Semantic Levels

Semantic Levels

Anyone interested in language and perception will recognize that there are different levels of analysis of the phenomenon in question. In linguistics we will distinguish phonetic, syntactic, and semantic levels (possibly others). In studies of visual perception we will distinguish the conscious percept, internal computations, the image on the retina, the proximal and distal stimulus, and the object of perception. If we compare the levels that concern the external object in the two areas, we notice a striking discrepancy: we speak of a single object of reference but several objects of perception. What did I refer to with “London”? I referred to a certain city and nothing else. What did I see when I flew over London? I saw the city, but I also saw a part of the city, the surface of the city, a facet of the city. In perception we readily speak of direct and indirect objects of perception—as when we say that I directly saw an elephant’s head but only indirectly saw the whole elephant, given that only the head was visible. We have the idea that you can see one thing in virtue of seeing another (one of its parts), as you can touch an object in virtue of touching a part of it (or the surface of part of it). Thus there are multiple perceptual objects in any visual encounter, but we suppose there to be only a single object in acts of reference. When I refer to London I don’t refer to its parts, surface, or facets. In the case of vision we might start out naively speaking of “seeing London”, but we quickly recognize that there is complexity here and that the objects of seeing are many—hence the talk of direct and indirect objects of perception. But we don’t think this way about reference: reference singles out a whole object not its parts of facets—hence there is no talk of direct and indirect reference analogous to the case of perception.

            We do suppose that the semantic level can be broken up into parts. In addition to the reference there is the sense, and sense itself can be broken into parts (character and content, say, or narrow and wide meaning). But we are not supposed to refer to sense—we express it. There is no dual reference, though the semantic level is composed of two sublevels. We might wonder whether the level of sense has all the complexity of the corresponding level in perception, i.e. the visual mode of presentation of the object. The latter divides into a central focal part and a peripheral part, but no one ever says that senses can be clearer at the center than at the periphery or that senses present many objects simultaneously. Senses seem simpler than percepts, as references seem simpler than perceptual objects. There is more structure in the one case than in the other. In particular, the semantic level concerned with reference to external objects is conceived as one-dimensional: we only refer to one thing at a time. Thus a theory of reference need only assign to words a single reference: the name “Saul Kripke” refers to Saul Kripke and to nothing else—not to his parts, surface, or facets. But the same is not true of seeing the man—here there are many objects of perception to be reckoned with. The obvious question is why the difference.

            Frege invoked the concept of an aspect in his classic discussion of sense and reference. The sense is said to contain an aspect–it presents or reveals or encodes an aspect. An aspect is an objective feature of the reference—something like the ensemble of properties apparent from a particular perspective. Objects can be seen from different perspectives and thus different aspects of the object are presented. This is what accounts for differences of sense (in central cases). A given object can present many aspects and sense can incorporate any of these. Now if we ask what relation a speaker has to an aspect, the answer will be that the aspect is presented to him, or perhaps that he apprehends the aspect. Maybe we can say that the name connotes an aspect—it somehow alludes to one. The aspect could be directly referred to, as in “the way the moon looks right now”: we can refer to the properties presented to us. The aspect belongs on the side of the world, along with the referent, not in the speaker’s mind. It is some kind of intentional object, in Brentano’s sense. So why not say that the aspect is referred to by the name? The name’s meaning identifies the aspect, alludes to it, specifies it—so doesn’t it refer to it? The sense of the name presents an object, but it also presents an aspect of the object; indeed, it presents the former by presenting the latter. I can be said to see an object and also to see an aspect of it, so why can’t I be said to refer to an object and also refer to an aspect of it? There is a kind of double denotation at work: the object and an aspect of the object are made objects of reference (intentionality). Even if the “folk theory” of reference doesn’t speak explicitly of this double denotation, closer analysis has revealed that there is more than just name and object; there is the aspect-presenting sense. So shouldn’t we revise the folk theory to take into account this further level of semantic structure? Wouldn’t that be good science?

            We can call the object itself the “secondary reference” and the aspect the “primary reference”. I choose “primary” for the aspect because it is in virtue of referring to the aspect that the object is referred to, just as in the case of perception. If we think of the aspect as captured by a definite description, then the aspect is clearly primary, since the description contains it en route to picking out an object. There is nothing to prevent us talking this way and it clarifies the structure of the referential semantic level. The aspect exists objectively alongside the object, not in the speaker’s mind, and our words (according to Frege) pick the aspect out; saying there is “reference” to the aspect is a small step. Thus “the Morning Star” denotes both Venus and the aspect of it presented in the morning. The proposition expressed by sentences containing the name will thus include both the object and the aspect. This allows us to explain the difference between “the Morning Star” and “the Evening Star” at the level of reference: different aspects referred to. A direct reference theory therefore permits a solution to Frege’s puzzle. No individual concepts need to be introduced or anything of a psychological nature; we just need to allow that reference can function like perception. We can invoke the apparatus of direct and indirect intentionality: I see an object by seeing an aspect of it, and I refer to an object by referring to an aspect of it. Whether we talk that way in our folk theory of reference is beside the point; the structure is there and needs to be articulated.  There may be possible perceivers who speak of perception in the simple way, as if there is nothing involved but the perceiver and the object, forgetting about perspective; but they would be wrong to insist that the object is the only thing that is perceived. We see objects by seeing aspects of them (surfaces of parts, roughly). It is the same with reference: there are nested levels of reference. Frege convinced us to accept that words can mean two kinds of thing—reference or sense—and now we should accept that words can refer to two kinds of thing—primary reference and secondary reference. This is scientific progress, though it may seem counterintuitive at first, as Frege’s theory of sense and reference did (e.g. to Russell).

            Suppose that a group of speakers came to the subject of semantics already equipped with Frege’s apparatus—they know all about aspects and objects. They refer explicitly to aspects all the time and are well aware of their role in determining the reference of names. They might introduce names by linking them to expressions that denote aspects: “Let ‘Hesperus’ be the name of the planet with that aspect”. They might even stipulate that “Hesperus” is to mean “the planet with that aspect”. Then the sense will include a specific reference to an aspect, so that we can readily speak of the primary and secondary reference of the name. These speakers have never entertained a single-reference folk semantic theory but have always allowed for double denotation. Shouldn’t we follow their example now that we have clearly discerned the semantic structure of object and aspect? Given Frege’s analysis, it is simply true that sentences containing names pick out both objects and aspects, and what point is there in denying that this “picking out” is the same as reference? When we speak of “the” denotation of a name we should really mean the pair of object and aspect. A theory of reference for names will accordingly assign both sorts of entity to a name. A Millian about names can agree with this double assignment, because the object itself enters the meaning of the name; but he can also solve Frege’s puzzle by appealing to the (implicit) reference to an aspect. We don’t need to bring in individual concepts or some such psychological thing; all the work can be done at the level of objective reference. It is just that we refer to more things than we realized. A name can have sense and references.

            The theory of reference therefore divides into two theories: object reference and aspect reference. How do these types of reference work? Should we have a causal theory for both, and how will the theories be related? Or should we have a description theory for both? We can see how aspects might figure in a description theory of object reference, but what about reference to aspects—is this mediated by description too? Can aspects have further aspects that figure in a theory of reference for them? Won’t that lead to an infinite regress? Is it possible that the basic theory will apply to aspect reference, with object reference carried by a description like “the object with aspect A”? Will the two theories be independent in the sense that neither determines the other, because different objects can instantiate the same aspect and different aspects can belong to the same object? Are aspects primarily referred to by means of demonstratives, so that object reference is carried by something of the form “the object with that aspect”? When a baby is baptized is the reference of the name fixed by “the human being with the babyish aspect now before us”? Should we say that we know aspects by acquaintance and objects only by description? What about the radical idea that all real reference is to aspects, with objects not strictly referred to at all? Frege spoke of the sense as “illuminating only a single aspect of the reference”, but should we infer that the sense doesn’t illuminate the reference as such? Is sense a conduit for aspects not objects, with the latter coming along for the ride? Is aspect reference where all the semantic action is? Maybe the object comes into the picture because we need it to account for truth conditions, but the basic semantic work goes into aspect reference. Senses tell us a lot about aspects, but relatively little about the objects that have the aspects. If we knew all about reference to aspects, what further task would the theory of reference have? All we would need to add is that the object of reference is simply the one that has the aspect already referred to. The theory of reference might be mainly a theory of aspect reference—as with the theory of perception. In any case, double denotation calls for a doubling of theory.

Colin McGinn