=====================================================================
Stanford Phonology Workshop

Christine Bartels
Cambridge University Press

Thursday May 20th, 7:30pm

=====================================================================
Towards a Formal Semantics of English Phrasal Intonation
(joint work Arthur Merin, IMS, Stuttgart)
====================================================================

English utterances exhibit characteristic pitch movements that contribute to utterance meaning in context. For instance, many (but not all) statements, wh-questions, and alternative-questions involve a salient fall from the target level of the last pitch accent to the speaker's pitch baseline ("declarative fall"). Many (but not all) yes/no questions, by contrast, show a final rise. A final rise is also observed in the common fall-rise contour, including the so-called contradiction contour.

The description and compositional phonological analysis of intonation contours cannot get far without assumptions as to what constitutes linguistically relevant contrasts in this domain - that is, a semantics. But the task of mapping sound into meaning is made difficult by the fact that a given intonation contour, or 'tune', can have very different connotations in different contexts. Simplistic meaning correlates such as 'statement' and 'question' turn out to be highly unreliable in predicting tune, regardless of how these utterance types are defined. Discourse-epistemic denotata such as those suggested by Pierrehumbert & Hirschberg (1990), involving 'given', 'new', and interclausal discourse dependencies, are much more successful; but they still fail to make (the right) predictions for all cases, such as the contrast between yes/no questions and wh-questions.

In this presentation we outline a social decision-theoretic model of intonational meaning that is more general though not less structured than current proposals based on discourse relations. We assume the Pierrehumbert (1980 et seq.) model of level tones as intonational phonemes, but we here take kinetic tones, i.e. transitions or movements of perceived pitch consisting of two or more tones from the Pierrehumbert inventory, to be the basic meaning-bearing units, i.e. intonational morphemes. In our model, the natural target domain for a semantics of intonation are the fundamental sociopolitical relations governing the establishment, maintenance, and negotiation of cooperation among potentially autonomous actors.

In its simplest form, our discourse model is built around a pair of cooperating actual or virtual persons, call them Ego (E) and Alter (A). A and E are engaged in joint decision making about an issue, P vs. non-P, as part of the goal of establishing a common ground (CG) of joint deontic-boulomaic or epistemic commitments. A and E are always in the business of persuading each other, rather than just speaking their minds. Their negotiations are in essence bargaining games, i.e. social situations in which interests are neither wholly opposed nor wholly consonant. Regarding the point at issue, P vs. non-P, we assume as default that their preferences are formally inverse; the paradigmatic question is always: 'Why () should I (do/believe that)?'. Negotiations on what becomes common ground proceed by Elementary Social Acts (ESAs) of Claim, Concession, Denial, Retraction (of a Claim). These are transitions to (and from) negotiation states. Negotiation states are characterized by n-tuples consisting of a proposition (P, or non-P) and four binary decision-theoretic parameters for negotiation states (Merin 1994). These parameters allocate ostensible actor-role, preference, dominance with respect to balance of incentives/warrant, and initiator role among E and A..

The formal ESA model turns out to have a broad range of uses in linguistic analysis, which are explored in recent works by Merin. The connection to intonation is this: intonational morphemes as we define them denote, in the first and core instance, (re-)allocations of the [D]-parameter value - i.e. of the power of choice - regarding the instantiation of variables under negotiation: either regarding propositions expressed by a whole sentence or clause or regarding focus-identified subsentential items that co-determine propositions. A rise (L* H) ostensibly alientates choice to an actual or virtual Alter, a Fall (H* L) appropriates it for the speaker, Ego. The complex movement of a Fall-Rise corresponds to a nesting of choice allocations.

In our presentation, we will first apply our model of intonational meaning to prosodically monophrasal statements and questions illustrating typical uses of falls, rises, and fall-rises; we will also examine certain phonologically and semantically distinctive subvarieties of these basic tunes that suggest refinements of the model. We will then offer further evidence for the validity of our decision-theoretic approach by applying the model to occurrences of those same basic tunes in utterance- and clause-medial position, specifically, in conditional clauses and topic phrases.