Weak and strong implicatures

↵ Back to class homepage

Thus far we have usually discussed implicatures by talking about what certain utterances "mean". But, as mentioned in the previous module, that is an oversimplification. We actually need to think about what utterances tell us the speaker believes. I.e., rather than just saying "utterance X means Y...", we should be saying "utterance X tells me that the speaker believes Y...". (Recall that, as we learned in the module about the Cooperative Principle, "meaning" can be defined as just what a speaker wants the hearer to believe and wants the hearer to recognize that the speaker wants the hearer to believe; so, thinking about the speaker's beliefs and intentions is a crucial part of figuring out what anything "means".)

With that in mind, for the next several modules we will take a deeper look about what we can learn about implicatures by considering what the speaker knows and believes. What someone knows or believes is often called their epistemic state. (More formally, an epistemic state is someone's attitude towards whether or not some proposition is true; statements like "I know P", "I'm not sure about P", "I have no idea whether P", "I am certain that P is wrong , etc., are all ways of describing my epistemic state towards P [which could be any proposition, like "there is a wig in this box", "3+3=10", or whatever, depending on the context].)

A sample implicature

In this scene near the end of the video game Horizon: Zero Dawn, the main character, Aloy, is talking to her occasional ally, Nil, who has come to offer his assistance in an upcoming battle for the fate of the world (what role-playing game doesn't have a climactic battle for the fate of the world?). Nil is an evil sociopath who loves fighting and killing, but in this particular moment he's a "lesser of two evils" (because the enemy they're about to face wants to destroy all life on Earth), so Aloy is nevertheless grateful for his offer to help in the important fight that's about to happen. After chiding him for being a creep, she acknowledges that his assistance will be valuable:

Here we see that Nil objects to Aloy's description of the coming battle as "hard", because he thinks it will be more than just "hard"; he thinks it will be "near impossible". In other words, he interprets Aloy's utterance (#1) as having the implicature shown in #2:

  1. Utterance: This battle will be hard.
  2. implicates: This battle will be hard, but not near-impossible.

Since Nil thinks the battle will be near-impossible and Aloy seems to implicate that it won't be, he disagrees with (or corrects or clarifies) her. Importantly, he's not actually disagreeing with what she said, he's disagreeing with what the implicated. What Aloy said, "This battle will be hard", literally just means that its difficulty will be at or above some certain level; it is logically consistent with the possibility that the battle might be near-impossible or even impossible. The idea that the battle will be "hard but not near-impossible" is an implicature, not a literal meaning; we can tell by using the implicature diagnostics we've seen before, such as the cancellability test ("This battle will be hard; in fact, it will be nearly impossible").

This kind of implicature is often called a scalar implicature (we will see this term come up again many times in the coming modules), and it seems to follow straightforwardly from Gricean logic. Per the maxim of quantity, if Aloy believed the battle would be almost impossible, she could have said "The battle will be near impossible"; but she didn't say that, so we infer that she means it's not near-impossible (i.e., she means it will hard but there's still a good chance that victory is possible).

For a very similar example, imagine that some lady Rebecca says "Josh is smart". People may often infer that she means he's just smart but he's not a brilliant super-genius. This implicature follows from the same logic just outlined above: if Josh were brilliant then Rebecca could have said "Josh is brilliant", but she didn't say that, so we infer that there must be a reason she chose not to say that, and then we infer that the reason is she really means he's not brilliant.

But I have oversimplified here. We actually need more steps of reasoning to infer that Rebecca saying Josh is smart means "Josh is smart but not brilliant" (or that Aloy saying This battle will be hard means "This battle will be hard, but not near-impossible").. The problem with the analysis I just described above is that I tried to describe the implicature vaguely as what the utterance "means", rather than putting "meaning" in terms of Aloy's and Rebecca's beliefs and intentions. (Recall that, as we learned in the first module about Gricean reasoning, "meaning" in Gricean theory is all about what the speaker believes and wants to make us know that they believe.)

Let's try analyzing the "Josh is smart" example again, and be more careful about delineating exactly what Rebecca (the speaker) believes or intends at each step of the process. Following the description outlined by Geurts (chapter 2), we can use the following sort of reasoning to interpret (1):

To understand the distinction here, it's necessary to keep in mind that "X does not believe Y" is not the same as "X believes not-Y". As I mentioned in step 2 above, "Rebecca does not believe Josh is brilliant" literally means that "Rebecca believes Josh is brilliant" is false: it could be false because Rebecca really believes Josh is not brilliant, or it could be false because Rebecca doesn't have any beliefs or opinions about this one way or another (maybe she doesn't even know Josh!).

Some helpful notation

I find this distinction clearer to express if we use some semantics notation instead of using plain English, because plain English is too ambiguous. Geurts expresses these two different meanings with the formulae ¬BELR(P) and BELR(¬P). I know semantics notation may look a bit intimidating, but let's take a moment to break these down, because I think they will be useful.

P is a placeholder for some proposition; in this example, we are using P to refer to the proposition that Josh is brilliant. BEL is short for "believes", and the subscript R is short for Rebecca. So BELR(P), in our current example, means "Rebecca believes that P", i.e., "Rebecca believes that Josh is brilliant." The most crucial piece left is the ¬ sign, which means "not". So we end up with two different formulae (representing two different meanings), depending on where we put the "not":

Now let's fit these back into the four-step reasoning procedure we outlined above:

Through those steps we are left with BELR(¬P), i.e., Rebecca believes that Josh is not brilliant.

Weak and strong implicatures

The important point here is that, if we hear Rebecca say "Josh is smart" and then we infer that she means he's not brilliant, then we have actually inferred two implicatures. In step 2, we infer that Rebecca does not believe Josh is brilliant (¬BELR(P)); we do this based on the Gricean maxim of quantity (we noticed that she could have said something more informative, but she chose not to, so apparently she flouted that maxim). Then, in step 4, we infer that Rebecca believes Josh is not brilliant (BELR(¬P)); we do this based on assuming that Rebecca has an opinion about whether or not Josh is brilliant.

Many people refer to the first implicature here (¬BELR(P)) as a weak implicature and to the second (BELR(¬P)) as a strong implicature (see, e.g., Geurts, chapter 2). This distinction is very important for pragmatics, but often overlooked; a lot of research and thinking about implicatures just looks at the strong implicature (just like we did with examples #1 and #2 at the beginning of this module) and neglects the weak implicature. (My own research on pragmatics has also been guilty of this.) But they are different, and are realized through different processes. For example, if we don't have reason to believe that Rebecca has any opinion on Josh's brilliance (for example, maybe she's only just met him, and has seem him do some simple tasks but has never yet had the opportunity to see him do anything that would require brilliance), then we might not get the strong implicature at all.

According to some pragmaticists, like Geurts, the ability to explain both strong and weak implicatures is one of the key advantages of the Gricean theory as opposed to others. You can decide for yourself whether or not you agree (maybe you can think of some other way to explain where these two implicatures come from), but you should always keep in mind that any theory of pragmatics will need to explain these two different kinds of implicatures (unless it turns out that these don't really exist or aren't really different).

(Note that here I am using the terms "weak implicature" and "strong implicature" in the sense that Geurts and some other Gricean pragmaticists use them, and not in the sense that some other theories do. For instance, Relevance Theory [another approach to pragmatics which is quite different from the Gricean approach that we are mostly focusing on in this class] also has concepts of "weak implicature" and "strong implicature", but those concepts mean something completely different in Relevance Theory than what they mean in this module. In Relevance Theory, a "strong implicature" is an implicature that you need to recover in order to figure out what message the speaker intends to communicate to you, and a "weak implicature" is some extra meaning that isn't totally necessary for the purpose of what the speaker is trying to communicate—this is similar to Noveck's distinction between "voluntary implicatures" and "imposed implicatures" that we discussed in the previous module. So just keep in mind that you might someday see the words "weak implicature" and "strong implicature" used to refer to a very different concept than what this module was about.

Video summary

In-class activities

Have students brainstorm at least one more example of an utterance that has both strong and weak implicatures, and think of a context where the strong implicature would arise and one where the strong implicature would not arise.

Gradient adjectives, like the one we focused on in this module, are an easy place to find strong and weak implicatures. Others are quantifiers like "some" modals like "might" ("some of the Xs" is often taken to imply "some but not all of the Xs", and "she might Y" is often taken to imply "she might Y but it's not certain"; these are both classical examples of "scalar implicatures" which will be discussed in more depth in later modules), and so-called "ad-hoc scales" (e.g., if I ask what classes your friend is taking this semester and you say "engineering", I might assume your friend is only taking engineering and nothing else).

I've sometimes wondered whether strong implicatures are even the same kind of implicature as the others we've seen.

This will ultimately probably depend on how we define "implicature". Certainly strong implicatures are something that is beyond "what is said", and the way they are derived is based on Gricean reasoning about the speaker's beliefs and intentions (recall the bit about the "what" intention and the "that" intention that we learned about in the module about the Cooperative Principle). So, if we use the term "implicature" to mean anything non-truth-conditional meaning that is based on Gricean reasoning about intentions, then strong implicatures seem to count. However, they are not derived from the Cooperative Principle; they're derived from assuming the speaker is competent (i.e., either BELR(P) or BELR(¬P); the speaker has an opinion one way or the other), which is different than the Cooperative Principle. (Although maybe someone could argue that this competency assumption is based on the Cooperative Principle, on the assumption that a cooperative speaker would keep their mouth shut if they have no opinion? This seems debateable to me.) Zufferey et al. (chapter 6.7) also recognize that the assumption of speaker competence (which they call opinionatedness, a clunkier but more accurate term) is "non-Gricean". So, even though strong implicatures seem to be implicatures, they seem to be derived in a fundamentally different way than other conventional and conversational implicatures we have seen.

A key distinction here is between different types of "implicature". We have already seen that not all implicatures are conversational implicatures; there are also conventional implicatures. But those might not be the only types of implicatures. Indeed, Levinson (chapter 3.1) divides implicature into "conventional" and "non-conventional", then divides non-conventional implicature into "non-conversational" and "conversational" (and further divides conversational implicature into generalized and particularized), as shown in the chart below (based on Levinson, p. 131). So this leaves a type of implicature that we don't seem to have seen yet: implicatures that are neither conventional nor conversational (the red circle in the image below). Levinson doesn't say anything about what these implicatures might be (and, interestingly, many other pragmatics texts, such as Zufferey et al chapter 6 figure 3, have charts very similar to this one, just with this one piece removed), but I wonder if "strong implicatures" fall into this group.

A tree-style representation of meaning; see description in the main text above.

Do any of the implicature diagnostics help us figure out if strong implicatures are conversational implicatures? Or are there any other tests or characteristics you can think of that distinguish between strong implicatures and implicatures based on the Cooperative Principle?

⟵ One utterance, multiple implicatures
Negative strengthening ⟶

by Stephen Politzer-Ahles. Last modified on 2022-04-15. CC-BY-4.0.