Context, relevant parts and (lack of) disagreement over taste
Cappelen and Hawthorne’s book provides a very rich and interesting set of challenges for those of us who have advocated semantic theories in which certain sentences express contents which are assigned truth values only relative to parameters such as a “judge” or standard of taste. It would be impossible to give a comprehensive assessment and response to Cappelen and Hawthorne’s program in the space allotted here; I will focus instead on a one specific argument they make with regard to the semantic analysis of expressions of personal taste, and attempt to show that what they suggest is a serious problem for relativist semantics is not, in fact, threatening to the relativist program at all.1
As Cappelen and Hawthorne point out, the primary motivation for a relativist semantic theory is to account for the phenomenon of “faultless disagreement,” in which two or more individuals seem intuitively to disagree with each other, but where neither of them seems to be making an error of fact. If, for example, John says “Roller coasters are fun,” and Mary says “Roller coasters are not fun,” we take them to be disagreeing—yet intuitively, it seems purely a “matter of opinion,” not a matter of fact, whether roller coasters are fun; likewise if they say “Licorice is tasty” and “Licorice is not tasty.” In a relativist semantic theory, we may allow that sentences like these express contents which may be true relative to one person but false relative to another; in other words, that there is no objective fact whether the content expressed by the sentence is true. A notion of contradiction may be defined for such contents; thus John and Mary may assert mutually contradictory contents, without either making an error of fact. Of course, no such claim is made for sentences expressing contents whose truth values do seem intuitively to depend entirely on matters of fact, such as ‘Roller coasters have been built in six amusement parks during the last year’, or ‘Licorice is sold in half-pound bags’; these are treated as expressing contents whose truth values do not vary from person to person.
The idea just sketched conflicts with the position Cappelen and Hawthorne defend, and which they call “simplicity”—the thesis that the semantic values of declarative sentences in context are propositions, which are true or false simpliciter, and which serve as the objects of illocutionary acts such as assertion and denial, as the objects of propositional attitudes, and as the objects of agreement and disagreement. It should be noted that this position does not deny that there may be a notion of truth-relative-to-a-parameter which a semantic theory may usefully employ; but Cappelen and Hawthorne argue that any such notion should be regarded as derivative from the more basic concept of monadic truth, and applicable to sentences as opposed to sentence contents; they are committed to the idea that the contents of declarative sentences (and hence the objects of assertion, denial, agreement, disagreement, etc.) are always assigned truth values simpliciter, i.e. with a monadic truth predicate.
The idea that declarative sentences express propositions, and that propositions are true or false simpliciter, is certainly a traditional, mainstream position in philosophy. But we should be careful not to allow the orthodoxy which this position has achieved lead us into thinking that it is just common sense. On the contrary, the idea that some assertions concern matters of fact, and others merely concern “matters of opinion” or “matters of taste” is firmly rooted in pretheoretic intuition and everyday experience.2 If we seriously advocate the claim that the contents of declarative sentences are always true or false simpliciter (and assume that monadic assignment of truth values is determined by matters of fact), this amounts to a denial of the legitimacy of this intuition, and a radical departure from the common sense view.
I do not mean to suggest that theoretical analysis should always conform to common sense, nor that a relativist semantic theory is the only one which can capture the distinction between sentence contents concerning matters of fact and sentence contents concerning matters of taste. But a theory which radically departs from our pretheoretic intuitions requires special justification, and owes an explanation why these intuitions arise.
In place of a relativist analysis of expressions of personal taste, Cappelen and Hawthorne advocate a contextualist analysis, in which sentences like ‘Roller coasters are fun’ or ‘Licorice is tasty’ express different contents relative to different groups or individuals with different standards of taste, with the relevant group or individual fixed by the context of use. Extended arguments against such an analysis have been given elsewhere (Kölbel 2002; Lasersohn 2005); here I will simply repeat one objection: In such an analysis, if one person says that something is fun, and another says that it is not fun, they are disagreeing only if both claims are made in contexts which fix the relevant parameter to the same value. In this case, the disagreement should be resolvable by consulting or observing the group or individual to whom the parameter is set. However, people may persist in disagreement based purely on their own tastes, and need not concede the point in response to information about the tastes of others. If I have experienced the roller coaster and know that I did not enjoy it (and that my experience was not affected by unusual factors such as illness), then my judgment will be that the roller coaster is not fun, and is unlikely to change simply because I learn that a survey has shown that most people do enjoy it; I simply disagree with them.
Of course this does not mean that our tastes never change; nor does it imply that we will always persist in debating those with whose tastes we disagree.3 But it is desirable for any theory dealing with these matters to explain—or at least show that it is compatible with—the fact that (1) we feel a license to make judgments of tastiness, fun, etc., based purely on our own experiences, but (2) in doing so we disagree with others who draw the opposite judgment based on their own experiences, and (3) we may persist in that disagreement even when it is clear to us that others have come to the opposite judgment based on their own experiences, no matter who they are or how many. As far as I can see, this pattern is completely unexpected on Cappelen and Hawthorne’s analysis (and on other contextualist analyses).
Suppose a caterer says of a certain party ‘That party is not going to be fun. I have to cook hors d’oeuvres all night.’ Suppose that, meanwhile, someone in a separate conversation says of the same party ‘That party is going to be fun. I get to meet lots of school buddies that I haven’t seen in a long time.’ In this case we have absolutely no strong sense at all that the people are in disagreement. Another example: a child says ‘The summer is going to be fun. I get to go to music camp.’ A parent, in a separate conversation, says, ‘The summer isn’t going to be fun. I have to work overtime to pay for my child’s music camp.’ Once again, it would be silly to claim there is a contradiction between the two speeches.
Before considering the implications of examples like this for relativist semantic theory, it is worth setting aside one potential objection as misguided: that the lack of disagreement in these examples is due to the fact that the speakers in each case are in separate conversations. As Cappelen and Hawthorne point out, the word ‘disagree’ is ambiguous: on one reading, it is an activity verb and means something like “engage in a dispute,” but on another it is a stative verb and means something like “hold mutually incompatible beliefs.” The second, stative, reading is at issue in the debates about faultless disagreement, and not merely the first; but in the stative sense, two people may disagree without ever engaging in conversation with one another, or even being aware of each other’s existence.5
A second point that bears emphasis in considering these examples is that they are problematic only for those versions of relativist semantic theory which claim that predicates like ‘fun’ are always interpreted relativistically. But it is perfectly compatible with the general program of relativist semantics to claim that such predicates are only sometimes assigned a relativistic reading (which may give rise to faultless disagreement), and sometimes assigned a contextualist or hidden-indexical reading (which does not). Indeed, this position is argued in some detail in Stephenson (2007), and mentioned briefly as a possibility in Lasersohn (2005).
In contrast, Cappelen and Hawthorne are committed to the position that the contents of declarative sentences, including expressions of personal taste, are always true or false simpliciter. Legitimate examples of sentences expressing contents which are true or false only relative to parameters thus pose a far deeper problem for their program than examples of sentences expressing personal taste whose contents are true or false simpliciter do for the relativist program.
But even the claim that predicates like ‘fun’ and ‘tasty’ are always interpreted relativistically can be adequately defended in the face of examples like Cappelen and Hawthorne’s, I think. All that is really necessary is to recognize that a single expression can show more than one kind of contextual sensitivity in its semantics, and to remember that predication in natural language is often done on the basis of only certain parts or aspects of the object to which the predicate is applied.
This latter point is probably most familiar from discussion of color predicates (e.g. Lahav 1989): We call a watermelon red or yellow depending on the color of its interior edible flesh, but call an apple red or yellow depending on the color of its exterior skin. But this phenomenon is not limited to color predication by any means; on the contrary, it is pervasive throughout the vocabulary: We call something a “brick building” if bricks form the primary material of the exterior walls, regardless of whether they are the primary material of the building as a whole; the mayapple may be found on lists of “edible plants” because of its fruit, even though the roots are poisonous.
The part of an object which is relevant for the applicability of the predicate to the object as a whole depends in part on the particular predicate and the kind of object involved, but is sometimes also dependent on contextual factors, as is familiar from the work of Charles Travis and others.6 The same watermelon which is described as red by a chef preparing a fruit salad may be called green by a photographer setting up a shot. Notice that in this case we do not consider the chef and the photographer to be disagreeing with each other about the color of the watermelon; we understand that different parts of the watermelon are relevant to its color classification in different pragmatic contexts, and that the content expressed by the chef’s and photographer’s descriptions varies accordingly. Of course this does not require us to analyze the phrase ‘that watermelon’ as having a different denotation in the chef’s utterance of ‘That watermelon is red’ than it does in the photographer’s utterance of ‘That watermelon is green’; we can attribute the contextual effect to the predicate—an idea that has already been worked out in a variety of different ways by different authors (Kennedy and McNally 2010; Rothschild and Segal 2009; Szabó 2001).
More to the point, a party or a summer may be regarded as a complex event, with a variety of different subevents as parts—events of guests mingling, caterers cooking, etc., are parts of the party; events of children attending camp, parents working overtime, etc., are parts of the summer. And just as we may apply a predicate to a watermelon or a building based on its applicability to a relevant part, we may also apply predicates to parties or summers based on their applicability to relevant parts: It is natural to describe a party as outdoors based on the location of the guests, regardless of the location of the caterers (even if the caterers outnumber the guests), for example. And here also we find that pragmatic factors sometimes play a role in determining which part is relevant, so that this varies from context to context: Suppose John invests in the stock market, and also in real estate, and that over the summer his stock investments did very poorly, but his real estate investments did well. In a conversation with his stockbroker, he might truthfully complain “This summer was not at all profitable for me,” but in a different conversation with a fellow real estate speculator he might just as truthfully brag “This summer has been very profitable for me.”
None of these examples involve personal taste or faultless disagreement; they lend themselves well to an ordinary contextualist analysis, in which context affects which content is expressed, and sentence contents themselves are true or false simpliciter. But there is no reason to expect that predicates of personal taste will not show a similar contextual effect, over and above the relativization to individuals or standards of taste which is the target of Cappelen and Hawthorne’s criticism. In particular, we may analyze the caterer’s and the party guest’s utterances as made in contexts which make different parts of the party (the food preparation and the socializing, respectively) relevant to the applicability of the predicate to the party as a whole, with the content varying accordingly—and similarly for the example of the parent and child talking about the summer.
It is clear enough on independent grounds that examples like these involve a different kind of contextual effect from the one used in Lasersohn (2005) and elsewhere to motivate a relativist semantics. There, the idea was that certain sentences express contents which may vary in truth value from individual to individual according to their personal tastes. Although it was not made explicit in that article, it is natural to assume, if one is relativizing truth values according to differences of taste, that if a sentence content ϕ is true relative to one individual x, but false relative to some other individual y, then x and y do not share exactly the same tastes. In fact one might, as MacFarlane (2007) has done, relativize to standards of taste rather than individuals at all.
But in Cappelen and Hawthorne’s examples, despite the fact that the caterer says “That party is not going to be fun” and the guest says “That party is going to be fun,” we need not conclude that the caterer and guest have different tastes at all. They might, in fact, enjoy precisely the same kinds of work, entertainment, etc. This shows that their assertions differ in some way other than the setting of the taste parameter, and render the example irrelevant to the issue Cappelen and Hawthorne are attempting to address. To give an authentic counterexample to the claim that expressions of personal taste are always interpreted relativistically, one would have to control for all other contextual effects, by setting up the example in such a way that the only difference in context between two assessments of a sentence for truth or falsity is in the personal tastes of the individuals performing the assessments. If, in such a case, it were found that one individual assessed the sentence as true, and the other as false, but we intuitively did not feel that they disagreed with one another, then the relativist would have to fall back to the position that expressions of personal taste are at best ambiguous between relativistic and contextualistic readings, and not unambiguously relativistic. My expectation is that no such example will be found.
Put somewhat differently, to show that the effect of personal taste on truth values should be treated using contextualist, rather than relativist, techniques, it would suffice to show that incompatibilities of taste do not constitute real disagreements (or that they are never faultless)—but it will not do to show that contextual differences that have nothing to do with personal taste fail to produce intuitions of disagreement.
Responses to several others of Cappelen and Hawthorne’s arguments are given in Lasersohn (2008).
Some caution is necessary in using the term ‘opinion’ in a way which opposes it to ‘fact’, since we can certainly have (and express) opinions about issues of fact. But it seems to me that the longer phrase ‘matter of opinion’, more than the lone word ‘opinion’, suggests an independence from purely factual issues; it seems perfectly normal, for example, to say that it is a matter of opinion whether the licorice is tasty, but it would be quite odd to say that it was a matter of opinion whether the licorice contains sugar.
As many people have pointed out to me, people often drop a debate as soon as it becomes clear that it is over a matter of taste. It seems to me, however, that we do not drop such debates because our sense of disagreement has disappeared, but rather because we know that it is not resolvable.
Oddly, Cappelen and Hawthorne also class ‘filling’ and ‘spicy’ as predicates of personal taste. It may be that these predicates also give rise to faultless disagreement, but only, I think, because they are scalar; it hardly seems a matter of taste whether something is filling or spicy. (Of course one person may find filling or spicy food to his or her taste, and another person not, but this is an entirely separate question.) For more discussion, see Lasersohn (2008).
The term ‘faultless disagreement’ is a handy one, but one that has the potential for causing confusion, not only because of the ambiguity that Cappelen and Hawthorne point out, but also because in ordinary, pretheoretic talk, we understand the word ‘disagreement’ more broadly than what is at issue in recent debates about relativism. If for example John says “I like roller coasters” and Mary responds by saying “Well, I don’t,” we intuitively take them to be disagreeing in some sense (and presumably neither is making an error of fact); but such examples do not pose any obvious problem for conventional semantic theories or provide any obvious motivation for relativist theories. In contrast, examples like “Roller coasters are fun”/“Roller coasters are not fun,” provoke a sense of direct contradiction, and not merely “disagreement” in the broad sense. The challenge for semantic theory is in accounting for this intuition of direct contradiction and simultaneously for the intuition of faultlessness which such examples produce.
I do not intend this as implying that I accept the conclusions which Travis draws from such examples.
- Kölbel, M. (2002). Truth without objectivity. London: Routledge.Google Scholar
- Stephenson, T. (2007). Towards a theory of subjective meaning. Cambridge, MA: Massachusetts Institute of Technology dissertation.Google Scholar
- Szabó, Z. (2001). Adjectives in context. In I. Kenesei & R. M. Harnish (Eds.), Perspectives on semantics, pragmatics and discourse: A festschrift for Ferenc Kiefer (pp. 119–146). Amsterdam: John Benjamins.Google Scholar