Travis Cases and Ambiguity

In ‘Pragmatics’, Charles Travis develops an argument against the view that linguistic meaning, which determines locutionary act content, also determines truth-conditional illocutionary act content. (Travis 1997.) The argument is based around cases like the following:

Pia’s Japanese maple is full of russet leaves. Believing that green is colour of leaves, she paints them. [Case 1:] Returning, she reports, ‘That’s better. The leaves are green now.’ She speaks truth. [Case 2:] A botanist friend then phones, seeking green leaves for the study of green-leaf chemistry. ‘The leaves (on my tree) are green,’ Pia says. ‘You can have those.’ But now Pia speaks falsehood. (Travis 1997: 89)

According to Travis, we can suppose that the same words, with the same meanings, are used in both cases. Moreover, we can assume that the operations of indexicals are confined to determining which leaves are being spoken about, and that the same leaves are being spoken about in both cases. Thus, we can assume that the same type of locutionary act, with the same locutionary content, is performed in case 1 and case 2. And since the same leaves, in the same condition, are being spoken about in both cases, if ‘The leaves are green’ were used in both cases to make the same statement and, so, to determine the same truth conditions, then Pia would either state a truth in both cases, or state a falsehood in both cases. Since it is plausible that Pia can state a truth (and no falsehood) in case 1 and that she can state a falsehood (and no truth) in case 2, even though the leaves are in the same condition in both cases, we have that ‘The leaves are green’ is not used in both cases to make the same statement. So, we have that the statement that Pia makes in case 1 is not the same as the statement that she makes in case 2. Finally, if we assume that the statements that Pia makes in both cases are basic illocutionary acts, we have grounds for rejecting the view that locutionary act content determines basic illocutionary act content. For we have that Pia performed locutionary acts with the same contents and basic illocutionary acts with different contents and, moreover, that the difference in illocutionary content was not due to the operation of indexicals. And since n-tuples of cases of this sort can be constructed for most, if not all, substantive expressions, there are grounds for generalizing Travis’ conclusion: most, if not all, substantive expressions function in a similar way to discretionary indexicals.

Travis’ argument depends upon three main assumptions. The first assumption is that the variation in truth conditions exhibited across cases 1 and 2 affects basic illocutionary acts performed in those cases. For if the variance affected only derivative illocutionary acts (e.g. Gricean implicatures), it would be consistent with maintaining that locutionary act content determines basic illocutionary act content. The second assumption is that the variation is not due to ambiguity. For if ‘The leaves are green’ was used with relevantly different meanings across the two cases, the variation in truth conditions might be traced to a difference in locutionary act content. The third assumption is that the variation is not due to the operation of non-discretionary—that is, meaning controlled—indexicality, like that plausibly exhibited by ‘I,’ ‘here,’ and ‘now.’ For if it were due to such non-discretionary indexicality, then the variation in truth conditions might again be due to variation in meaning-controlled locutionary act content. Travis’ second and third assumptions have been subjected to interesting challenges in recent years. In this note, I briefly consider one such challenge.

Christopher Kennedy and Louise McNally have argued that ‘is green’ carries two kinds of meanings, one kind on which it is used to attribute a gradable quantity or quality of perceptible greenness, and another on which it is used to attribute a property that is conventionally correlated with the colour green, for example a biochemical property. They argue moreover that a gradable meaning is operative in case 1, which therefore comes out true because a sufficient quantity of the surface of the leaves is perceptibly green, while the correlational meaning is operative in case 2, which therefore comes out false because the biochemical property conventionally correlated with greenness in leaves—leaves with which property are sought by Pia’s botanist friend—is absent with respect to the painted leaves. Thus, they seek the explain the divergence in truth values across the two cases by appeal to a subtle ambiguity. (Kennedy and McNally 2010.)

One worry about this type of challenge is that an unsupplemented appeal to an n-way ambiguity can explain only n-way variation in truth conditions. If Travis’ argument were based exhaustively on the pair of cases with which we began, then appeal to a two-way ambiguity would suffice. However, it is a straightforward task to multiply cases. For one example amongst very many, we might do so by appeal to cases in which the leaves look green in natural light, but yellow in the artificial light: depending on operative conversational ends—for instance, on whether the leaves will end up indoors or outdoors—it might be reasonable to use, ‘The leaves are green,’ in order to say something true or something false. In order to capture all such variations, we would be forced to appeal to a distinct meaning for each case, and so forced to appeal to a vast—perhaps indefinitely large—variety of distinct meanings for each susceptible expression.

Such an account would be unacceptable. It is, at best, unclear how we could gain facility with such an array of meanings. And even if it were possible to acquire competence with each of the meanings, doing so would supply, at best, very little guidance to what speakers were up to. For, as the number of distinct meanings increases, the task of discerning which meaning a speaker intends to deploy on a particular occasion approximates to the task of discerning the speaker’s basic illocutionary intentions without any guidance at all from the meanings of the words they use. Finally, it is far from clear that if were able to focus attention on one amongst the array of meanings carried by ‘The leaves are green,’ we would thereby have eliminated the sort of variation in truth conditions to which Travis appeals.

In order for a meaning to determine truth conditions in a way that eliminated all such variation, it would have to determine, for all possible variations in prevailing conditions, precisely how those variations would bear on truth-value. For suppose that the meaning failed to determine the bearing of a specific form of variation amongst possible cases—for instance, by failing to determine precisely how much of a leaf would need to look green, for how much of its career, for what reasons, in which viewing conditions, and so forth. There might then be pairs of cases differing in that specific way such that the meaning of the target expression left open whether the expression would be used to say something true on each of those occasions. And in that case, there would be room for speakers’ ends, in addition to the end of according with the meanings of their words, to figure in determining whether what was stated on each occasions was true. It is difficult to understand how the meanings of any expression that we could usefully employ could include sufficient detail to foreclose on speakers’ wider ends playing some role in determining the truth conditions of what they state.

Kennedy and McNally don’t claim that the ambiguity that they posit is the only source of variation in truth conditions, even with respect to colour adjectives. They allow that the applicability of gradable adjectives, including their gradable form of ‘is green,’ depends upon a standard of comparison that isn’t determined by meaning. So, they allow that even if we hold fixed that ‘The leaves are green’ is used with its gradable meaning, and that it is used with respect to the same leaves in the same condition, there can nonetheless be variation in truth value due to variation in the operative standard of comparison. And they allow that the applicability of their correlated property form depends upon the selection of one or another correlated property, again allowing for case-by-case variation in truth value even holding fixed the use of that form. Even if successful, then, their proposal serves not to eradicate, but only to mitigate, the sort of variation on which Travis’ argument is based. Thus, considered on its own, their proposal lacks the resources to save the view that Travis advises us to reject. (Kennedy and McNally suggest that their proposal might be combined with the proposal that some of the variation in truth conditions exhibited in Travis’ cases is to be explained by appeal to non-discretionary indexicality.)

Furthermore, it is far from clear that Kennedy and McNally’s proposal should be accepted. There are two main difficulties.

The first difficulty is that even if we were to confine attention to perceptible greenness, and so to do our best to focus on the gradable meanings, the possible variations in truth-value seem to go beyond what would be predicted by simple appeals to variation in an operative threshold of quantity or quality of colour coverage. For example, suppose that an object were coloured in such a way that it appeared uniformly green only when viewed from certain distances, or certain angles, otherwise appearing yellow. It seems plausible that, depending on the mutually manifest ends of a speaker and their audience, ‘The object is green’ might be used on some occasions to speak truth about the object and on others to speak falsehood. (Suppose, for instance, that what is wanted is a green ornament, and that the agreed location for the ornament will constrain the angle and distance of viewing.) And yet it is far from obvious that such variation must be due to variation in an operative threshold quantity (or quality) of colour coverage. So, it is plausible that the meaning or meanings of ‘is green’ provide even less detailed guidance to truth-conditions than Kennedy and McNally suggest.

The second difficulty is that the evidence to which Kennedy and McNally appeal in support of there being distinct gradable and correlational meanings for ‘is green’ is less clear-cut than they suggest.
With respect to the circumstances of case 1, it would seem perfectly acceptable for Pia to qualify her colour attribution: ‘These leaves are greener than those,’ ‘These leaves are not green enough,’ and so forth. By contrast, with respect to the circumstances of case 2, in which Pia’s overriding interest is supposed to be in the biochemical properties of the leaves, Kennedy and McNally claim that attempts at such qualifications would be unacceptable. Their conclusion is that the difference in acceptability of qualifications of this sort is to be explained by a difference in the operative meaning of ‘is green,’ specifically that such qualifications are acceptable only with respect to the gradable meanings.

One concern here is that the clarity of the contrastive judgment to which Kennedy and McNally appeal seems to depend upon treating case 2 as one on which what matters to Pia is the presence or absence of a biochemical property that is not itself gradable. However, if, say, the quantity and quality of natural greenness were correlated with the quantity and quality of the sought for property, it would be less obvious that Pia couldn’t acceptably qualify what she said to her botanist friend: ‘These leaves are green enough,’ ‘These leaves are less green than those.’ (It might remain true that, as Kennedy and McNally observe, it would not be natural for Pia to explain to her botanist friend why her painted leaves were unsuitable by saying that they were ‘not green enough.’ However, that might plausibly be explained by appeal to the fact that, for those purposes, the painted leaves would not count as green at all, and so it would be misleading to claim that they were insufficiently green.)

A second concern is that it is not obvious why the contrast to which Kennedy and McNally appeal is better explained by appeal to ambiguity than by appeal to what it is possible for Pia to say. We are assuming that, in uttering, ‘The leaves are green,’ Pia aims to be saying something about the leaves’ natural colour, where being so coloured is (let us suppose) not a gradable quality. But if being so coloured is not a gradable quality, then it is not clear what Pia could be saying, or trying to say, about that quality in using the problematic qualifiers. So, it is not obvious that the best way to explain the contrasting applicability of qualifiers is by appeal to meanings rather than by appeal to what it would make sense to say.

In summary, then, it is not obvious that Kennedy and McNally’s proposal, according to which ‘is green’ is ambiguous, should be accepted. Furthermore, even if accepted, the proposal is consistent with Travis’s claim that meaning, and so locutionary act content, serves as a guide to, rather than as a determinant of, the contents of basic illocutionary acts. Their disagreement with Travis concerns only the quantity and quality of guidance that meanings provide. Thus, accepting their proposal is consistent with rejecting the view that linguistic meaning, which determines locutionary act content, also determines truth-conditional illocutionary act content.


Kennedy, C. and L. McNally, 2010. “Color, context, and compositionality.” Synthese 174: 79–98.
Travis, C. 1997. “Pragmatics.” In B. Hale and C. Wright eds. A Companion to the Philosophy of Language. Oxford: Blackwell.

  1. John Collins said:

    I very much agree. Two further thoughts.

    (1) In the clause for the non-gradable case ([[green]] = λx. Pi(x) & cor(Pi, green)) it seems that the correlation relation is itself open to occasion-sensitive interpretation, so if that is the lexical entry on one disambiguation, one still doesn’t have a determinate contribution to truth conditions even after disambiguation.

    (2) It is unclear what to do about copredication on the K&M account, e.g., ‘The car and traffic light are red’. That seems fine to me, but then what is ‘red’ here?

  2. Thanks. Yes, both comments sound exactly right to me.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: