It is a long-held intuition in linguistics that negative sentences somehow "respond to" their corresponding affirmatives (Givón 1978; Horn 1989). Thus, the act of uttering a sentence like I'm not in love is presumed to be licensed by the presence of the underlying affirmative counterpart ("the speaker is in love") somewhere in the speaker's discourse model, whether explicitly via some prior utterance, or implicitly as an expectation shared by some other person(s) in the discourse. This is what I will term the Standard Position in the study of sentential negation.

Such widespread intuitions about sentence negation, however, have not found convincing empirical support in synchronic language use. In fact, a number of studies examining how sentential negation is used in naturally-occurring discourse have revealed that negative constructions are only rarely employed to counter previous assertions or even to challenge unstated assumptions (Tottie 1991; Thompson 1998). In this tradition of research, which I term the Conversationalist Position, negation is therefore not invariably tied to the presence of an underlying affirmative.

In this presentation, I use cross-linguistic evidence from English and several Romance languages to shed light on this ongoing debate. I show how both Positions ultimately make some sense. However, the relevant evidence is not to be found by analyzing canonical or standard sentential negative constructions, as has typically been done in the literature emerging from both Positions. Rather, it is found in the behavior of a set of non-canonical negative constructions, such as English not ... either and Portuguese não ... nada ('not ... nothing'), which owe their existence to dialogal (=dyadic) contexts where they express interlocutor denials. I show that these constructions and others like them can be placed on a synchronic continuum whose ordering metric is the degree to which a given negative "responds to" a salient affirmative proposition that is accessible in the discourse context and, especially, the degree to which a given negative is sensitive to a dialogic context. I illustrate the close connection between these negative constructions and information structural properties of the discourse, specifically the sensitivity that these negative constructions display to HOW the negated propositional content has entered into the current discourse model.