatically relevant; each focuses on a distinct cognitively salient facet of events. The first, the localist approach, highlights the notions of motion and location. The second, the aspectual approach, suggests that the temporal properties of events, including their mereological (part–whole) structure, are central.

As discussed in chapter 1, generative grammar traditionally assumed the realization of a verb's arguments to be determined by information registered in a structured lexical entry for that verb. In early versions of generative grammar, the syntactic expression of a verb's arguments is directly encoded in its lexical entry in the form of a subcategorization frame. Later, semantically based representations of argument-taking properties replace subcategorization frames (Bresnan 1982b; Marantz 1984; Pesetsky 1982; Rappaport and B. Levin 1988; Stowell 1981; Williams 1981), and the realization of a verb's arguments is calculated from these “argument structures” via mapping algorithms in ways illustrated in chapter 5. In Rappaport Hovav and B. Levin (1996, 1998a), we call such theories projectionist. There are a variety of projectionist theories which all share the fundamental assumption that a verb's lexical entry registers some kind of semantically anchored argument structure, which in turn determines the morphosyntactic expression – or projection – of its arguments.

In this chapter we consider a set of phenomena involving argument realization which pose a challenge to projectionist theories, the phenomena falling under the label multiple argument realization. Perhaps its most intensively studied manifestation is what we called argument alternations. This form of multiple argument realization is typified by pairs of sentences with the same verb, related by paraphrase or subsumption. The dative and locative alternations are illustrated in (1) and (2). A third alternation is illustrated in (3); as it lacks a conventionally accepted name, we refer to it as the with/against alternation (B. Levin 1993).

One of the most widely adopted forms of lexical semantic representation is what we term a semantic role list, also known as a “case frame” (Fillmore 1968) or a “theta-grid” (Stowell 1981). The best-known instantiations of this approach are Fillmore's Case Grammar (1968) and Gruber and Jackendoff's thematic relations (Gruber 1965; Jackendoff 1972, 1976), but such representations have a long history, going back to the Sanskrit grammarian Pānini. The modern interest in semantic role lists originated in Fillmore's work, especially as presented in “The Case for Case” (1968). In a semantic role list, grammatically relevant facets of a verb's meaning are represented by a list of labels identifying the role that each of the verb's arguments plays in the event it denotes. For example, the verbs break and put might be associated with the semantic role lists “Agent, Patient” and “Agent, Theme, Location,” respectively.

Rather than comparing and contrasting the myriad proposals for semantic role list representations, this chapter provides a critical discussion of these approaches. First, in section 2.1 we present the essential properties of semantic role lists. Then, in section 2.2 we discuss their most fundamental limitations. In section 2.3 we review some attempts aimed at overcoming these shortcomings, while maintaining many basic assumptions underlying such approaches.

The properties common to semantic role list approaches

The components of a semantic role list are a predetermined set of labels that identify arguments according to the semantic relation they bear to their verb; each verb is associated with the relevant list of semantic roles.

Over thirty-five years have passed since the publication of Fillmore's “The Case for Case” (1968), the first substantive generative attempt to lay out a program of deriving the morphosyntactic realization of a verb's arguments from relevant semantic properties. In this book we have reviewed a considerable number of issues pertaining to this research program. In this postscript we step back from the myriad details and take stock of the overall picture which has emerged, asking how our understanding of this complex area has advanced since Fillmore's seminal paper. In the introduction we set out the issues which a comprehensive theory of argument realization must address. We now use these issues as a framework for summarizing the results presented in our survey of argument realization, reviewing what kind of consensus has emerged with respect to each. As we stressed at the outset, we do not present results meant to be embedded in any particular theory of grammar, and we have tried to abstract away from issues which divide linguists of different theoretical persuasions. We believe that a body of knowledge has accumulated about which there should be little dispute and which needs to be taken into consideration in developing a theory of argument realization regardless of the larger grammatical framework in which it is couched.

Which elements of meaning are relevant for the mapping from lexical semantics to morphosyntactic expression? As mentioned in the conclusion to chapter 4, broadly speaking these are aspectual notions, causal notions, and notions related to sentience or potential volitionality.