3.4 Conclusion
4.1.2 Discourse contexts
When we speak, we usually do not utter meaningful expressions in a vacuum. Instead, our utterances have a certain effect on the context in which we utter them—at least we may hope so. Usually when conversation progresses, interlocutors build up a shared information base. In fact, increasing the amount of shared information in the discourse context is often taken to be the main goal of conversation. Traditionally, this effect is modeled as update on the common ground (cg), where the common ground is defined as a set of propositions shared by all the discourse participants. Updating thiscgsimply amounts to adding the proposition expressed by the utterance to this set (Stalnaker, 2002).
However, it is not always the case that the common ground is updated immediately after an utterance is made; performing a speech act does not always immediately result in a shared commitment. Such a view of discourse effects as immediately updating the cg may work well for assertions, but is hard to extend to other types of speech acts. Gunlogson (2004) therefore proposes that speaker’s individual commitments should be kept track of in a discourse context as well. The main effect of making an assertion is then to add its propositional content to the speaker’s individual commitment set. Also in other type of speech acts, like questions, the individual commitment set is affected before any changes are made to the common ground.
Farkas and Bruce (2010) find a middle ground between the initial idea of a shared set of commitments and individual commitments by suggesting that speech acts must be viewed as proposal to update thecg. In making such a proposal, the speaker also makes an individual commitment that can be turned into a shared commitment by the other interlocutors by accepting the proposed update of the common ground. As is shown in Farkas and Roelofsen (2017), such a perspective on discourse contexts and
the effects of speech acts interfaces very neatly with the inquisitive semantics frame- works, since issues can be viewed as proposals to update the common ground in dif- ferent ways: each info state contained in the proposition that is expressed represents a proposed update of thecg. I will adopt a version of this view on discourse contexts too.
In Farkas and Roelofsen (2017), discourse contexts keep track of the commitments of individual speakers and of the proposals that these speakers have made to update the commong ground. A discourse context is therefore defined as follows.
Definition 4. A context C is a triplehA,T,CSi, where a. Ais the set of discourse participants;
b. T is a stack of propositions, representing the proposals that have been made in the context so far.
c. CS is a function that maps every participanta∈Ato a set of info states, those info states thatais publicly committed to.
Here,T is called the Table, and may contain both inquisitive and non-inquisitive propo- sitions. The discourse effects of questions and assertions are the same in the sense that both have the same effects onCS andT. In other words, questions and assertions are distinguished by means of their semantics, but after the semantic content of these expressions are determined, the conventions of use of both can be defined in the ex- act same way. More specifically, performing a speech act always has two effects on the discourse context: (i) the commitment set of the speaker is updated and (ii) the proposition expressed by the speech act is put on the Table.
Definition 5. If a discourse participant a utters a sentence ϕ, the discourse context is affected as follows:
1. The proposition expressed byϕ,~ϕ, is added to the Table.
2. The informative content ofϕ, info(ϕ) is added to CS(a)
Speech acts would then be indexed for the specific discourse participant that performs them. We therefore only update the specific commitment set of that speaker.
Some auxiliary notions are defined too. The context set of an agenta, denoted by cs(a), will be the set of worlds that are compatible with the agent’s public commitments at some point in the conversation. This can be derived fromCS:cs(a)=∩CS(a).
Using this notion of a context set, the Stalnakerian notion of a common ground is recoverable too: cg=S
a∈A{cs(a)}. This means that the common ground is the smallest
set of info states that all participants agree on the actual world to be part of. For example, if agent a is publicly committed to believing that the actual world is a p- world, and agent b is committing to believing the actual world to be a q-world, the common ground will consist of states|p|and|q|.
Even when having this distinction between a proposition and a speech act in place, we can ask: why is it that speech acts cannot be disjoined? Perhaps one of the most articulated accounts of this can be found in Krifka’s work, in which it is argued that uttering a disjoined speech act will result in an ill-formed commitment state. I therefore discuss this proposal below and I will show that even though this account might work well for disjoined assertions, it does not straightforwardly extend to disjoined questions in an inquisitive setting. I will therefore adjust Krifka’s basic intuition in such a way that it can be implemented into a framework similar to that of Farkas and Roelofsen (2017).