A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Understanding an utterance requires inferring speaker intention beyond literal meaning. When someone asks 'Can you pass the salt?' they're not truly requesting information about your ability—they're implicitly requesting that you pass the salt. Pragmatic interpretation depends on shared context, mutual knowledge, and cooperative principles. Violations of these create comprehension problems or are deliberately used (e.g., irony) to communicate different meanings.
Analyze real conversations showing pragmatic inference and contrast with literal interpretations. Use examples like indirect requests, irony, sarcasm, and metaphor to show how context licenses non-literal meaning. Conduct experiments showing that listeners derive implicatures rapidly and unconsciously.
You have studied language comprehension — how listeners parse sentence structure and recover literal semantic content from words and syntax. But you know from everyday experience that what a speaker means routinely exceeds what their words literally say. "Nice weather we're having" said during a storm means the opposite. "Can you pass the salt?" is a request, not a question about your motor ability. The philosopher H.P. Grice proposed that this gap is bridged by the Cooperative Principle: speakers and listeners implicitly assume that conversations are governed by rational cooperation. This generates four Gricean maxims — quantity (be as informative as required, not more), quality (say what you believe to be true), relation (be relevant), and manner (be clear and orderly). These maxims are not rules people consciously follow; they are assumptions that license inference.
The key inferential mechanism is: when a speaker appears to violate a maxim, the listener doesn't conclude the speaker is irrational — they infer a meaning that rescues the assumption of cooperation. Suppose you ask "How did Sarah do on the exam?" and the reply is "She didn't miss a single lecture." The response seems irrelevant to the question (relation maxim apparently violated). But a cooperative listener infers: "This reply must be relevant in some way — perhaps it implicates that diligence led to a good result, or that the speaker won't commit to a stronger positive claim." The inference is a conversational implicature — a meaning communicated but not semantically encoded. Crucially, implicatures are cancelable: you could add "…but she still failed" without logical contradiction, which distinguishes them from semantic entailments that cannot be canceled.
Your prerequisite in theory of mind is directly relevant here. Computing implicature requires modeling the speaker's mind: "Given that this person is cooperative and has chosen *these* words in *this* context, what must they intend me to infer?" This is second-order mental state reasoning — you attribute to the speaker an intention to communicate a particular meaning, and you recover that meaning by reasoning about what a rational agent would mean by saying this here. Irony and sarcasm demand even higher-order reasoning: the speaker says something they know to be false (violating quality), and the listener must detect the deliberate violation, attribute ironic intent, and recover the speaker's actual attitude. Research shows that individuals with autism spectrum conditions often process implicature more slowly or less accurately — consistent with the theory-of-mind demands of pragmatic inference.
A common intuition is that context *refines* meaning after the fact — you get the literal reading first, then adjust. Research on real-time language processing shows this is wrong. Contextual constraint is applied *immediately*, in parallel with lexical and syntactic processing, not as a post-hoc correction. In a strongly constraining context ("He spread the butter with the…"), readers begin fixating on the contextually appropriate word before encountering it. When an utterance is pragmatically anomalous — when what was said is implausible for the conversational context — processing difficulty increases measurably: reading times slow, and ERP measures show N400 effects (the neural signature of semantic processing difficulty). Context is not an optional layer added on top of semantic comprehension; it shapes the comprehension process from the first word. Meaning is always meaning-in-context, and there is no prior context-free stage that yields "the literal interpretation" to be then adjusted.