5.1 Informativity and conversational implicature

Example (2), repeated here as (10), works because of some shared expectations that speakers A & B have regarding the appropriate amount of information that is normally needed to answer a question.

In (10), speaker A asks about two things (taking Pragmatics and liking it) and an appropriately informative answer would address both pieces. What speaker B does instead is answer only the first part of the question, giving a seemingly underinformative answer. Although the literal meaning of B's utterance may be underinformative, it conveys a good deal more meaning (namely that speaker B didn't like Pragmatics). This additional meaning arises based on a calculation about what a full answer would have been and why an answer to the second part of the question might not have been included, despite the violation of the expectation that answers be informative. In general, statements that portray something in a negative light are dispreferred.

To talk about how information is measured, we can use what we know about entailment as a (somewhat crude) measure of informativity. Consider the example in (11):

In general, when a proposition is true in fewer states of affairs (fewer possible worlds), it is said to be more informative. Because (11a) is true in fewer worlds than (11b), we say that (11a) is semantically stronger (more informative) than (11b). This is also the case for (12-14):

For (12-14), make sure you understand why (a) is more informative (rules out more possible worlds) than (b). Now, reconsider (10) and two possible answers to Speaker A's question:

In (15), (a) entails (b) but not vice versa. With the measure of informativity in terms of semantic strength, we can say that (15a) is a stronger and more informative statement than (15b). But speaker B chose to say only (15b). This raises an implicature that some part of (15a) makes it not true, namely the second half. This kind of implicature is called a conversational implicature because it depends on the particulars of the conversational context in order to arise. In the exchange in (10), the conversationally implicated proposition is that the speaker did not like Pragmatics.

What are the properties of conversational implicatures and how are they diagnosed? First, conversational implicatures can be denied without contradiction. Although speaker B's response (15b) implicates that the speaker did not like Pragmatics, that meaning can be denied without creating a situation where a speaker sounds like they are reversing what they just said, as in (16). This words precisely because the implicature was never stated overtly and is hence not 'on the record' as what was said.

Second, conversational implicatures can be reinforced without redundancy. Again, because the speaker is not on record as having stated the implicated meaning, a subsequent statement that makes that meaning explicit does not sound redundant. Compare the overt statement of the implicated meaning in (17) with the (re-)statement of the literal meaning in (18).

Third, conversational implicatures are calculable, meaning that rather than arising automatically, they must be derived via a chain of reasoning. In the case of the Pragmatics example, the reasoning involves expectations of information quantity among cooperative speakers. Other implicatures arise via other cooperative principles, as will be discussed in the next section.

Lastly, conversational implicatures are context-dependent. They don't arise automatically but require the right context to give rise to them. See how the same statement from speaker B in (19) yields no implicature.

To go on to section 5.3 on "Grice's maxims", click here.