1. Introduction
Can one be courageous in pursuit of evil ends? James Rachels illustrates the question like this:
Consider a Nazi soldier […] who fights valiantly – he faces great risk without flinching – but he does so in an evil cause. Is he courageous? (Reference Rachels1999, 179)
Many people are reluctant to answer ‘Yes’ to Rachels’ question. But many are also reluctant to answer ‘No’. The question thereby illustrates a two-part challenge, which shall be called the Courage Problem. The first task is to commit to a view on whether gutsy villains are courageous.Footnote 1 The second task is to explain away people’s reluctance to respond as we have to the first task.
One approach to the Courage Problem has potentially unwanted consequences. Suppose we accept that the Nazi is courageous. If we also hold that he has no moral virtues, then we’re forced to accept a radical conclusion – that courage is not a moral virtue. Of course, this radical conclusion could be avoided if we insist that the Nazi has at least one moral virtue, namely courage. But this forces us to reject a classical position, that the virtues are unified – having one moral virtue requires having them all (Aristotle Reference Ostwald1962 NE 1145a1–2). Even if the Nazi has courage, he surely lacks other moral virtues, such as justice.Footnote 2 Therefore, if the Nazi is courageous, then either courage is not a moral virtue or the virtues are not unified. Voltaire infers the first conclusion from his view that gutsy villains can be courageous (1905, 223); James Wallace infers the second from the same view (Reference Wallace1978, 77).
However, if this paper is successful then these conclusions should not be based on the view that gutsy villains are courageous. This paper provides reason to think gutsy villains are not courageous, and it does so independently of whether courage is a virtue or whether the virtues are unified.
The Courage Problem shall be examined in light of a controversy sparked by Bill Maher, who claimed that the 9/11 terrorist attacks were ‘not cowardly.’ Maher’s comment generates another explanatory task – that of explaining why people were offended by it. This task shall be called Maher’s Puzzle. In Section 3, we see that Maher’s Puzzle is fundamentally related to the Courage Problem; it is thus assumed that if a given explanation can solve both Maher’s Puzzle and the Courage Problem then this explanation is better than ones that cannot solve both (other things being equal). In Sections 4 and 5, it’s argued that the view that gutsy villains are courageous does not provide a feasible solution to both problems. Then, Section 6 outlines the preferred solution, one that denies the courage of gutsy villains. This solution integrally involves a traditional view of courage – that courage requires the pursuit of good ends. In Section 7, objections are addressed.
The main argument of this paper is an inference to the best explanation of the linguistic behaviors associated with the Courage Problem and Maher’s Puzzle. The goal is to explain why speakers are reluctant to assert or deny ‘courage’ of gutsy villains, and why hearers were offended by Maher’s use of ‘not cowardly’. Widely accepted linguistic mechanisms are employed to explain these behaviors. And it’s assumed that the best explanation of such behaviors can supply prima facie support for a view about the correct use of ‘courage’ and, in turn, for a view about what courage requires.Footnote 3
2. Courage problem
The Courage Problem typically centers on a person who faces great risk for an end that is morally bad. Paradigmatic examples include Nazi soldiers and terrorists. But the problem can also arise when the relevant end is neutral. Peter Geach describes ‘a fundamentalist’ who maintains ‘to the death against an infidel government that π is equal to 3’ (Reference Geach1977, 159–160). His end – the defense of a mathematical principle – can be seen as neutral, neither good nor bad. Henceforth, let ‘villain’ refer to any gutsy person with morally bad ends, and let ‘fundamentalist’ refer to any gutsy person with neutral ends. And let’s stipulate that these characters are disposed to face risk for the sake of some such end.
The Courage Problem is partly concerned with whether such persons are courageous, but it can also arise with regard to actions. For example, consider a risky action of the Nazi soldier (e.g. running across a battlefield) which is aimed at an evil cause. The Courage Problem is also concerned with whether these actions should count as courageous.Footnote 4
Henceforth, let ‘A’ refer to a villain, fundamentalist, or risky action aimed at a bad or neutral end. Which of the following is true?
-
(1) A is courageous.
-
(2) A is not courageous.
Many people express reluctance to assert either claim. Rachels sums up this reluctance as follows:
Calling the Nazi soldier “courageous” seems to praise his performance, and we should not want to praise it […]. Yet neither does it seem quite right to say that he is not courageous – after all, look at how he behaves in the face of danger. (Reference Rachels1999, 179–80)
Despite this reluctance, one may eventually take a stance. Some people – such as Rachels (Reference Rachels1999, 180) and Foot (Reference Foot2002, 15–7) – assert the truth of (1). Let’s call them assenters. Others – such as Geach (Reference Geach1977, 160) and Walton (Reference Walton1986, 99) – assert that (2) is true. Let’s call them dissenters. The first task of the Courage Problem is to commit to being either an assenter or a dissenter.
Our second and more formidable task is to explain away people’s reluctance to respond as we have to the first task. Many explanations could be offered. Assenters might claim that (1) conversationally implies a proposition that many people reject, and that these people are reluctant to accept (1) because they’re misled by this conversational inference. Alternatively, one could posit multiple senses of the word ‘courageous’ and hold that (1) and (2) are each false under different senses, which may explain why people are reluctant to assert them. These explanations are discussed in Sections 4 and 5.
Ultimately, however, this paper argues that dissenters are correct. Section 6 explains why people are reluctant to assert (2), and it does so without positing extra senses of ‘courageous’. This explanation gains additional support from the fact that it also solves Maher’s Puzzle.
3. Maher’s puzzle
Shortly after the September 11 attacks, Bill Maher publicly disagreed with U.S. President George W. Bush, who called the terrorists ‘cowards’ in a public address (Jones Reference Jones2010, 73). On his show Politically Incorrect, Maher said:
We have been the cowards, lobbing cruise missiles from 2,000 miles away. That’s cowardly. Staying in the airplane when it hits the building, say what you want about it, it’s not cowardly.Footnote 5
Many people were offended by Maher’s statement. Indeed, his statement elicited censure from White House Press Secretary Ari Fleischer. And eventually the comment got him fired (Jones Reference Jones2010, 8, 73). But why did Maher’s comment offend people? His first two sentences cannot be wholly responsible for the fact that people took offense. After all, the criticism of US forces as cowardly hardly causes uproar; US drone strikes are often called ‘cowardly’ without raising similar controversy.Footnote 6 Instead, it’s clear that Maher’s third sentence – saying that the terrorists’ acts were ‘not cowardly’ – is at least partly responsible for the fact that people were offended. Had Maher refrained from uttering it, he would have caused less of an uproar.
Maher’s Puzzle challenges us to explain why people were offended by Maher’s third sentence. This puzzle is distinct from the Courage Problem. For one thing, Maher’s Puzzle does not challenge us to explain a speaker’s reluctance to make an assertion, but only the offense taken by those who heard it.Footnote 7 For another, Maher never used the word ‘courageous’ or its synonyms in describing the terrorists’ actions; indeed, his assertion doesn’t even entail that the terrorist’s acts are courageous. It’s certainly possible for an act to be neither cowardly nor courageous (e.g. brushing one’s teeth).
Still, Maher’s Puzzle is fundamentally related to the Courage Problem in two important ways. (i) Maher sparked the controversy by implying what the Courage Problem takes as counterintuitive – that a villain can be courageous.Footnote 8 In particular, the people who reacted to Maher’s assertion understood him as implying that the terrorists’ acts were courageous, even though he never said this explicitly.Footnote 9 And, according to Rachels, this implication was what sparked the controversy: ‘Maher implied that they were [courageous], and consequently his television show […] got canceled’ (Reference Rachels and Rachels2015, 163). Moreover, (ii) these two problems jointly reveal that it is controversial to deny courage of villains and can also be controversial to deny them the opposite trait – cowardice. It’s unlikely that this commonality is mere coincidence, given that ‘courageous’ and ‘cowardly’ are antonyms that pick out logical contraries.
Given the connections outlined in (i) and (ii), it would be surprising if there were no common explanation underlying our two problems. So, it will be assumed that if a given explanation can solve both Maher’s Puzzle and the Courage Problem then this explanation is better than ones that leave one of our two problems unresolved (other things being equal). The former sort of explanation would have more explanatory power because it would explain the linguistic behaviors associated with both problems rather than only one. In Section 4, it’s shown that many potential solutions to the Courage Problem are deficient along this line – they’re useless in solving Maher’s Puzzle.
This paper takes no stand on whether Maher’s third sentence was true or false. And it’s important to note that the offense taken by it cannot be explained merely by citing what people believed about its truth or falsity. Some people believed the assertion was false. But this doesn’t by itself explain why such people would take offense. People are often not offended by assertions they believe to be false – e.g. had Maher asserted that 2 + 2 = 5, no one would have been offended. Others believed the utterance was true; and some likely suspended judgment. But clearly these doxastic states don’t explain why such people would take offense. So, in order to explain the offense, we cannot focus merely on the believed-truth-value of Maher’s utterance.
A better explanation might claim that Maher’s audience saw his assertion as praising the terrorists or their ends. But this explanation falls short in two ways. First, it’s not immediately clear how the mere denial of a negative attribute (i.e. cowardice) can amount to praise. Moreover, this explanation makes no mention of courage, and therefore fails to explain why the offended parties took issue with the idea that the terrorists are courageous. The final solution in Section 6 avoids both concerns.
Let’s now examine how assenters may solve our two problems. It turns out that their most feasible solutions to the Courage Problem are useless in solving Maher’s Puzzle.
4. Assenters
Philippa Foot is an assenter who attempts to explain our reluctance to assert (1). Her explanation shall be developed into a general model for how assenters may solve the Courage Problem.
Foot’s solution to the Courage Problem hinges on an analogy between courage and poisons. Just as poisons may not always operate characteristically (e.g. if they don’t cause illness), so too might courage fail to operate characteristically (e.g. if it doesn’t produce good action). Foot uses this analogy to explain our reluctance to assert (1). She writes, ‘the fact that courage does not here have its characteristic operation is a reason for finding the description strange’ (Reference Foot2002, 16).
But the analogy between courage and poison is not enough to explain our reluctance to assert (1). If a person remains healthy after ingesting hemlock, we would have no reluctance to say that she ingested poison. So, the analogy needs to be supplemented. For us to explain the reluctance to assert (1), we must also assume that courage-ascriptions typically imply that the courage has its characteristic operation of producing good action. Since this implied claim is false with regard to villains, its falsity could explain why we’re reluctant to assert (1).
On Foot’s view, the implied evaluative claim is one that positively evaluates the same item to which courage is ascribed. That is, utterances of ‘x is courageous’ imply the following:
E: that x is a good action (or person).
In Section 4.2, this claim shall be reformulated as a general principle, but for now let’s assume that E is the appropriate formulation.
How exactly do courage-ascriptions imply E? Assenters have not clarified what this relation is. But let’s consider the options, beginning with H.P. Grice’s distinction between what is said and what is implicated (Reference Grice1989, 24–25). What is said by an utterance includes the truth-conditional content of the sentence uttered. Obviously, assenters cannot claim that E is part of, or entailed by, what is said by courage-ascriptions, because then (1) could not be true while E is false. But they could claim that E is implicated by courage-ascriptions. For Grice, what is implicated includes conventional and conversational implicatures, which are not part of what is said, and are not entailed by what is said. It will now be argued that neither approach helps assenters solve Maher’s Puzzle.
4.1. Troubles with Maher’s puzzle
How might assenters solve Maher’s Puzzle? Ideally, they would claim that Maher’s utterance somehow implied that the terrorists’ acts were courageous, and that this in turn implicated E. And people were offended because they took issue with E. This would be a unified solution to both problems. The trouble is that the assenter’s chosen relation between courage-ascriptions and E cannot be explained by Grice’s notion of implicature without leaving Maher’s Puzzle unsolved. Let’s begin with conventional implicature.
Conventional implicatures are components of meaning that do not contribute to the truth-conditions of the sentence uttered, but which are nonetheless regularly associated with particular linguistic expressions (e.g. ‘but’ and ‘therefore’). Compare ‘Paul is poor but honest’ with ‘Paul is poor and honest’. These sentences plausibly have the same truth-conditions, but only the first sentence implicates that honesty contrasts with poverty. This implication is thought to be a conventional implicature, which is triggered by the word ‘but’ in the first sentence.
Could assenters claim that E is a conventional implicature of courage-ascriptions? If they did, then this solution to the Courage Problem would be useless in solving Maher’s Puzzle. The reason is that conventional implicatures are always detachable (Blome-Tillmann Reference Blome-Tillmann2013, 173). A proposition P is a detachable implication of sentence S only if there could be another sentence propositionally equivalent to S that does not imply P. In this case, P must only be an implication of sentence S, not of the proposition expressed by S. Recall the above example: ‘Paul is poor but honest’ implies that honesty contrasts with poverty; but the propositionally equivalent sentence ‘Paul is poor and honest’ does not have this implication. So this implication must only be an implication of the first sentence, not of the proposition expressed by it. Now consider the assenter’s approach. If E were a conventional implicature of ‘A is courageous’, then E would only be implied by the sentence ‘A is courageous’, rather than the proposition expressed by it. But Maher didn’t utter any such sentence, which means E would be useless in solving Maher’s Puzzle.
Now consider conversational implicature. Conversational implicatures are propositions that hearer’s are defeasibly licensed to infer from the saying of what is said, based on the assumption that certain conversational maxims are being observed (Grice Reference Grice1989, 26–31). For example, typical utterances of ‘Some students attended’ have the conversational implicature that not all students attended. We’re inclined to infer the latter because we’re assuming the speaker is observing certain principles of conversation (e.g. she would not make such a weak claim if she thought all students attended). And the inference passes all tests for conversational implicature, two of which shall be relevant for our purposes. First, the implicature is cancelable (or defeasible) – the speaker could felicitously deny or disown the implicature after her utterance (e.g. ‘Some students attended; in fact they all did’). It’s also reinforceable – the speaker could affirm the implication after her utterance without awkward redundancy (e.g. ‘Some students attended, but not all’).Footnote 10 Cancelability and reinforceability are common tests for conversational implicature, which shall be relevant later.
For now, let’s observe that if E is a conversational implicature of courage-ascriptions (as Grice understood conversational implicature), then this view would be useless in solving Maher’s Puzzle. On Grice’s view, conversational implicatures must be inferable from the saying of what is said (Grice Reference Grice1989, 39). But, again, Maher never said the terrorists’ acts were courageous. So, this approach provides no means for assenters to solve Maher’s Puzzle.
Many other connections between courage-ascriptions and E would be equally useless in solving Maher’s Puzzle, and for the same reason just mentioned. Consider Kent Bach’s conversational impliciture (Reference Bach1994), François Recanati’s pragmatically-enriched-said (Reference Recanati2004, chap. 2), and Sperber and Wilson’s explicature (Reference Sperber and Wilson1986/1995, 182). These views have subtle differences, but here’s the basic idea. Each view holds that the literal meanings of certain declarative sentences do not correspond to complete propositions, because these meanings leave open certain gaps that must be filled before the meanings can yield a proposition. But, according to these theorists, hearers can fill in these gaps by observing the context of utterance, and thereby inferring a complete proposition. For example, the literal meaning of ‘John was late’ doesn’t specify what John was late for, and this may need to be specified for hearers to attach a complete proposition to the sentence uttered. But, with the help of context, hearers can specify this and thereby infer a complete proposition. For instance, from an utterance of ‘John was late’, a hearer could infer that John was late for dinner, in certain contexts.
But these mechanisms don’t help assenters solve Maher’s Puzzle. The inferences in question always require the hearer to infer something from the speaker’s utterance. So, even if E is conveyed by utterances of ‘A is courageous’ via one of these mechanisms, there would still be no reason to think Maher’s utterance conveyed E, because Maher did not utter any such sentence.
In short, if assenters solve the Courage Problem by way of conversational impliciture, pragmatically-enriched-said, explicature, conventional implicature, or conversational implicature (as Grice understood it), then this solution would be useless in solving Maher’s Puzzle.Footnote 11 This leaves assenters with few ways of providing a unified solution to both problems. We shall now see that the remaining options don’t supply adequate solutions to the Courage Problem.
4.2. Troubles with the courage problem
Above, we saw that Grice’s notion of conversational implicature would be useless to assenters in solving Maher’s Puzzle. But assenters could diverge from Grice and hold that conversational implicatures need not be inferred from what is said. Here’s one possibility: let’s suppose the following proposition is a conversational implicature of Maher’s utterance:
P: that staying in the airplane when it hits the building is courageous.
(It will be argued in Section 6 that P is a conversational implicature of Maher’s utterance). In this case, assenters could claim that E is a conversational implicature of proposition P. Here, assenters would be assuming that conversational implicatures can be inferred from other conversational implicatures (which are unsaid). This assumption diverges from the dominant view, according to which conversational implicatures are only inferable from utterances, not propositions, and from the saying of what is said, not what is implicated. But some linguists accept this assumption (e.g. Burton-Roberts Reference Burton-Roberts1984, 194).
The trouble is that it’s implausible for assenters to appeal to conversational implicature, regardless of Maher’s Puzzle. It’s widely agreed that thick evaluative terms, like ‘courageous’, somehow convey evaluations such as E. But (as pointed out in Väyrynen Reference Väyrynen2013, 104–105 and Kyle Reference Kyle2013, 3), the connections between thick terms and evaluations are neither reinforceable nor cancelable. This is also true in the present case. The attempt to cancel E in (3), and to reinforce E in (4), both seem awkward:
-
(3) # It was courageous, but I don’t mean to imply it was good.
-
(4) # It was courageous, and it was good.
(Henceforth let ‘#’ indicate that the sentence is linguistically awkward). Cancelability and reinforceability are seen as the most reliable tests for conversational implicature. Since the relation between courage-ascriptions and E fails these tests, we should look elsewhere for a solution to the Courage Problem.
Consider another linguistic relation – pragmatic presupposition. If an utterance of sentence S pragmatically presupposes a proposition, then this proposition is something we would normally expect to be in the common ground whenever S is uttered (Stalnaker Reference Stalnaker1999, 49). The common ground is the mass of knowledge, beliefs, and assumptions that conversational participants believe they share with one another. For example, an utterance of ‘Smith regrets drinking Pabst’ presupposes that Smith drank Pabst; this proposition is expected to be part of the common ground whenever the sentence is uttered.
Could assenters hold that ‘A is courageous’ pragmatically presupposes E?Footnote 12 No. This putative presupposition lacks the most distinctive feature of presupposition – ’constancy under negation’ (Huang Reference Huang2007, 67). The negation ‘Smith does not regret drinking Pabst’ also presupposes that Smith drank Pabst. But it’s implausible to hold that the negation ‘A is not courageous’ presupposes E. Recall that E is a claim that positively evaluates A. But utterances of ‘The Nazi is not courageous’ surely don’t presuppose a positive evaluation of the Nazi.
The above problem only arises because E positively evaluates the very same item to which ‘courageous’ applies. This was Foot’s view. But perhaps the assenter’s appeal to presupposition would be improved if we replace E with a general claim like the following:
E*: that people who are disposed to confront risk are good.
It’s slightly more plausible that E* is presupposed by typical utterances of ‘A is courageous’. And this approach may help assenters solve the Courage Problem. Assenters believe (1) is true, and they could claim that we’re reluctant to assert (1) because (1) presupposes E*, and the combination of E* with (1) entails that the villain or fundamentalist is good. It is because we deny the goodness of these characters that we’re reluctant to utter (1).
The trouble is that this approach mistakenly predicts that those who reflect on the Courage Problem will become reluctant to apply ‘courage’ to anyone, not merely to the villain or fundamentalist. On this view, we’re reluctant to assert (1) because we recognize that the villain or fundamentalist is a counterexample to what (1) presupposes – E*. So, upon reflecting on the courage-problem, we see that E* is false. But E* is also presupposed by courage-ascriptions in many other contexts, such as when we ascribe courage to Martin Luther King Jr. So, now that we recognize E* as false, we should be reluctant to ascribe courage, not just to villains, but also to paradigmatically courageous people. But this plainly does not happen to those who reflect on the Courage Problem. So, either we’re all being inconsistent by continuing to use the word ‘courageous’ without reluctance, or the putative presupposition does not explain our reluctance to utter (1). The latter is obviously most plausible.
Notice that this problem has nothing to do with the particular relation we appealed to – presupposition. The problem would arise for any relation, provided a general claim like E* is the implied evaluative claim. On the other hand, if we hold that courage-ascriptions imply a specific evaluative claim, like E, then we run into the problems discussed earlier. It appears there is no suitable relation that allows assenters to solve both problems.
5. Physical and moral courage
Fortunately, there’s an approach that does not posit pragmatic relations between ‘courage’ and E or E*. The Courage Problem is often addressed by distinguishing between physical and moral courage (e.g. Toner Reference Toner2000, 114). Physical courage requires risk of life and limb but does not require the agent to have good ends. Moral courage is just the opposite – it requires the agent to have good ends but does not require risk of life and limb. Given this distinction, one may hold that the villain and fundamentalist have physical courage, but lack moral courage, because they don’t have good ends. Does this help solve the Courage Problem?
It depends. On this view, ‘courageous’ could be either ambiguous or univocal. In other words, physical and moral courage could mark out two different senses of the word ‘courageous’. Or, they could be two different kinds of courage, where ‘courage’ is used in the same sense in reference to both. In what follows, we shall see that the ambiguity view violates Grice’s Razor by positing senses beyond necessity, whereas the univocal view provides no solution to the Courage Problem.
Let’s begin with the ambiguity view. How would this view solve the Courage Problem? If ‘courageous’ has more than one sense, then it’s possible to claim that there’s a sense in which (1) is false and also a sense in which (2) is false. (1) is false when ‘courageous’ is used in the moral sense; (2) is false when ‘courageous’ is used in the physical sense. It’s because both claims are false under some reading that we’re reluctant to assert them.
The basic problem with the ambiguity view is that there’s no reason to posit an extra sense of ‘courageous’, except for the fact that it may solve the Courage Problem. In particular, the putative ambiguity of ‘courageous’ cannot be detected by any tests for ambiguity.Footnote 13
Consider the conjunction-reduction test. On this test, we take two sentences that use a term in its supposedly different senses, and then conjoin the sentences in a way that uses the term only once. If the term is ambiguous, we can expect the conjoined sentence to be awkward (Zwicky Reference Zwicky1975, 17–18). For example, ‘call’ is ambiguous. It can mean ‘request’ as in ‘Sue called him a taxi’, but it can also mean ‘label’ as in ‘Sue called him a bigot.’ If we conjoin these sentences while using ‘call’ only once, the result is awkward:
-
(5) # Sue called him a bigot and a taxi.
The awkwardness of (5) suggests that ‘call’ is ambiguous. But now consider ‘courageous’. Proponents of the ambiguity view hold that ‘The Nazi soldier is courageous’ is true in the physical sense. And they would accept that ‘MLK Jr is courageous’ is true in the moral sense. But the conjoined sentence in (6) does not seem awkward:
-
(6) The Nazi soldier and MLK Jr were both courageous.
So, the conjunction-reduction test cannot detect the supposed ambiguity of ‘courageous’.
The contradiction test also fails to detect any ambiguity. According to W.V.O. Quine, ‘an ambiguous term […] may be at once clearly true of various objects […] and clearly false of them’ (Reference Quine1960, 129). Here, we try to observe a lack of contradiction in sentences that would be contradictory if they involved no ambiguous terms. Consider
-
(7) The minister married Sue, and he didn’t marry Sue.
(7) can be interpreted as non-contradictory; the first conjunct can mean that the minister performed the marriage ceremony, whereas the second can mean that he didn’t get married to Sue. By contrast, there appears to be no established reading of (8) that would be non-contradictory:
-
(8) # The Nazi soldier is courageous, and he’s not courageous.
Thus, the contradiction test also fails to detect any ambiguity in ‘courageous’.
The above tests provide no reason to posit an extra sense of ‘courageous’. Can we instead posit an extra sense solely to solve the Courage Problem? No, doing so would violate Grice’s Razor: ‘Senses are not to be multiplied beyond necessity’ (Reference Grice1989, 47). For our purposes, we can assume a minimal version of Grice’s Razor: other things being equal, we should not posit an extra sense of a word if we can explain the same data by citing forms of conversational implicature that we already have within our linguistic theory.Footnote 14 The next section solves both of our problems, without positing an extra sense of ‘courageous’, and it does so by citing widely-accepted forms of conversational implicature. If it’s successful, then we cannot accept the ambiguity view.
What about the univocal view? Simply put, the univocal view provides no means for solving the Courage Problem. If ‘courageous’ has only one sense that applies equally to physical and moral courage, and the villain and fundamentalist each have physical courage, then there is no sense in which these characters lack courage. So, there’s no sense in which (1) is false, and no reason why we would be reluctant to utter it. The univocal view has no solution to offer.
6. Solution
Unlike assenters, let’s suppose courage-ascriptions semantically entail evaluations.Footnote 15 Unlike the ambiguity view, let’s not posit an extra sense of ‘courageous’. And unlike the univocal view, let’s suppose ‘courageous’ expresses only moral courage. Together, these assumptions generate the following postulate:
Semantic Postulate: ‘x is courageous’ semantically entails that x pursues an end that is good.
More precisely, if x is an action, then x must be intended to produce some end e such that e would be good if e were to come about. If x is a person, then x must be disposed to have such intentions.Footnote 16
The Semantic Postulate is nothing new. Hints of it are found in Aristotle (Reference Ostwald1962 NE, 1115b20) and Aquinas (Reference Aquinas1922, ST, II.II Q123 A.1). What’s new is the following argument in its favor. It shall be argued that the Semantic Postulate is a necessary part of the best explanation of the linguistic behaviors associated with the Courage Problem and Maher’s Puzzle. However, this argument only provides prima facie reason to accept the Semantic Postulate. Objections shall be addressed in Section 7.
Let’s start with the Courage Problem. Since A’s ends are not good, the Semantic Postulate entails that dissenters are correct – (1) is false and (2) is true.Footnote 17 According to this postulate, our reluctance to assert (1) is due to our semantic competence with ‘courageous’. What then explains our reluctance to assert (2), if this claim is indeed true?
Our reluctance to assert (2) can be explained by a particular form of conversational implicature. Notice that it’s tempting to infer the following proposition from (2):
Q: that A is cowardly.
This tendency is even more noticeable when we consider sentences that are truth-conditionally equivalent to (2), such as the following:
(2′) A has no courage.
(2″) A lacks courage.
(2), (2′), and (2″) each seem to imply Q in some sense. The trouble is that these inferences are not logically valid. Many actions are neither courageous nor cowardly (e.g. feeding a kitten). And there are persons who are neither courageous nor cowardly (e.g. foolhardy persons).
Nevertheless, the temptation to infer Q from (2), (2′), or (2″) can be explained as a form of conversational implicature known as negative strengthening. When a negation is combined with certain words, like ‘happy’ and ‘good’, this combination conversationally implicates something stronger than what is literally said (Horn Reference Horn1989, 331ff; Levinson Reference Levinson2000, 127ff). For example, an utterance of
-
(9) Sue is not happy.
seems to imply something stronger, namely
R: that Sue is unhappy.
Clearly, (9) doesn’t entail R, since (9) allows that Sue could be so-so rather than unhappy. Still, utterances of (9) seem to imply R in some sense. And linguists typically hold that this occurs via a form of conversational implicature, known as negative strengthening (Horn Reference Horn1989; 331ff; Levinson Reference Levinson2000, 127ff). Notice the inference is cancelable – ’Sue isn’t happy, but she’s not unhappy either.’ It’s also reinforceable – ’Sue isn’t happy, in fact, she’s downright unhappy.’ As we’ll see, negative-strengthening is crucial for solving our two problems.
The inference from ‘not courageous’ to ‘cowardly’ can also be explained by negative strengthening. This inference closely resembles the inference from ‘not happy’ to ‘unhappy’. In both cases, we infer a marked expression from the denial of its unmarked antonym.Footnote 18 Furthermore, the inference from ‘not courageous’ to ‘cowardly’ is cancelable:
-
(10) Jill isn’t courageous, but neither is she cowardly.
It’s also reinforceable:
-
(11) Jack isn’t courageous; in fact, he’s downright cowardly.
So, the inference from ‘not courageous’ to ‘cowardly’ is plausibly an instance of negative strengthening.
We can now explain why we’re reluctant to assert (2): (2) conversationally implicates something else that we’re reluctant to accept, namely Q. The villain and fundamentalist do not seem cowardly – ’after all, look at how [they behave] in the face of danger,’ they face ‘great risk without flinching’ (Rachels Reference Rachels1999, 179–180). As suggested, this solution helps to shed light on Rachels’ expressed reluctance to assert (2).
On this view, the reluctance to assert (2) is rooted in an inference that is not logically valid, but is only a conversational implicature. It’s not uncharitable to think we’re misled by the conversational implicature in question. The present form of negative strengthening is an instance of what Grice calls ‘generalized conversational implicature’ – implicatures that are associated with the relevant expressions in all ordinary contexts (Grice Reference Grice1989, 37). And, according to Stephen Levinson, generalized conversational implicatures are ‘hard to distinguish from the semantic content of linguistic expressions’ because they are routinely associated with such expressions (Reference Levinson1983, 127). So, these generalized implicatures can mislead. A common example is provided in Section 7.1.
Of course, some people are not reluctant to think the 9/11 attacks are cowardly (e.g. George W. Bush). But this is not a problem for the above solution, since these people presumably have no reluctance to assert (2) whenever ‘A’ refers to the 9/11 attacks. After all, their firm view about the 9/11 attacks being cowardly logically entails that these attacks are not courageous. So, in this case, there’s likely no reluctance for dissenters to explain.Footnote 19 Of course, this doesn’t mean such people are never reluctant to deny that A is courageous; they may be reluctant when ‘A’ refers to a Nazi soldier or fundamentalist, and in these instances the above explanation would apply.
What about Maher’s Puzzle? Those who believe the 9/11 attacks are cowardly are committed to the falsity of Maher’s utterance. But, as noted in Section 3, this doesn’t by itself explain why these people would take offense. We also have no explanation for why offense would be taken by those who suspended judgment or believed Maher’s utterance was true. But it will now be shown that the above solution to the Courage Problem provides the needed explanation for all such individuals.
The fact that Maher’s third sentence caused offense can be explained by the Semantic Postulate along with another form of negative strengthening – litotes. Litotes is a form of understatement in which an affirmative is implied, though not entailed, by the negation of its contrary (Horn Reference Horn1989, 356). For example, in some contexts, ‘not unhappy’ implies the affirmative ‘happy’. If one utters
-
(12) Say what you want about her, Sue is not unhappy.
with a rise-fall tone of voice on ‘not unhappy’, this utterance would likely imply
S: that Sue is happy.
Linguists typically hold that litotic inferences are also a form of conversational implicature (Horn Reference Horn1989, 356). Litotic inferences, along with the Semantic Postulate, provide a basis for solving Maher’s Puzzle.
Notice that Maher’s audience could easily interpret his utterance as an instance of litotes. Recall Maher’s third sentence:
-
(13) Staying in the airplane when it hits the building, say what you want about it, it’s not cowardly.
(13) is structurally similar to the understatement in (12). In particular, it’s feasible to interpret Maher in (13) as understating his point for rhetorical effect (with no intention to misinform). And this is precisely what understatement is (Cruse Reference Cruse2006, 186). Moreover, it’s plausible that his utterance was seen as implying the affirmative
P: that staying in the airplane when it hits the building is courageous.
Recall that many people took him as ascribing courage to the terrorists’ acts, even though he never did so explicitly. Thus, it’s plausible that Maher’s utterance was interpreted as an understatement that implies P. And this amounts to interpreting it as an instance of litotes.Footnote 20
Why then were people offended? On the present view, Maher’s utterance offended people because those people interpreted it as implying P (via litotes, a form of negative strengthening). They were offended by P for precisely the same reason that people are reluctant to assert (1). P and (1) each entail that the intended ends are good, according to the Semantic Postulate. And these interpreters took issue with the putative goodness of the terrorists’ ends. In this way, we have an explanation for how Maher’s utterance could be interpreted as praising the terrorists’ ends. Most importantly, we have a solution to Maher’s Puzzle that is fundamentally unified with our solution to the Courage Problem – both solutions appeal to negative strengthening and the Semantic Postulate.
7. Objections
It has been argued that the Semantic Postulate is a necessary part of the best solution to Maher’s Puzzle and the Courage Problem. In particular, the Semantic Postulate explains our reluctance to assert (1), and it also plays a key role in explaining why people were offended by Maher’s utterance. In Sections 4 and 5, it was argued that these linguistic behaviors cannot be adequately explained without the Semantic Postulate. Still, this argument provides only prima facie reason to accept the Semantic Postulate – we can accept it only if there are no successful objections. Three objections shall now be addressed.
7.1. Semantic competence
The Semantic Postulate asserts that ‘courageous’ semantically entails the pursuit of good ends. This postulate requires that our semantic competence with ‘courageous’ disposes us to accept the following:
T: that, necessarily, if x is courageous then x pursues an end that is good.
However, assenters tend to reject T. So, one might object: the Semantic Postulate entails that either these assenters lack semantic competence with ‘courageous’ or they use ‘courageous’ in a different sense and are therefore talking past dissenters. This paper has disavowed any extra senses of ‘courageous’. So, the present view predicts that assenters lack semantic competence, which is absurd.
But this objection would prove too much. Indeed, it would show that we’re not disposed to accept purported laws of logic by our semantic competence with logical connectives. Quoting Matti Eklund,
There are long-standing philosophy disputes over the status of some purported logical laws. We would not want to say that in every such dispute, either some disputant lacks semantic competence or the disputants speak past each other, speaking different languages. It is possible to reject some law of logic without thereby manifesting [lack of] semantic competence. (Eklund Reference Eklund2002, 262)
Let’s assume the relevant disputants are not speaking past each other. How can one such disputant not lack semantic competence?
There are various ways of answering this, but Eklund’s approach will suffice. He holds that semantic competence involves dispositions that are defeasible – semantic competence disposes one to accept a proposition P in the absence of what one takes to be evidence against P (Reference Eklund2002, 252). Eklund doesn’t elaborate on what this perceived counterevidence must be. But clearly it can emerge from the fact that P is associated with a generalized conversational implicature. For example, if a speaker takes herself as having evidence against inclusive disjunction, she need not lack semantic competence with ‘or’ by rejecting inclusive disjunction. Indeed, the traditional Gricean explanation is that speakers who reject inclusive disjunction have misleading counterevidence. ‘P or Q’ conversationally implicates ‘not both P and Q’. These speakers are misled by a generalized conversational implicature, and thereby take themselves as having evidence against inclusivity.Footnote 21 On Eklund’s model, these people need not count as lacking semantic competence with ‘or’.
This approach also explains why the above objection fails against the Semantic Postulate. This postulate need only require that our semantic competence with ‘courageous’ involves a defeasible disposition to accept T. Since this disposition can be defeasible, the objector must show that those who reject T are not misled by putative counterevidence. But the proposed solution to the Courage Problem integrally involves misleading counterevidence. Since ‘A is not courageous’ implies that A is cowardly, and assenters are reluctant to think A is cowardly, they could be inferring by modus tollens that A is courageous, which is a counterexample to T. The trouble is that the inference from ‘not courageous’ to ‘cowardly’ isn’t logically valid, but is only a conversational implicature, which means this evidence against T would be misleading. So, these assenters can very well have the defeasible disposition to accept T, which means they need not count as lacking semantic competence.Footnote 22
A close rival to the Semantic Postulate might claim that speakers are reluctant to assert (1) because they merely believe that T is true. This alternative hypothesis makes no claim about semantic competence. The trouble is that there are many assenters who register reluctance to assert (1) but do not believe T (e.g. Foot Reference Foot2002, 16; Rachels Reference Rachels1999, 179–180). This rival hypothesis leaves their reluctance unexplained. But the Semantic Postulate can explain it: these assenters are reluctant because they have a defeasible disposition to accept T (by way of semantic competence). And this can be true even if they ultimately reject T due to misleading counterevidence.
7.2. Counterexamples
Although we’re reluctant to call the villain and fundamentalist courageous, we agree without reluctance on many paradigmatic cases of courage. It’s sometimes objected that these agreed-upon cases provide counterexamples to the Semantic Postulate.Footnote 23 For example, most people agree, without reluctance, that the first ascent of Everest was courageous (Walton Reference Walton1986, 53–54). But this act does not seem aimed at a good end.
The success of this counterexample depends on what is meant by ‘good’ within the Semantic Postulate. But ‘good’ can just mean good in a way (i.e. good in some respect). And most things are good in some way or other, since there are trivial ways of being good. As Judith Thomson notes, everyone is good in some way or other (Reference Thomson2008, 10). If you can think of someone who seems not to be good in any way, then he’s surely good in at least this way – he’s good to cite as an example of someone who seems good in no way. The point is that it’s very easy for something to count as good in a way. So, the first ascent of Everest will surely also count. The problem of counterexamples can be avoided.
But now the opposite problem arises. Too many things count as good in a way, indeed, so many things that the Semantic Postulate can no longer exclude the villain or fundamentalist from counting as courageous. For instance, the Nazi soldier’s ends are surely good in some way or other – they’re good things to mention in a discussion on courage.
Can the Semantic Postulate allow that the Everest-ascent is courageous without allowing the Nazi’s behavior? More generally, can the Semantic Postulate make the right predictions about possible cases? Yes. This postulate asserts that courage entails the concept expressed by ‘good in a way’. But it’s misleading to speak about the concept expressed by ‘good in a way’, because this phrase expresses many different concepts in different contexts of utterance.
One mechanism responsible for the context-sensitivity of ‘good in a way’ is the embedded quantifier ‘a’ or ‘some’. The bus clerk may say ‘There’s no way to get from Ithaca to New Haven,’ while the gas clerk may say ‘There is a way to get from Ithaca to New Haven.’ But it’s plausible that both could be right. Their utterances express different propositions, relative to different contexts. The bus clerk asserts that there’s no way by bus, while the gas clerk asserts that there is a way by car. Neither person asserts that there is (or isn’t) a logically possible way. They instead assert that there is (or isn’t) a way within a restricted class. And context determines what that restrictive class is.
The same is true when we say that x is good in a way. We are rarely asserting that there’s a logically possible way in which x is good, but are instead asserting that there’s a way within a restricted class. And, again, the context determines what the restrictive class is.
The Semantic Postulate should be restricted to a select class of ways in which the agent’s ends are good. Otherwise, it cannot exclude the villain and fundamentalist from being courageous, which means it would be useless to dissenters. But since there is no context of utterance associated with the Semantic Postulate, we must restrict the relevant class artificially. The clarified version of the Semantic Postulate places this artificial restriction by speaking of the relevant ways (which shall be explained):
Clarified Semantic Postulate: ‘x is courageous’ semantically entails that x pursues an end that is good in one of the relevant ways.
It’s now possible to exclude the Nazi’s end from being good in one of the relevant ways, while including the Everest-climber’s end.
What are the relevant ways of being good? The relevant class is whatever class allows the Semantic Postulate to explain, as best it can, our linguistic intuitions about when the term ‘courageous’ is and isn’t correctly applied. Parsimony is a virtue when it comes to fixing this class – we should be partial to the smallest possible class that allows the Semantic Postulate to explain our linguistic tendencies to apply or withhold the word ‘courageous’. If we include trivial ways of being good, which don’t explain these tendencies, we might thereby fail to exclude the Nazi from being courageous.
At the very least, the relevant class should include moral goodness, because this inclusion allows the Semantic Postulate to explain our reluctance to ascribe courage to the villain and fundamentalist. However, if moral goodness is the only way within the relevant class, then the Everest example might be a counterexample to the Semantic Postulate (assuming his ends lack moral goodness). Nevertheless, there is a non-moral way in which the Everest-ascent is good – it’s an achievement. So, we can expand the relevant class to include at least one non-moral way of being good – namely that particular way in which achievements are good.Footnote 24
This illustrates a strategy for dealing with potential counterexamples. The basic idea is that we use uncontroversial examples, like the Everest case, as means for discovering which ways of being good are within the relevant class. There might be other agreed-upon instances of courage that don’t involve achievement or moral goodness, and, if so, then the relevant class can be expanded out accordingly.Footnote 25
7.3. Speaker-relativism
A final objection proposes a view that fellow dissenters might hold. Here’s the rough gloss: Dissenters may hold that courage entails that the agent’s ends are believed to be good by the speaker. Call this the Speaker-Relative View.Footnote 26 This view requires us to be dissenters, because we don’t believe the villain’s and fundamentalist’s ends are good. But the Speaker-Relative View allows other speakers, who value the ends of the villain or fundamentalist, to be assenters. Is there reason to prefer the Semantic Postulate over the Speaker-Relative View?
It depends on how we’re meant to understand the Speaker-Relative View. The rough gloss above has problems with disagreement. Suppose a Neo-Nazi utters
-
(14) The Nazi soldier is courageous.
And I reply:
-
(15) The Nazi soldier is not courageous.
The Neo-Nazi and I genuinely disagree. Yet, if the rough gloss of Speaker-Relativism is true, then we’re talking past one another. (14) expresses a proposition that requires the ends to be valued by the Neo-Nazi, while (15) denies a proposition that requires the ends to be valued by me. So, we don’t genuinely disagree. For this reason, we cannot accept the rough gloss of Speaker-Relativism.Footnote 27
But there is a sophisticated form of Speaker-Relativism, which is said to avoid the problem of disagreement (e.g. MacFarlane Reference MacFarlane2007). On this view, evaluative propositions – not merely sentences – are true or false relative to speakers. So, the following proposition can be true relative to the Neo-Nazi and false relative to me:
U: that x pursues an end that is good.
Most importantly, this view permits the Neo-Nazi and me to express a proposition that entails U when making courage-ascriptions. So, the Neo-Nazi can express the same proposition in (14) that I deny in (15). The problem of disagreement might be avoided.
What does this mean for the Semantic Postulate? – Nothing. This sophisticated form of speaker-relativism is compatible with the Semantic Postulate. The Semantic Postulate says that ‘courageous’ semantically entails U. And this form of speaker-relativism merely introduces the caveat that U is true or false relative to speakers. But the Semantic Postulate is perfectly compatible with this caveat.
This paper has argued that courage requires the pursuit of good ends, and that gutsy villains are therefore not courageous. This requirement is an integral part of the best solution to two closely related problems – the Courage Problem and Maher’s Puzzle. Bill Maher unwittingly raised an issue related to the Courage Problem, though he did so without using the word ‘courage’. The absence of this word from his utterance makes it difficult for opponents of this requirement to solve both problems.
Acknowledgments
Special gratitude goes to the cadets in my sections of Ethics at USAFA, whose questions, comments, and insights about courage led to the development of this project. Thanks also to Matti Eklund and Kent Dunnington for helpful feedback and willingness to discuss various drafts of this paper. I am also indebted to three anonymous reviewers, as well as USAFA’s Philosophy Department, for their helpful comments. And finally, I thank the participants of two events – the 2013 Conference for the International Society for Military Ethics and the 39th Conference on Value Inquiry – for providing feedback on early versions of this article. The viewpoints expressed in this article do not reflect the official positions of the U.S. Air Force Academy or any other government agency.
