Neo-Gricean pragmatics

↵ Back to class homepage

In the years following the proposal of Grice's theory, many new theories and ideas have been suggested. Some of these are meant as alternatives to Grice's theory (i.e., by people claiming that Grice's theory—or at least some parts of it—was wrong and some new theory can explain pragmatics better), but others are meant as sort of refinements or improvements to Gricean theory. This latter group of theories, called "neo-Gricean" (i.e., new Gricean) theories, generally accept the general premises of Gricean theory, but have made some tweaks to address some of its problems or to better explain the phenomena. In some cases that has meant introducing new concepts, and in some cases that has meant simplifying or combining some of Grice's different concepts.

In this module we won't review everything there is about neo-Gricean pragmatics, but we will take a closer look at one particular neo-Gricean idea that is sort of a little of each. It's a way of simplifying the Gricean maxims into a smaller number of principles, but it also raises an important new point that we didn't make explicit when we looked at Gricean ideas. The summary presented in this module is based on the way these concepts are described by Zufferey et al., chapter 6.4, and Huang, chapter 2.2.

The Goldilocks problem with informativeness

Recall that one of the key parts of Grice's theory is the Cooperative Principle, which he divides into a bunch of conversational maxims (which we have generally lumped into four groups of maxims: maxims pertaining to Quality, Quantity, Manner, and Relation). As we've already seen, sometimes it's hard to draw clear lines between them, and sometimes it's clear that an utterance violates something but it's hard to tell exactly which maxim it violates.

Starting in the 1980s, neo-Gricean pragmaticists tried to clean up this system by reducing it down to a few major principles (which are still sub-parts of the overarching Cooperative Principle) rather than a whole mess of lots of maxims. Some of the most influential theories along these lines were the ones proposed by Laurence Horn and Stephen Levinson, two of the biggest names in neo-Gricean pragmatics and really in pragmatics in general (Levinson literally wrote the book on pragmatics—his textbook, cited often within this class, remains probably the most influential pragmatics textbook ever written—and Horn's research about scalar implicatures was so seminal that the scales these implicatures are based on are now called "Horn scales").

Horn suggested that we should ditch all the Gricean maxims and replace them with just two principles, which he called the Q-principle (short for "quantity") and the R-principle (short for "relation"). Levinson suggested something very similar, but his principles were instead called Q (again for "quantity") and I (for "informativeness"), and he also had a third principle (the M-principle, for "manner", which we won't discuss in this module). Horn's Q- and R-principles are pretty similar to Levinson's Q- and I-principles, but the Levinsonian ones are the ones that have been used more in subsequent pragmatic theory, so those are the ones I will use here.

Both systems have the same basic idea: speakers have to balance two conflicting goals. On the one hand, they want to be as informative as possible (for the purposes of the conversation); if they are trying to be cooperative, they won't want to say too little. On the other hand, they don't want to be over-informative in a way that would be irrelevant and annoying. In other words, if someone asks me what time it is, I don't want to say "It's the 21st century", because that would be not enough information; I also don't want to say "It's 2:57.15" because that would be too much information. I want to say something that's just the right amount of information (for example, right now I might say "It's three o'clock", even though it's actually 2:58 and 37 seconds, because I think "three o'clock" is probably close enough for normal conversational purposes). In other words, I want to find a "Goldilocks" level of informativeness: not too little, not too much, but just right.

The Q-principle and I-principle are meant to represent that. They are sort of like the Gricean maxims, but much broader. The Q-principle says, roughly, that a speaker should say as much information as they can honestly say (as Levinson puts it: "Do not provide a statement that is informationally weaker than your knowledge of the world allows") as long as it's relevant. The I-principle, on the other hand, says, roughly, that speakers should keep the stuff they say informative and relevant, and not say a bunch of irrelevant and unnecessary stuff.

Taken to their extremes, these two principles are incompatible: the Q-principle tells you to say as much as you can and the I-principle tells you to say as little as you can, and it's impossible for you to do both. So, as a speaker, you always have to figure out a way to balance these—you have to figure out what's the "Goldilocks" amount of information you can say to best satisfy both principles. (If you've ever studied phonology and learned about Optimality Theory, this idea should sound similar; just as OT desribes phonology by describing how speakers resolve the conflict between markedness and faithfulness, neo-Gricean pragmatics describes meaning by describing how speakers resolve the conflict between the Q- and I-principles.)

These principles are from the perspective of the speaker (i.e., they describe what a speaker tries to do), but they also lead to corollaries about what a listener can do, as long as the listener assumes that the speaker is following those principles. If a listener assumes that the speaker is following the Q-principle and saying as much as they can, then the listener can assume that anything the speaker didn't say is stuff that the speaker doesn't think is true (or doesn't think is relevant). On the other hand, if a listener assumes the speaker is following the I-principle, then the listener can assume that whatever the speaker says should be "enriched" to make it as relevant and informative as possible (because a speaker following the I-principle would say as little as possible).

Q-implicatures vs. I-implicatures

Based on these ideas, we can see how different sorts of implicatures can arise from the Q-principle and the I-principle.

The Q-principle supposedly explains stuff like scalar implicatures (i.e., if a speaker says "Josh is smart" the listener might infer that the speaker does not believe Josh is brilliant). These implicatures happen because, if a person is following the Q-principle and they also believe Josh is brilliant, they would say "Josh is brilliant"; if they don't say that, then that seems to mean either they don't believe Josh is brilliant or they aren't following the Q-principle (or there is some reason that whether or not Josh is brilliant is irrelevant to the conversation).

The I-principle supposedly explains a wide variety of implicatures, all of which involve somehow enriching the meaning of something the speaker said that was vague. For example, it might explain the conventional implicatures we've seen with and, where e.g. "Sammy and Chris fell in love and got married" may imply that they fell in love first and got married later; that extra bit of information about the order is part of the interpretation that the speaker may have chosen not to say explicitly (because of the I-principle, which enjoins the speaker to say as little as possible) but which the listener might infer. It might also explain a bunch of other sorts of implicatures and related phenomena, such as indirect illocutionary acts (e.g., asking "Can you pass the salt?" is usually interpreted as a request to pass the salt) and negative strengthening (e.g., interpreting "I don't like superhero movies" as meaning the speaker dislikes superhero movies). What these have in common is that they all involve somehow strengthening the utterance (making it more specific) in order to somehow make it more relevant.

Recall that, at the end of the module on negative strengthening, we noted an interesting difference between scalar implicatures and negative-strengthening implicatures: scalar implicatures involve denying a stronger alternative that was not said (e.g., Rebecca didn't say "Josh is brilliant" so we think she doesn't believe it), whereas negative-strengthening implicatures involve accepting a stronger alternative that was not said (e.g., I didn't say "I dislike superhero movies", but you think I dislike superhero movies anyway). At the time we explained that by saying that negative strengthening is based on the Competence Assumption, whereas scalar implicatures (at least the 'weak' versions of them) are based on the Gricean maxim of quantity. Here, though, we can see that the Q- and I-principles can offer a different explanation of the same phenomenon: we could claim that scalar implicatures are based on the Q-principle (i.e., they are "Q-implicatures") whereas negative strengthening is based on the I-principle (i.e., it is an "I-implicature").

Huang notes that Q-implicatures tend to be "negative", in that they usually involve inferring that some other proposition (i.e., something that the speaker could have said but chose not to say) is not true. On the other hand, I-implicatures aren't like that; they involve inferring that some extra idea is true, not that some extra idea is false.

Another interesting difference between Q- and I-implicatures is their different susceptibility to metalinguistic negation. Metalinguistic negation is a way of using negation not to negate some proposition, but to negate an utterance itself. Huang illustrates the difference between metalinguistic negation and normal negation nicely with the following pair:

In normal negation you are saying some proposition is wrong, whereas in metalinguistic negation you are objecting to some utterance, on the grounds of its form or pronunciation or grammar or something like that. It's almost as if you are putting quotes around the part you are negating (as I have done in this example). It also tends to be pronounced with stress (as I have indicated in this example). And it tends to be followed up with some correction; but, unlike normal negation, which can be followed up by something that corrects the wrong proposition, metalinguistic negation tends to be followed up by something that corrects the way of saying it. In other words, in the normal negation example the speaker is saying "It's not true that Zhangsan lives in Peking; here's the place he actually lives", but in the metalinguistic example the speaker is saying something more like "'Zhangsan lives in Peking' is not the right way to say it; you should say it this other way instead.'" In fact, metalinguistic negation seems to require some sort of follow-up; the normal negation example would sound find even if we didn't say "he lives in Shanghai" at the end, but the metalinguistic example only makes sense when it includes the follow-up "correction" at the end.

Importantly, Q-implicatures can be cancelled with metalinguistic negation; for example, we can say "Josh isn't smart, he's brilliant!" (We will see in a later module that this also works on conventional implicatures and on another kind of pragmatic meaning called presuppositions.) I-implicatures, however, do not seem susceptible to metalinguistic negation (according to Huang). For example, saying "Sammy and Chris didn't fall in love and get married" doesn't seem to cancel the implicature that they fell in love first and get married second (in fact, I'm not sure what it does, and I can't think of what the follow-up would be... maybe "Sammy and Chris didn't fall in love and get married, they fell in love but got married!"... but this isn't really cancelling an implicature, it's introducing a new conventional implicature). So there is another interesting bit of evidence that these types of implicatures form two different categories. (However, contra this claim from Huang, it does feel like I can metalinguistically negate that I-implicature if I put stress on the whole verb phrase rather than just the word "and": "Sammy and Chris didn't fall in love and get married, they got married and fell in love!" So I'm not sure this metalinguistic thing is really a reliable difference between Q- and I-implicatures across the board.)

Overall, I don't often use these neo-Gricean concepts and I rarely find myself needing to talk about Q- and I-principles to explain or analyze some implicature (honestly, I had been doing pragmatics research for over 10 years before I even learned what these are); I consider these less important for pragmatics than the Gricean maxims. Nevertheless, even if we don't care about these principles themselves, the neo-Gricean project highlights a few important ideas that any pragmatic theory ultimately needs to be able to handle. Particularly, the idea that implicatures can do different things (i.e., there are implicatures that are based on narrowing down the meaning by rejecting some alternative that wasn't said, based on implicatures that narrow down the meaning by adding extra information to it) and the idea that speakers and listeners have to deal with an irreconcilable conflict over how much to say in a normal conversation. These issues are both quite important for any full understanding of how implicatures work. Research on neo-Gricean pragmatics has also made a major contribution to revealing many of the issues that pragmatic theory needs to be able to explain and outlining important issues in pragmatics; for example, the topics of scalar implicatures and clausal implicatures (and I think maybe also negative strengthening) are all ones that attracted more widespread attention as a result of their treatments in neo-Gricean theories.

In-class activities

Try going back through a bunch of the conversational implicatures we have seen in this class (including in modules and in previous discussions) and try to determine whether they are Q- or I-implicatures.

⟵ Alternatives and context
Relevance Theory ⟶

by Stephen Politzer-Ahles. Last modified on 2022-03-24. CC-BY-4.0.