H. Clark and Wilkes-Gibbs (1986/1990), e.g. working with the well known tangram figure experiments, found out that referringly used NPs are introduced into discourse according to a presentation-acceptance cycle. This implies that one agent suggests some NP to another agent in order to refer to some tangram figure. It is then negotiated whether this NP is to be used in the sequel. In order to indicate that NPs are not set once and for all but open to further consideration, they are phonologically marked. The marking is used to indicate the suggesting agent's estimation concerning the applicability of the NP under consideration.
In a more recent study, Brennan and Clark (1996) maintained that agents establish a conceptual pact about how they conceptualise an object, which, of course, determines lexicalisation.
In two studies based on the maze task, Garrod and Anderson (1987) and Garrod and Doherty (1994), it was shown, that agents coordinate on the description of local positions of objects. In addition, the 1994 study revealed that coordination spreads fairly quickly within groups of agents and may thus be considered a convention in the Lewisian (1969) sense.
Coordination is also relevant concerning mental states like belief and knowledge: Heydrich and Rieser (1995), studying directives and answers by agents in task-oriented dialogue, arrived at the conclusion that agents in discourse point out what they consider to be public information and what must hence go into the common ground. Using eye-tracking technology, Pomplun, Rieser, Ritter and Velichkovsky (1997) discovered that agents even coordinate their foci of attention, mainly by controlling each others direction of gaze.
These and other empirical studies led to the assumption of various general principles concerning agents' behaviour in dialogue: Clark and Wilkes-Gibbs, e.g. maintain that two principles are active in dialogue, the principle of mutual responsibility and the principle of least collaborative effort. Clark and Brennan postulate a "grounding principle concerning reference", which amounts to the following:
"When speakers present a reference, they do so provisionally, and they then work with their addressees to establish that it has been understood. When speakers first refer to an object as the loafer, they are proposing to their addressees that it be conceptualised as a loafer. The addressee can ratify the proposal ("okay"), modify it ("you mean the man's shoe?"), or solicit another proposal ("which one?") in the process of grounding that reference."On their way to capture semantic coordination among agents in dialogue, Garrod and Anderson (1987) established an "output-input-coordination principle". It entails that the current speaker in formulating his utterance will try to match the lexical, semantic and pragmatic decisions used in the interpretation of the previous speakers' last utterance as closely as possible. "Bilateral conformity to the principle quickly produces convergence on a common description schema of the kind observed in the maze game dialogues". Heydrich and Rieser (1995) based their turn-exchange-model on a "principle of Lewisian coordination": An agent's belief in public information concerning the situation brought about by his last directive is a sufficient condition for the production of his next turn.
In the research reported, semantic and pragmatic matters have been focussed on, although indirectly syntactical form also played some role. But so far, there has been no systematic work done on coordination of morphology and syntax in dialogue. In this paper I will concentrate on matters of syntax, but at least some remarks concerning morphology and the syntax-morphology interface should be made. First a cautionary remark: Agents in dialogue show a language behaviour far more creative than usually assumed. This can be seen from their coining new words related to the task at hand. These ad-hoc coinings are then subject to the various principles referred to above, especially to the Clark/Brennan "grounding principle". This does not imply, however, that coordination of morphological forms must needs take place. Speakers may stick to their own morphological options, but if they do so, they still have to coordinate their choice of form with a suitable semantics, perhaps along the lines of the Garrod/Anderson "output-input-coordination principle". Roughly the morphological processes involved either follow the structural rules of the natural language in question or they follow some sort of new rule which is perhaps restricted to spoken language. I will provide examples to make clear the issue involved: In airplane out of a set of wooden materials consisting of bars with holes in them, bolts, nuts, and screw-threaded wooden cubes. At some point of the construction procedure this involves that they must name the parts they need. That is where morphological matters enter. Frequently we have the integration of formal aspects (shape) and functional aspects (role in aggregate building up the airplane) within one compound. The examples in (1) are consonant with morphological composition in German, those under (2) seem to be doubtful:
| (1) | Siebener | sevener |
| Siebenerstange | sevener-bar | |
| Siebenerverlängerung | sevener-extension | |
| Siebenerlangstück | sevener-longpiece | |
| Siebenerflügel | sevener-wing | |
| Siebenertragfläche | sevener-wing
| |
| (2) | Baufixgelbschraube | Baufix-yellow-screw |
| Siebenerlöcherndingern | sevener-holes[dative] things[dative] |
To round off the picture, I mention, thereby anticipating the introduction of phenomena playing a central role in the discussion of syntactic matters, that we also have self-repairs concerning morphological processes as in (3) and (4) as well as cooperative production of compounds as in (5). Below, I and C stand for Instructor and Constructor, respectively. The significance of these role indications will become clear in Sect. 2.1.
| (3) | Siebenerlöchern-äh-plättchen | |
| sevener-holes uhm bars
| ||
| (4) | I: | Eine Lorenz Baufixgelbschraube. |
| A Lorenz Baufix-yellow-screw. | ||
| C: | [laughs] Aha. | |
| I: | Oder eine gelbe Lorenz Baufixschraube. | |
| Or a yellow Lorenz Baufix-screw.
| ||
| (5) | C: | Ich habe eine rote Sechskantschraube in der Hand |
| I hold a red six-edge-screw in my hand | ||
| und eine Drei. | ||
| and a three. | ||
| I: | Ja, genau. | |
| Right, exactly so. | ||
| I: | Löchrige Schiene, so. | |
| with-holes bar kind of. | ||
| C: | Dreilöchrige Schiene. | |
| with-three-holes bar. |
Observe that in (5), the final morphological form is dreilöchrige/with-three-holes, where one part, drei/three, is contributed by the Constructor, and the other löchrige/with-holes is provided by the Instructor, who also does the final compounding. Roughly, we have the Clark-Gibsian "presentation-acceptance cycle" here but with respect to compound-formation.