Week 7: Pragmatics for Empirical Methods

This year's Empirical Methods project is in the domain of Pragmatics, which is the study of language in context. The web reading here provides theoretical background to the field of Pragmatics, particularly with reference (hah, pun) to the topic of Reference. Before starting this reading, please make sure you've read the two entries from the Encyclopedia of Language and Linguistics (the entry on "Reference: Psycholinguistic Approach" and the one on "Reference: Psycholinguistic Approach"). These readings are listed on Learn. This web reading will cover the following four topics.

1. Intro to Pragmatics

Other areas of linguistics (e.g., phonetics, phonology, morphology, syntax, semantics) target the underlying systematicity of sounds and words and sentences. However, those sounds and words and sentences must be interpreted in context. Pragmatics is the subfield of linguistics that aims to understand how context affects production and comprehension. Consider the example in (1).

The sentence in (1) can of course be analysed at multiple levels. At the level of sound, we have tools from acoustic phonetics to analyse the phonemes within the words and the prosody of the sentence. At the level of meaning, we rely on tools from morphology, syntax, semantics and pragmatics. The formal structural connections within and between words is the domain of morpho-syntax. The relation between words and objects in the world is the object of study for semantics. In terms of the empirical questions that one can pose with regards to (1), researchers in morpho-syntax ask questions like What morphological patterns hold across words of these types?, What syntactic structure underlies the words in this sentence? or Is this sentence acceptable given this language’s grammar? Semanticists ask What meaning is composed from the words in this sentence? or Is this sentence evaluated as true in this world?

What pragmatics targets is the relation of words to those who use and interpret the words. The questions in pragmatics include ones for which context is key: What additional meaning is conveyed when this sentence is used in context? or Is this sentence felicitous as a way to convey that meaning in this context?

In a sense, each level of linguistic analysis is context dependent; the questions posed above for morphology/syntax/semantics must be evaluated relative to the grammar of a particular language or relative to the truth of situations in the world. These are fairly stable contexts: A language's grammar provides the consistent framework for composing sentences; our experience in the world provides the knowledge base for evaluating sentence meaning. For pragmatics, what's important is the context of use, something which changes very quickly. Speakers and listeners in conversation must keep track of the discourse context and must keep updating their understanding of the context as new situations and utterances are encountered. As such, pragmatics is inextricable from the people who are using the language to communicate. Speakers make choices of how to package what they want to say in a particular context (given all the ways they could express their intended meaning) and listeners have to figure out how to read between the lines and recover meaning that often goes beyond what was explicitly said (the recovery of another person's intention).

Communication depends on both what is said (the words a speaker produces) and what is meant (the meaning a speaker intends). Either of these components can be associated with misunderstanding -- a listener can mishear the words that were produced or misparse the sentence structure, but even when the listener can compute what was explicitly said, there is still ample room for miscommunication if a listener misinterprets why the speaker said what they said.

The following examples illustrate how what is meant often goes beyond what is said. In each case, consider how context is driving the recovery of the intended meaning.

In example (2), there is missing linguistic material that must be inferred. The place that Hannah went to is not specified (i.e., there is more that is meant beyond what is said), but it is easily recoverable given the linguistic context of the preceding clause. But to assess sentence (2) as true or false (its semantic meaning), one would need to know more than what the surface form provides. There's the potential ambiguity that arises with the use of the names Stefano and Hannah (which Stefano? which Hannah?) and the vagueness that arises with the use of words like Paris and went (which part of Paris? how did they travel there?) and even the time frame (past tense but when in the past?). In other words, there is a lot of meaning packed into (2) that reflects the speaker's knowledge of the situations they're describing but that is not made explicit in what the words say.

Example (3) shows another way that the meaning goes beyond what is explicitly said, and that comes down to the function of the utterance. The syntactic form of (3) is that of a yes/no interrogative question, but its function is ambiguous -- it could be used by a speaker to ask the listener whether they are physically able to open the door, but it's more likely that it's being used to try and get the listener to open the door (a request, one of whose necessary precondition is that the listener be physically able to execute the request). For (3), recovering the intended meaning requires going beyond the form of the utterance (a speaker's information-seeking question) to the underlying intention (a speaker's request for an action on the part of the listener). Distinguishing between a question about ability and a request for action depends on the listeners' knowledge about speaker conventions and also on the context of use.

Lastly, here in example (4), the meaning may depend on what the sentence is an answer to. If the sentence is answering the question "Who teaches Empirical Methods?", then its content is a direct answer to that question. However, if a speaker utters (4) as an answer to the question "Do you know Hannah?" and the listener believes the speaker is attempting to make a relevant contribution to the conversation, then (4) can be interpreted as conveying meaning quite separate from the explicit literal meaning associated with the words in the sentence. As an answer to the question "Do you know Hannah?", the speaker might use (4) to implicate the meaning "Yes" (yes, I know Hannah since she's my instructor for Empirical Methods). It's also possible that (4) would fail to convey that meaning given the existence of the more direct answer "Yes", in which case (4) might implicate instead that the meaning "Yes" is not entirely warranted -- for example, maybe the speaker has uttered (4) to convey that they know who Hannah is but haven't met her in person (though they've been subjected to her video lectures).

As (4) shows, there are a range of questions that could plausibly precede a given utterance and thereby influence its interpretation. These different questions constitute different contexts of use. Understanding how such contexts influence meaning is a core part of pragmatics.

Another core part of pragmatics is understanding which parts of an utterance's meaning constitute the explicit literal meaning and which parts have a different status. One type of non-literal status is implicated meaning, like the speaker-intended meanings in (3) and (4) that are distinct from the literal content of the utterances. There is also backgrounded meaning that we are often quite unaware of, which is also within the purview of pragmatics. For example, sentence (2) contains the clause "Stefano went to Paris"; this clause has truth-conditional meaning describing a situation in the past in which some referent named Stefano traveled to some place called Paris, but this clause also carries presupposed meaning, including that there exists a person named Stefano and a place called Paris. What implicated meaning and presupposed meaning have in common is that they are both very hard to deny.

If someone hears "Stefano went to Paris" and responds "No, that's not true", their denial seems only to be a denial of the explicit literal meaning (that Stefano went to Paris) rather than a denial that there exists a person named Stefano or a place called Paris. The implicated meaning that can be associated with (4) is similar: If a speaker says "Well, Hannah's the instructor for Empirical Methods" in response to the question about whether or not they know Hannah, someone who overhears this exchange and chimes in to say "No, that's not true" will likely only succeed at denying the explicit content (that Hannah is the instructor), not the implicated content about whether the speaker knows Hannah. This is because the implicated content was never explicitly asserted and is not part of the formal record of what has been said, and so it therefore is difficult to deny.

As you will see later in this reading (in Section 3 on "Definites and indefinites"), understanding backgrounded pragmatic meaning is useful for thinking about what a speaker conveys when they choose how to talk about something. The use of particular referring expressions ("a cup" versus "the cup", for example) carries meaning (e.g., "a cup" signals that the intended referent is a cup that the listener is unfamiliar with or that is non-unique, whereas "the cup" signals that the referent is one that is familiar and unique), but the meaning is similar to other pragmatic meanings in being hard to deny and not explicitly part of what the speaker is on record as asserting.

All of these issues raise questions about how speakers and listeners draw these kinds of context-driven inferences in real-time during communication. Such questions are addressed by researchers in experimental pragmatics, who use psycholinguistic and corpus tools. The following list shows some of the breadth of topics in experimental pragmatics:

• Discourse coherence: How do speakers/listeners draw connections between related topics and related ideas?
• Information structure: What forms convey the discourse status of information (new/old/inferred/focus/topic/etc.)?
• Literal vs implicated meaning: When and how do listeners assign these meanings?
• Ambiguity resolution: What contextual cues guide listeners' interpretation of what is said?
• Choices in production: What contextual cues guide speakers' decisions of how to express what they mean?

All of these topics have connections to research in other subfields of linguistics. In the area of first and second language acquisition, there are questions of how learners grapple with pragmatic constraints alongside other acquisition challenges. In computational linguistics, there are big challenges in building machines that can read between the lines to understand human language and produce natural-sounding conversation. At interfaces with other levels of linguistic representation, there are questions about how pragmatic information influences and is influenced by cues like the syntactic structure of a sentence or the prosody a speaker uses.

To go on to section 2 on "Reference", click here.