The World Philosophy Made

by Scott Soames

  Indeed, without the models provided by these truth-theoretic investigations plus philosophical theories of speech acts and philosophical accounts of conversational or communicative implications of our assertive utterances, today’s thriving empirical sciences of linguistic meaning and language use—practiced by distinguished linguists including Barbara Partee, Angelika Kratzer, Irene Heim, Craige Roberts, Geoffrey Pullum, and Paul Elbourne—wouldn’t exist. Thus if we ask, “What did philosophers of language from Frege through the mid-twentieth century contribute to our understanding of language?” the answer is that they gave us the theoretical core of the scientific study of linguistic meaning as we know it today.

  Nevertheless, we now recognize certain limits to work done in that philosophically dominant paradigm. Because sentences that impose necessarily equivalent conditions that the world must satisfy in order to conform to how the sentences represent it to be can have substantially different meanings, sentence meanings can’t be identified with such truth conditions. Nor can the things we assert, believe, and know be identified with sets of possible world-states satisfying the conditions. When it became clear in the mid-1980s that these problems couldn’t be solved by swapping theories of truth at possible world-states for theories of truth at circumstances of any sort, there was a revival of interest in Frege-Russell propositions, as additions to (rather than wholesale replacements of) theories of truth at possible world-states.10

  Nevertheless, the mere passage of time—more than a half century—didn’t obliterate the problems that previously undermined Frege-Russell propositions. For Russell, propositions had been structured complexes of objects and properties combined in unexplained ways. For Frege, mysterious entities called “senses” were thought to somehow fit together like pieces in a jigsaw puzzle, despite being nonspatial. For both, the representational properties of propositions, in virtue of which they had truth conditions, were taken for granted rather than explained, while agents were said to represent things as being certain ways by virtue of bearing an unexplained relation to mysterious, ill-understood propositions. In short, Frege’s and Russell’s otherworldly conception of propositions as sui generis abstract objects (existing independently of us outside of time and space) made it impossible to explain how propositions manage to represent anything, how agents succeed in believing them, or why recognizing them at all should be needed for agents to represent things as being certain ways.

  At this point it began to dawn on some philosophers and theoretical linguists that minds themselves might be the source of representation, and that the things asserted, believed, and known—which are also primary bearers of truth or falsity as well as being meanings of (some) declarative sentences—might be cognitive in nature. Although different theorists conceptualized this in different ways, the general idea was, and is, that propositions are cognitive acts, operations, states, products, or processes.11

  According to one main development of this idea, agents represent things as being various ways when they perceive, visualize, imagine, or otherwise think of them as being those ways. Propositions are then identified with repeatable, purely representational, cognitive act types or operations. For example, when one perceives, imagines, or thinks of B as hot, one predicates being hot of B, and so represents B as hot. This cognitive act represents B as hot in a sense similar to the derivative senses in which various socially significant acts can be said to be insulting or irresponsible. Roughly put, an act is insulting when for one to perform it is for one to insult someone; it is irresponsible when to perform it is to neglect one’s responsibilities. A similar derivative sense of representing can be used to assess the accuracy of cognitions. When to perceive or think of o as F is to represent o as it really is, we identify an entity, a particular cognition—a mental operation or doing—plus a property it has when it is accurate. The entity is a proposition, which is the mental act or operation type of representing o as F. The property is truth, which the act has if and only if for one to perform that representational act or operation is for one to represent o as o really is.

  To judge that B is hot is to predicate being hot of B in an affirmative manner, which involves forming or reinforcing dispositions to act, cognitively and behaviorally, toward B in ways conditioned by one’s reactions to hot things. To believe that B is hot is to be disposed to judge it to be hot. To know that B is hot is for B to be hot, to believe it is, and to be safe or justified in so believing. Since believing p doesn’t require cognizing p, any organism that can perceive or think of the objects and properties in terms of which p is defined can believe p, whether or not the organism uses language or can predicate properties of propositions. Knowing things about propositions requires distinguishing one’s cognitive acts or operations from one another. Self-conscious agents who can do this can ascribe attitudes to themselves and others, and predicate properties of propositions. Focusing on their own cognitions, they identify distinct propositions as different thoughts, which leads them to conceive of truth as a form of accuracy.12

  In this way, we explain how an organism without the concept of a proposition, or the ability to cognize one, can know or believe one. We also explain how sophisticated agents acquire the concept, and come to know things about propositions. What about meaning? What is it for a proposition p to be the meaning of a sentence S? It is for speakers following the linguistic conventions of their community to use S to perform the representational cognitive act p; learning a language is learning how to use its sentences to perform the mental acts conventionally associated with them. For example, one who understands ‘Plato was human’ uses the name to pick out the man, the noun to pick out humanity, and the phrase ‘was human’ to predicate the property of the man, thereby performing the cognitive act—predicating humanity of Plato—that is the meaning, p, of the sentence.

  Surprisingly, however, p isn’t the only proposition the speaker thereby expresses. Because using the sentence ‘Plato was human’ to predicate humanity of Plato is itself a purely representational cognitive act, it also counts as a proposition p*, distinct from, but very closely related to, p. Since to perform p* is to perform p, but not conversely—just as to perform the act driving to work is to perform the act traveling to work, but not conversely—propositions p and p* are cognitively distinct, even though they represent the same thing as being the same way, and so are representationally identical.

  The importance of representationally identical but cognitively distinct propositions is illustrated by sentences (1a–b), both of which express the proposition p that is the act predicating being a planet of Venus.

  1 a.  Hesperus is a planet.

     b.  Phosphorus is a planet.

  As with traveling to work, there are many ways of performing the predication, including (but not limited to) one that involves identifying the predication target (Venus) by using a name for it—e.g., ‘Hesperus’, ‘Phosphorus’, ‘Venus’—or by seeing it and focusing one’s attention on it. With this in mind, consider utterances of (1a) and (1b). In addition to the proposition p that consists in predicating being a planet of Venus (no matter how one identifies it), one who assertively utters (1a) also asserts the proposition pH, which is the act predicating being a planet of Venus, using ‘Hesperus’ to identify it. In addition to p, one who assertively utters (1b) asserts the cognitively distinct but representationally identical proposition pP, which is the act predicating being a planet of Venus using ‘Phosphorus’ to identify it. Since pH and pP are different, the pair of propositions—pH and p—asserted and believed by someone sincerely uttering (1a) differs from the pair pp and p asserted and believed by someone sincerely uttering (1b), despite the fact that the common member, p, of each pair is the meaning, or semantic content, of both sentences.

  Examples like these illustrate another point: there is more to understanding a sentence than being able to use it with its linguistically determined semantic content. Understanding a sentence also involves the ability to use it in conformit
y with widely shared presuppositions in one’s linguistic community. For example, even though those well enough informed to employ the names ‘Hesperus’ and ‘Phosphorus’ are not required to know that the two names designate the same thing, they are expected to know that those who use the names typically presuppose that ‘Hesperus’ stands for something visible in the evening sky, while ‘Phosphorus’ stands for something visible in the morning sky. One who mixed this up—taking ‘Hesperus’ to be the morning designation and ‘Phosphorus’ the evening designation—would misunderstand the names. Because of this, normal utterances of (1a) assert and convey the information that the body Hesperus that is seen in the evening is a planet, while normal utterances of (1b) assert and convey the information that the body Phosphorus that is seen in the morning is a planet.13

  The treatment of uses of (1a–b) extends to all forms of cognition that give rise to representationally identical but cognitively distinct propositions, and to uses of sentences to express or report cognitive attitudes. Let p be the simple proposition the performance of which involves focusing on object o and predicating property F of it. Suppose in a particular case that the sub-act of focusing on o can be performed either by using a name of o, or by perceiving o (visually or aurally). Taking only this variation into account gives us several representationally identical but cognitively distinct propositions—the proposition p of predicating F of o in some way or other, the proposition p* of predicating F of o, cognizing o by using its name, and the proposition p**, cognizing o visually, and so on.

  Next, suppose one predicates some property of oneself, cognizing oneself introspectively, which is different from any other way of cognizing oneself. Here again, a basic cognitive sub-act of a larger propositional act of predication is performed in a special way, with the result that the agent simultaneously performs both the general act of predicating F of someone or something, and the more specific act of doing so in a certain way, in this case by identifying the predication target introspectively. Because of this, one may fail to bear certain cognitive attitudes—e.g., belief or knowledge—to a proposition that predicates F of oneself, cognized introspectively, even though one does believe or know propositions that ascribe F to one, cognized in some other way. For example, if, after being stunned by a blow on the head, I forget my name—while continuing to believe that ‘Scott Soames’ names some professor or other—I may continue to believe that Scott Soames is a professor, while temporarily failing to believe that I am Scott Soames and wondering whether I am a professor.

  Similar points can be made about propositions in which one predicates a property of the present moment, cognized as subjectively present, versus cognitively distinct but representationally identical propositions in which the property is predicated of the very same moment, cognized via some objective presentation, like a date (and time). Just as for each person p there is an introspective first-person way of cognizing p no one else can use to cognize p, so, for each time t there is an immediate “present-tense” way of cognizing t at t that can’t be used at other times to cognize t. Suppose I plan to attend a meeting that will start at t—noon on March 31st. Not wanting to be late, I remind myself of this that morning. Nevertheless, as the morning wears on, I lose track of time. So, when I hear the clock strike noon, I say “The meeting starts now!” and change my behavior. Coming to believe of t in the present-tense way that the meeting starts then motivates me to hurry off. Had I not believed this, I wouldn’t have done so, even though I would have continued to believe, of that very time t, that the meeting starts then. In this case, I believe something new by coming to believe something old in a new way. What makes it true for me to say, at t, that I only just realized that the meeting starts now is that the proposition to which I have only just come to bear the realizing relation requires cognizing t in the immediate present-tense way. By using considerations like these, investigators are beginning to make progress in solving problems heretofore often deemed intractable.14

  A related development is the recognition that the linguistic meaning of (some) sentences is best thought of not as a single proposition that utterances of the sentence are typically used to assert, but as a set of constraints that determines a range of propositions that utterances in different circumstances may be taken as asserting. The simplest examples of this type have long been recognized. As the philosopher David Kaplan pointed out, to know the meanings of sentences containing indexicals—e.g., ‘I am hungry’ and ‘The meeting starts now’—is to know that one who uses the former asserts of oneself that one is hungry and that one who uses the latter at time t asserts of t that the meeting starts then—with different propositions asserted by different agents or at different times.15 But as a few examples will show, the point generalizes much further. The old idea that contents of our assertions—i.e., what we say or assert in uttering a sentence—are usually fully determined by the linguistic meanings of the sentences we assertively utter has given way to the idea that they are determined by the complex interaction of austere linguistic meanings with contextual information and contextually mandated reasoning.

  For example, sentences (2a–c) are grammatically but not semantically complete; (2a) requires the second argument of the finishing relation, (2b) requires an activity on which the semantic content of ‘ready’ must operate, and (2c) requires a reference point: nearby what—our present location, Bill’s present location, a location that he, or we, will be visiting next week? It all depends on context.

  2 a.  Bill is finished.

     b.  Bill is ready.

     c.  Bill is going to a nearby restaurant.

  When these sentences are used, the needed completion is sometimes provided by aspects of the context of utterance—e.g., the activity that Bill has been engaged in, the one he has been preparing for, or, in the case of (c), the location of the speaker, or the addressee. However, it can also be provided by activities and locations mentioned in the larger discourses of which utterances of (2a,b,c) are parts, or by the contents of shared presuppositions of speaker-hearers. Since there are indefinitely many possible completions of these utterances, this isn’t a matter of linguistic ambiguity (which arises from multiple linguistic conventions governing particular words). It is simply one way in which linguistic meanings can be underspecified, and so require contextual completion in order to determine what proposition is asserted.16

  Sentences containing bare numerical quantifiers N Fs, in which N is a numeral and F is a predicate nominal (two children, three dogs, four bicycles), are similar. Depending on the context of utterance, the quantifier can be interpreted as at least N Fs, exactly N Fs, at most N Fs, or up to N Fs. As before, this is not ambiguity; it is non-specificity. The meanings of these sentences leave open which of several possible completions may be dictated by the context.17

  A different, but related, kind of under-specification concerns sentences containing ordinary quantifiers—e.g., phrases of the form every/any/some/no so-and-so. The linguistic meanings of these phrases determine their use in talking about so-and-so’s, but the contributions these phrases make to what is asserted by uses of sentences containing them may be further restricted in specific contexts by the clearly discernible point of the speaker’s remark. For example, parents whose children are holding a sleepover in the basement might utter (3a,b,c) to make the italicized assertions, which don’t concern all people, or even all in the house, but merely the children downstairs.

  3

  a.

  Everyone is asleep.

  Everyone downstairs is asleep.

  b.

  Someone is lying on the floor.

  Someone downstairs is lying on the floor.

  c.

  No one wants to get up before 9 AM.

  No one downstairs wants to get up before 9 AM.

  Similarly, a football coach wishing to keep the opposing team from learning his strategy for the big game might use (4):

  4.  No one may, without my permission
, speak to any reporter.

  to say to his players that no team member may, without his permission, speak to any reporter about the upcoming game. This doesn’t restrict team members running for positions in student government from speaking to reporters about their candidacies.

  It is just this kind of linguistic phenomenon that we find in the compact clause of Article I, Section 10 of the U.S. Constitution, “[N]o State shall, without the Consent of Congress … enter into any Agreement or Compact with another State,” which asserts the more fully specified proposition that No state shall, without the consent of Congress, enter into any agreement or compact with another state that diminishes federal supremacy, or undermines the federalist structure of this Constitution. The fact that this is the proper interpretation of the compact clause—which doesn’t prohibit all agreements between states—isn’t news. What is news is that the interpretation is not, as prominent legal commentators have sometimes labeled it, a judge-made constitutional construction, which modifies original content of a clause in light of new facts or further judicial reflection prompted by litigation. Rather, the so-called construction is simply the mislabeled recognition of the original asserted content of the clause. The reason for the mislabeling is that the Supreme Court of the United States, when it ruled on this, didn’t realize that it was merely recognizing original asserted content because it didn’t have the sophisticated understanding of the relationship between linguistic meaning and asserted content that we do now.