Towards flexible types with constraints for manner and factive adverbs Introduction

In this chapter, I discuss the syntax and semantics of so-called factive and manner adverbs such as in (1a) and (2a), with their respective paraphrases in (1b) and (2b).¹

(1) a. Stupidly, Bill hit Jill.

b. The fact that Bill hit Jill was stupid.

(2) a. Bill hit Jill passionately.

b. The manner in which Bill hit Jill was passionate.

I will provide syntactic and semantic evidence to support the distinction between these two classes. Other factive adverbs are wisely, rudely, and appropriately; other manner adverbs are gently and roughly. For the moment, let us say that stupidly in (1a) is a sentence adverb, while passionately in (2a) is a verb phrase adverb. Some factive adverbs are polysemous between a factive and a manner interpretation (e.g., rudely), but I do not address polysemy in this chapter.²

In a recent overview of research on the syntax and semantics of adverbs, Lang, Maienborn and Fabricius-Hansen (2003: 1, 25) present two widely accepted assumptions concerning the syntax and semantics of adverbs:

Almost any treatment of adverbials starts from a long-established classification of adverbials that is somehow based on semantic intuition. . . . Normally, syntacticians and semanticists make different choices in selecting a subset of these types, by starting their approach with a division into, say VP- vs. sentential adjuncts or predicates vs. operators, and then concentrate on finding and justifying refined subdivisions below that intuitively assumed level. (Lang et al. 2003: 4)

They assume that one should distinguish classes of adverbials in terms of the syntactic and semantic category they modify; for example, factive adverbs are modifiers syntactically of sentences and semantically of propositions, while manner adverbs are modifiers syntactically of verb phrases and semantically of predicates. Moreover, given the classification, they implicitly assume that these types are fixed; that is, once an adverbial type is ascribed to an adverb, it cannot be changed. These two assumptions are based on syntactic tests and semantic intuitions, some of which I review later. I refer to these two assumptions together as the “fixed types hypothesis” of adverbial modification and a theory which makes use of this hypothesis as a “fixed types theory.” In other words, put broadly, if an adverb is a factive adverb, it must apply to the sentence level and not the VP level, while if it is a manner adverb, it must apply to the VP level and not the sentence level. In general, adverbs can only appear adjoined to syntactic and semantic positions associated with their syntactic and semantic category. If they do not, the resulting sentences are ungrammatical. Where adverbs do in fact appear in alternative positions, some other explanation must be given, for example with movement or as exceptions. These basic assumptions appear in a broad spectrum of syntactic and semantic research, as I report later.

Other grammatical and semantic frameworks do not adhere to the fixed types hypothesis, for example allowing an adverb to apply to a range of syntactic positions and correlated semantic expressions. I refer to this as the “flexible types hypothesis,” and a theory which makes use of this as a “flexible types theory.” For instance, in such an approach, a manner adverb might have both an S- and a VP-level expression, so the adverb could apply in either position. Such an approach could account for a range of alternative positions that are problematic for a fixed types theory. However, it has a reciprocal problem: it must then account for the syntactic tests and semantic intuitions which support the fixed types theory. One way to do so is to provide constraints on the application of the types. Yet, so far as I am aware, this possibility has largely gone unexplored for the problems I examine.

To decide between the two types of theories, I consider cases of discourse anaphora which paraphrase the sentential examples. I observe that the available and unavailable cases of discourse anaphora parallel available and unavailable cases of adverbial modification within the sentence. These observations lead to the most important contribution of this chapter, which is the identification of formal mechanisms from the semantics of discourse which are used to constrain the distribution and interpretation of factive and manner adverbs within the sentence. A fixed types theory cannot account for the discourse cases since it is a theory of sentence grammar. A flexible types theory can accommodate both discourse and sentence facts given the constraints. I then suggest a theory which unifies the interpretations and distributions in both the sentence and discourse cases. The analysis indicates that for the syntax and semantics of factive and manner adverbs, sentence grammar and discourse grammar share some key components.

As I have made this argument in detail previously (Wyner 1994, 1998b), in this chapter, I only briefly present the main lines of the argument again, focusing on the conceptual and empirical issues. In particular, I emphasize what is at stake in the choice between fixed types and flexible types with respect to adverbial syntax and semantics. I leave the interested reader to consult Wyner (1998b) for an explicit formal analysis within the approach of Discourse Representation Theory (Kamp and Reyle 1993 and Asher 1993).

1 Such paraphrases are found in Reichenbach (1947: 301–307), Quirk, Greenbaum, Leech, and Svartvik (1985: 557, 612–631), Jackendoff (1972: 49, 52), Parsons (1990: 62–64), Asher (1993), and Moore (1993: 136), among others.

2 There are a range of other approaches to what I call here factive adverbs, in terms of vagueness (Barker 2002), conventional implicatures (Potts 2005), and so-called disjuncts (Espinal 1991), among others. These approaches are tangential to the core facts and issues I discuss here.