# 8.6: How to tell one kind of inference from another

$$\newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}}$$

$$\newcommand{\id}{\mathrm{id}}$$ $$\newcommand{\Span}{\mathrm{span}}$$

( \newcommand{\kernel}{\mathrm{null}\,}\) $$\newcommand{\range}{\mathrm{range}\,}$$

$$\newcommand{\RealPart}{\mathrm{Re}}$$ $$\newcommand{\ImaginaryPart}{\mathrm{Im}}$$

$$\newcommand{\Argument}{\mathrm{Arg}}$$ $$\newcommand{\norm}[1]{\| #1 \|}$$

$$\newcommand{\inner}[2]{\langle #1, #2 \rangle}$$

$$\newcommand{\Span}{\mathrm{span}}$$

$$\newcommand{\id}{\mathrm{id}}$$

$$\newcommand{\Span}{\mathrm{span}}$$

$$\newcommand{\kernel}{\mathrm{null}\,}$$

$$\newcommand{\range}{\mathrm{range}\,}$$

$$\newcommand{\RealPart}{\mathrm{Re}}$$

$$\newcommand{\ImaginaryPart}{\mathrm{Im}}$$

$$\newcommand{\Argument}{\mathrm{Arg}}$$

$$\newcommand{\norm}[1]{\| #1 \|}$$

$$\newcommand{\inner}[2]{\langle #1, #2 \rangle}$$

$$\newcommand{\Span}{\mathrm{span}}$$ $$\newcommand{\AA}{\unicode[.8,0]{x212B}}$$

$$\newcommand{\vectorA}[1]{\vec{#1}} % arrow$$

$$\newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow$$

$$\newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vectorC}[1]{\textbf{#1}}$$

$$\newcommand{\vectorD}[1]{\overrightarrow{#1}}$$

$$\newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}}$$

$$\newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}}$$

$$\newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} }$$

$$\newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}}$$

$$\newcommand{\avec}{\mathbf a}$$ $$\newcommand{\bvec}{\mathbf b}$$ $$\newcommand{\cvec}{\mathbf c}$$ $$\newcommand{\dvec}{\mathbf d}$$ $$\newcommand{\dtil}{\widetilde{\mathbf d}}$$ $$\newcommand{\evec}{\mathbf e}$$ $$\newcommand{\fvec}{\mathbf f}$$ $$\newcommand{\nvec}{\mathbf n}$$ $$\newcommand{\pvec}{\mathbf p}$$ $$\newcommand{\qvec}{\mathbf q}$$ $$\newcommand{\svec}{\mathbf s}$$ $$\newcommand{\tvec}{\mathbf t}$$ $$\newcommand{\uvec}{\mathbf u}$$ $$\newcommand{\vvec}{\mathbf v}$$ $$\newcommand{\wvec}{\mathbf w}$$ $$\newcommand{\xvec}{\mathbf x}$$ $$\newcommand{\yvec}{\mathbf y}$$ $$\newcommand{\zvec}{\mathbf z}$$ $$\newcommand{\rvec}{\mathbf r}$$ $$\newcommand{\mvec}{\mathbf m}$$ $$\newcommand{\zerovec}{\mathbf 0}$$ $$\newcommand{\onevec}{\mathbf 1}$$ $$\newcommand{\real}{\mathbb R}$$ $$\newcommand{\twovec}[2]{\left[\begin{array}{r}#1 \\ #2 \end{array}\right]}$$ $$\newcommand{\ctwovec}[2]{\left[\begin{array}{c}#1 \\ #2 \end{array}\right]}$$ $$\newcommand{\threevec}[3]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \end{array}\right]}$$ $$\newcommand{\cthreevec}[3]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \end{array}\right]}$$ $$\newcommand{\fourvec}[4]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}$$ $$\newcommand{\cfourvec}[4]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}$$ $$\newcommand{\fivevec}[5]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}$$ $$\newcommand{\cfivevec}[5]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}$$ $$\newcommand{\mattwo}[4]{\left[\begin{array}{rr}#1 \amp #2 \\ #3 \amp #4 \\ \end{array}\right]}$$ $$\newcommand{\laspan}[1]{\text{Span}\{#1\}}$$ $$\newcommand{\bcal}{\cal B}$$ $$\newcommand{\ccal}{\cal C}$$ $$\newcommand{\scal}{\cal S}$$ $$\newcommand{\wcal}{\cal W}$$ $$\newcommand{\ecal}{\cal E}$$ $$\newcommand{\coords}[2]{\left\{#1\right\}_{#2}}$$ $$\newcommand{\gray}[1]{\color{gray}{#1}}$$ $$\newcommand{\lgray}[1]{\color{lightgray}{#1}}$$ $$\newcommand{\rank}{\operatorname{rank}}$$ $$\newcommand{\row}{\text{Row}}$$ $$\newcommand{\col}{\text{Col}}$$ $$\renewcommand{\row}{\text{Row}}$$ $$\newcommand{\nul}{\text{Nul}}$$ $$\newcommand{\var}{\text{Var}}$$ $$\newcommand{\corr}{\text{corr}}$$ $$\newcommand{\len}[1]{\left|#1\right|}$$ $$\newcommand{\bbar}{\overline{\bvec}}$$ $$\newcommand{\bhat}{\widehat{\bvec}}$$ $$\newcommand{\bperp}{\bvec^\perp}$$ $$\newcommand{\xhat}{\widehat{\xvec}}$$ $$\newcommand{\vhat}{\widehat{\vvec}}$$ $$\newcommand{\uhat}{\widehat{\uvec}}$$ $$\newcommand{\what}{\widehat{\wvec}}$$ $$\newcommand{\Sighat}{\widehat{\Sigma}}$$ $$\newcommand{\lt}{<}$$ $$\newcommand{\gt}{>}$$ $$\newcommand{\amp}{&}$$ $$\definecolor{fillinmathshade}{gray}{0.9}$$

The table in Table 8.1 summarizes some of the characteristic properties of entailments, conversational implicatures, and presuppositions.11 In this section we will work through some examples showing how we can use these properties as diagnostic tools to help us determine which kind of inference we are dealing with in any particular example.

Two general comments need to be kept in mind. First, before we begin applying these tests, it is important to ask whether there is in fact a linguistic inference to be tested. The question is this: if a speaker whom we believe to be truthful and well-informed says p, would this utterance in and of itself give us reason to believe q? If so, we can apply the tests to determine the nature of the inference from p to q. But if not, applying the tests will only cause confusion. For example, if our truthful and well-informed speaker says My bank manager has just been murdered, it seems reasonable to assume that the bank will soon be hiring a new manager.12 However, this expectation is based on our knowledge of how the world works, and not the meaning of the sentence itself; there is no linguistic inference involved. If the bank owners decided to leave the position unfilled, or even to close that branch office entirely, it would not render the speaker’s statement false or misleading.

Second, any one test may give unreliable results in a particular example, because so many complex factors contribute to the meaning of an utterance. For this reason, it is important to use several tests whenever possible, and choose the analysis that best explains the full range of available data. Presuppositions are especially tricky, partly because they are not a uniform class; different sorts seem to behave differently in certain respects. Some specific issues regarding presuppositions are discussed below.

Let us begin with some simple examples. If our truthful and well-informed speaker makes the statement in (22), we would certainly infer that the wasp is dead. We can test to see whether this inference is cancellable/defeasible, as in (22a); the result is a contradiction. We can test to see whether the inference can be suspended, as in (22b); the result is quite unnatural. We can test to see whether the inference is reinforceable, as in (22c); the result is unnaturally redundant.

(22) stated: John killed the wasp.
inferred: The wasp died.

a. # John killed the wasp, but the wasp did not die.

b. # John killed the wasp, but I’m not sure whether the wasp died.

c. ?# John killed the wasp, and the wasp died.

d. Did John kill the wasp?

e. John did not kill the wasp (and the wasp did not die).

In applying the final test, we are asking whether the same inference is created by a family of related sentences, which includes negation and questioning of the original statement. Clearly if someone asks the question in (22d), that would not give us any reason to believe that the wasp died. Similarly, the negative statement in (22e) gives us no reason to believe that the wasp died. We can demonstrate this by showing that it would not be a contradiction to assert, in the same sentence, that the wasp did not die; note the contrast with (22a), which is a contradiction. We have seen that all four tests in this example produce negative results. This pattern matches the profile of entailment; so we conclude that John killed the wasp entails The wasp died.

Now let us apply the tests to Grice’s example (23); specifically we will be testing the inference that arises from B’s reply, There is a garage around the corner. The sentences in (23a–c) show that this inference is defeasible (additional information can block the inference from arising), suspendable, and reinforceable. Neither the question in (23d) nor the negative statement in (23e) would give A any reason to believe that he could buy petrol around the corner. (The phrase any more could be added in (23e) to make the negative statement sound a bit more natural. In applying these tests, it is important to give the test every opportunity to succeed. Since naturalness is an important criterion for success, it is often helpful to adjust the test sentences as needed to make them more natural, provided the key elements of meaning are not lost or distorted.)

(23) A: I am out of petrol.
B: There is a garage around the corner.
inferred: You can buy petrol there.

a. There is a garage around the corner, but they aren’t selling petrol today.

b. There is a garage around the corner, but I’m not sure whether they sell petrol.

c. There is a garage around the corner, and you can buy petrol there.

d. Is there a garage around the corner?

e. There is no garage around the corner (any more).

In this example the first three tests produce positive results, while the last one (the “family of sentences” test) is negative. This pattern matches the profile of conversational implicature; so we conclude that There is a garage around the corner (when spoken in the context of A’s statement) conversationally implicates You can buy petrol there. Of course, we already knew this, based on our previous discussion. What we are doing here is illustrating and validating the tests by showing how they work with relatively simple cases where we think we know the answer. This gives us a basis for expecting that the tests will work for more complex cases as well.

Finally consider the inference shown in (24). The sentences in (24a–c) show that this inference is not defeasible (24a) or reinforceable (24c), but it is suspendable (24b). Both the question in (24d) and the negative statement in (24e) seem to imply that John used to chew betel nut. These results match the profile of a presupposition, as expected (stopped chewing presupposes used to chew).

(24) stated: John has stopped chewing betel nut.
inferred: John used to chew betel nut.

a. # John has stopped chewing betel nut, and in fact he has never chewed it.

b. John has stopped chewing betel nut, if he (ever/really) did chew it.

c. ?# John has stopped chewing betel nut, and he used to chew it.

d. Has John stopped chewing betel nut?

e. John has not stopped chewing betel nut.

Recall that we mentioned in Chapter 3 another test which is useful for identifying presuppositions, the “Hey, wait a minute” test.13 If a speaker’s utterance presupposes something that is not in fact part of the common ground, it is quite appropriate for the hearer to object in the way shown in (25a). However, it is not appropriate for the hearer to object in this way just because the main point of the assertion is not in fact part of the common ground (25b). In fact, it would be unnatural for the speaker to assert something that is already part of the common ground.

(25) statement: John has stopped chewing betel nut.

a. response 1: Hey, wait a minute, I didn’t know that John used to chew betel nut!

b. response 2: # Hey, wait a minute, I didn’t know that John has stopped chewing betel nut!

We mentioned above that it is important to use several tests whenever possible, because any one test may run into unexpected complications in a particular context. For example, our discussion in §4.1 would lead us to believe that the word most should trigger the generalized conversational implicature not all. The examples in (26) are largely consistent with this prediction. They indicate that the inference is defeasible (26a), suspendable (26b), and reinforceable (26c). However, the “family of sentences” tests produce inconsistent results. The question in (26d) fails to trigger the inference, as expected, but the negative statement in (26e) seems to entail (not just implicate) that not all of the boys went to the soccer match.

(26) stated: Most of the boys went to the soccer match.
inferred: Not all of the boys went to the soccer match.

a. Most of the boys went to the soccer match; in fact, I think all of them went.

b. Most of the boys went to the soccer match, if not all of them.

c. Most of the boys went to the soccer match, but not all of them.

d. Did most of the boys go to the soccer match?

e. Most of the boys didn’t go to the soccer match.

f. If most of the boys went to the soccer match, dinner will probably be late this evening.

As mentioned in Chapter 4, combining clausal negation with quantified noun phrases often creates ambiguity; we see here that it can introduce other complexities as well. This is a situation where preservation under negation is not a reliable indicator. However, other members of the “family of sentences”, including the question (26d) and conditional clause (26f), can be used, and show that the inference is not preserved. So the overall pattern of results confirms that this is a conversational implicature.

The table in Table 8.1 indicates that presuppositions are normally preserved under negation, and this is the first (and often the only) test that many people use for identifying presuppositions. But as we have seen, negating a sentence can introduce new complications. In discussing the presupposition in (24) we noted that the negative statement (24e), repeated here as (27a), seems to imply that John used to chew betel nut. This is true if the sentence is read with neutral intonation; but if it is read with what Jespersen (1933) calls “the peculiar intonation indicative of contradiction”, indicated in (27b), it becomes possible to explicitly deny the presupposition without contradiction or anomaly. This is an instance of presupposition-cancelling negation.

(27) a. John hasn’t stopped chewing betel nut.
b. John hasn’t stopped chewing betel nut, he never did chew it.

Horn (1985; 1989) argues that cases of presupposition-cancelling negation like (27b) involve a special kind of negation which he refers to as metalinguistic negation. Metalinguistic negation is typically used to contradict something that the addressee has just said, implied, or implicitly accepted.14 The negated clause is generally spoken with the special intonation pattern mentioned above, and is typically followed by a correction or “rectification” as in (27b).

Some additional examples of metalinguistic negation are presented in (28). These examples show clearly that metalinguistic negation is different from normal, logical negation which is used to deny the truth of a proposition. If the negation used in these examples was simply negating the propositional content, the sentences would be contradictions, because horrible entails bad, all entails most, etc. Horn claims that what is negated in such examples is not the propositional content but the conversational implicature: asserting bad implicates not horrible; asserting most implicates not all. Metalinguistic negation is used to reject the statements in the first clause as being inappropriate or “infelicitous”, because they are not strong enough.

(28) a. That [1983] wasn’t a bad year, it was horrible.15
b. I’m not hungry, I’m starving.
c. Most of the boys didn’t go to the soccer match, all of them went.

For our present purposes what we need to remember is that, in testing to see whether an inference is preserved under negation (one of the “family of sentences” tests), we must be careful to use normal, logical negation rather than metalinguistic negation.

11 Thanks to Seth Johnston for suggesting this type of summary table.

12 This example comes from Saeed (2009: 54).

13 Von Fintel (2004)

14 Karttunen & Peters (1979: 46–47).

15 A quote from the famous baseball player Reggie Jackson, cited in Horn (1989: 382).

This page titled 8.6: How to tell one kind of inference from another is shared under a CC BY 4.0 license and was authored, remixed, and/or curated by Paul Kroeger (Language Library Press) via source content that was edited to the style and standards of the LibreTexts platform.