Life and Language: Strange Singularities

It is difficult to explain the origin of life on earth. Presumably, inorganic molecules at some point made the transition to organic molecules, allowing for self-replication. The state of the earth at some time and place must have been conducive to this event, while it wasn’t before and elsewhere. Even if life on earth was seeded by organic materials arriving on meteors, these seeds must themselves have had an inorganic origin somewhere else in the universe. This is admitted to be a difficult problem. But there is another problem in this neighborhood that is equally difficult: why did life on earth arise only once? Evidently, conditions on earth made life possible at a certain time, but why didn’t those same conditions lead to life at other times? We are told that life arose at a particular place and time, so that everything now living traces back to this singular origin, but why is this so—why didn’t it arise many times once the appropriate conditions obtained? Why the singularity? It is as if the switch went on once but then the earth forgot how to switch it on again. The events of the universe are mainly repetition, but in this case we have a one-off event—a completely unique occurrence. Once life arose from non-life, why didn’t it keep on so arising? The conditions were conducive, so why didn’t they conduce more often?

Suppose you travel to a distant planet and find life there. You investigate this life and determine that it arose n billion years ago: but it arose not just once but multiple times. It has kept on arising, perhaps millions of times, since the conditions for life persisted after the initial rising; there are thus a great many chains of living things that trace back to these many points of origin. There is nothing physically impossible about this; indeed, it is what you would expect given the conditions on this planet. So why isn’t earth more like that? Why the stinginess? Life on earth is all about repetition–organisms duplicate, cells divide into copies of themselves, the same kinds of biological events occur over and over—so why is the origin of life on earth a singularity? This fact cries out for explanation, but no explanation is forthcoming. Something special must have happened–but what? And how could anything that happened at life’s inception be so special? The singularity seems inexplicable, strange, mysterious. Imagine if life on earth had evolved three times, or seventeen times, but then stopped evolving: wouldn’t that be peculiar, calling for explanation? One time seems no less arbitrary, no less improbable. It might be more explicable if life had evolved by a huge cosmic accident—say, a fully formed pair of tigers, male and female, are created from inorganic materials by chance: that is not likely to happen again! But this isn’t how life on earth began: the process was incremental, with only bacteria at the early stages, preceded by organisms yet more simple. There is nothing intrinsically chancy about the earliest forms of life; yet they arose only once. It’s like a mountain arising only once or rain falling only once.[1]

I rehearse these points in order to draw an analogy I have not seen drawn: between life and language. The origins of language are notoriously problematic and subject to much controversy, but less often noted is the problem of singularity: why did language evolve only once? That is, why did language with the properties possessed by human language arise only once? I don’t mean communication systems in general, such as are possessed by whales, dolphins, bees, and ants; I mean the kind of grammatically complex systems possessed by humans. Clearly, the human language faculty has a tremendous adaptive advantage—some say it is the key to our dominance—so it must be a question why it has not evolved several times.[2] Other highly adaptive traits have evolved many times, as in so-called convergent evolution, but in the case of language this is not the case; we don’t find a plethora of species speaking a language like ours. In particular, we are the only hominid species with a language faculty: it is as if we were the only such species with eyes or ears. According to modern conceptions, the language faculty is an autonomous organ grounded in the genes, analogous to other organs; but it is an organ that appears only in our species.[3] Why? Its existence is a singularity, like life, but this is puzzling. One might expect that such a useful organ should have evolved many times, especially in our nearest relatives, but evidently not: we are its sole proud possessor. It arose at a certain late point of human evolution and only in us, but it has proven its worth a hundredfold, so why isn’t it more common. Like life, it is a strange singularity. You might suggest it is like the Mona Lisa—a sublime work of art, understandably rare—but that is very unconvincing: the Mona Lisa is not that unique among works of art, and is anyway a human product of creative genius. The language faculty, by contrast, is a biological organ evolving by mutation and natural selection just like other biological organs; so its occurrence should be governed by the same laws—hence its uniqueness is perplexing. On other planets there may be many speaking species—it may be the norm—but on our planet language is confined to a single species and arising at a specific time. If its origin is a mystery, so is its uniqueness. And if the mystery of origin is resolved, that will only intensify the mystery of uniqueness, since it will explain how natural processes of a non-mysterious sort account for the origin of language. An explanation of the origin of infinite recursive productivity, for example, will accentuate the question of why this property is not found elsewhere. It is as if no eyes ever evolved for billions of years, despite favorable conditions, and then a single species suddenly evolved them, never to appear again. Maybe language will evolve again in the distant future—as life may conceivably evolve again on planet earth—but heretofore we have a marked absence of both recurrences. Why is nature behaving so sparingly, so miserly? Life and language are now everywhere on the planet, but they refuse to evolve afresh as they once did. They seem determined to belong to an elite club of one. Why the exclusiveness?

It might be replied that appearances are misleading: life and language have come into existence many times, by chemical concatenation or genetic mutation, but they have not been selected for, and hence do not exist in full form in multiple cases. But there is no evidence for this, and anyway it just raises the original question in a revised form–namely, why has there been no natural selection for these common uprisings? Obviously there was selection in the case of the life and language we see around us, so why not for these other alleged fledging cases? Nor, of course, would it be remotely plausible to suggest that life and language have arisen many times and been selected for but we have just not noticed it: where might these elusive realities be hiding—at the center of the earth maybe?  No, both have arisen only once, puzzlingly so. We have here two “mysteries of singularity”: not why did they arise at all but why did they arise so sparsely. It almost seems as if there is some force preventing them from arising more than once. I have no idea how to answer this question and I doubt that anyone else does either, but the question is clearly worth asking. It reveals a serious limitation in our understanding of natural history.


[1] We might compare the origin of life with the origin of the universe. The big bang was a singularity too—a major event that has not been repeated. We haven’t had a series of big bangs (in our universe) since that initial one. Why? What explains the uniqueness? If there is a multiverse each with its own big bang, then we have a plurality of cosmic origins; but that doesn’t account for why our universe has only experienced one big bang. Is it because the nature of the universe was so changed by the big bang that the laws of nature ruled out a recurrence thereafter? That kind of explanation would clearly not apply to the origin of life—or of language (see below).

[2] I would speculate that consciousness (sentience, awareness) has arisen independently many times: it does not all trace back to a single evolutionary origin. Consciousness in humans and consciousness in the octopus are cases of convergent evolution. This is because consciousness is very widespread, unlike language. If consciousness were confined to a single species, that would be very surprising and cry out for explanation. Language is the outlier, not the conscious mind in general.

[3] It has struck many people as strange that more animals don’t speak—hence those fantasies about speaking animals. It is as if they lack an obvious ability, unaccountably granted to us. Speaking seems natural, only to be expected. What if language had evolved in monkeys instead of us? We might still have high general intelligence, while theirs might be inferior to ours (compare young children who can speak well but aren’t all that bright). That would seem very strange: if they have it, why don’t we? Isn’t it just anthropocentrism that makes us think that we alone are sophisticated enough to speak? What is really surprising is that monkeys don’t speak, given that there was nothing preventing them from evolving the capacity. Why should we be the chosen ones?


On Denoting and Describing

According to Russell’s theory of descriptions, the word “the” contains two conceptual elements: existence and uniqueness. It implies that there is a certain something, and that there is only one such thing. The indefinite article “a” contains only the first element; the definite article contributes the second. So the short word “the” is conceptually quite rich; it carries demanding implications. Through error these implications can fail, as when someone says, “The unicorn ate the grass” or “The teacher thinks he’s funny”: for there are no unicorns and there are many teachers. It is easy to make false statements (or neither true nor false statements) using definite descriptions because we can make errors of existence and also fail to specify a uniquely identifying predicate. You have to be on your toes with “the”. This insight of Russell’s is a point about the meaning of “the”, its semantics not its pragmatics. We can put it by saying that “the” is analyzable into “exists” and “unique”: “the F is G” means “an existent and uniquely F thing is G”. Strictly speaking, this claim about the meaning of “the” is limited to its conceptual composition and does not entail the standard conjunctive analysis deriving from Russell. You could agree with Russell about the meaning of “the” but dissent from his analysis of the logical form of sentences containing this word as comprising a threefold conjunction in first-order predicate logic. That analysis makes the definite description, apparently a referring expression, dissolve into quantifiers, and detects a hidden repeated occurrence of “and”. You might find that objectionable without disagreeing about the essential meaning of “the”: you might prefer not to express the semantic point in such conjunctive terms (you might favor a part-whole conception of semantic complexity). Similarly, you might be sympathetic to the idea that the meaning of “know” includes truth, belief, and justification without accepting that the logical form of sentences containing “know” is literally a conjunction—though that is certainly one way to express the semantic point. Thus we can divide Russell’s theory into two separate ideas, usually run together: the idea that the meaning of “the” involves existence and uniqueness, and the idea that the logical form of sentences containing “the” is a conjunction of quantified propositions. In either case, however, the word carries quite demanding and specific conditions as a matter of semantic analysis. It is certainly not redundant or dispensable.

The reason I am emphasizing what should be obvious is that Russell’s theory has a bearing on the question of the semantics of names and demonstratives. Suppose we conjoin Russell’s analysis with the description theory of names (or demonstratives)—and how could we not, given the cogency of that analysis? Then we have it that the meaning of a name is equivalent to a description having the analysis in question. Thus a name logically implies the existence and uniqueness of something satisfying a certain condition: for example, “London” implies that England has a capital and that it has only one capital, given that “London” means “the capital of England.” A speaker who uses the name “London” is asserting these two things, just as a speaker who explicitly uses the phrase “the capital of England” is asserting these two things. If he doesn’t think England has a capital, or thinks that it has more than one, then he has no business using the name “London”—for he doesn’t accept its logical implications. If you use the phrase “the capital of England”, you are committed to two propositions: that there exists something that is a capital of England, and that there is only one such. Your speech act fails if either condition fails to be met. But then, if this phrase gives the meaning of “London”, you make exactly the same commitments when you use “London”. There is, admittedly, no use of “the” in the employment of the name—no overt use, that is—but the theory is that this word lurks in the underlying meaning. The meaning of a name is a combination of the meaning of “the” and the meaning of a predicate attached to it. The question then is whether this is true: is that what names mean?

Critics of the description theory make much of the point that the speaker might be wrong about the predicates she ascribes to the bearer of a name she uses. You might, for example, be wrong to suppose that Godel proved the incompleteness of arithmetic, but you would still refer to Godel with “Godel”.[1] But you might also be wrong about existence and uniqueness: you might think a certain something exists when it doesn’t, or that there is only one thing of a certain kind and there isn’t. Suppose you believe that Sherlock Holmes was a real detective and bandy his name about under that misguided impression: you don’t hesitate to speak of the world’s greatest detective, Sherlock Holmes, living at 10 Baker Street. However, other members of your speech community correctly believe the description “the fictional detective created by Arthur Conan Doyle” to be true of Holmes. Don’t you still refer to the fictional character despite your false existence belief? The speech community’s reference carries the day. You might also be wrong about uniqueness in the case of Godel (maybe several people proved the incompleteness theorem, including him), but it wouldn’t affect your ability to refer using the name “Godel”. These beliefs are extraneous to your use of the name to refer to a certain individual. But they are not extraneous to the use of the associated description: you can’t use a description to successfully refer if its existence and uniqueness conditions fail. You can’t use “the aristocrat of France” to refer if there are no aristocrats in France or if there are many—you won’t succeed in referring to anything. This suggests the following thesis: names do not have the semantic implications of descriptions with respect to existence and uniqueness, and hence they cannot be analyzed by means of descriptions. We are accustomed to the idea that names lack connotation, i.e. they carry no descriptive or predicative content; the thesis to be considered is that they also lack the semantic content carried by “the”, as that word is analyzed by Russell.

It will help if I introduce some terminology: I will call a designator loaded if it expresses the kind of content attributed by Russell to definite descriptions; and I will call a designator desiccated if it does not express such content. Then my thesis is that only definite descriptions are loaded; names and demonstratives are not–they are desiccated. The intuitive point is that descriptions carry a rich and demanding semantic content—the existence and uniqueness of something of a certain kind—while names and demonstratives carry no such commitments (we could also call them “non-committal designators”). Consider uniqueness first. Suppose I point to a dog at the park and say, “that dog is lively”: I make no claims of uniqueness at all with respect to the predicate I use—and indeed there are many dogs milling around in front of me. I simply single out a particular dog by using the apparatus of demonstrative reference, in which context plays a vital role. I make no attempt to describe the dog in question uniquely. It is no objection to my speech act to say, “But there are lots of dogs in the park!” It is otherwise with “the dog whose owner is Bill Smith”—here we take on the burden of supplying individuating conditions. Similarly with names: if I say, “Bill Smith has a lively dog”, I don’t purport to provide individuating descriptive conditions for the bearer of the name I use—I may have no idea of such conditions. It is quite otherwise with “the inventor of bifocals” where I do purport to provide a uniquely identifying description (if more than one person, or no one at all, invented bifocals, I have misspoken). This designator is loaded, fully committed—while demonstratives and names are desiccated, non-committal about the properties of the designated object.

The question of existence is trickier because we normally do assume existence for the things we refer to in a desiccated manner. But it is noteworthy that there are many contexts in which this assumption is suspended. If I am under the influence of hallucinogenic drugs, I might remark, “That pink elephant is looking thoughtful”, knowing quite well that I am hallucinating said elephant. But I can’t say, “The pink elephant is looking thoughtful” without implying that it exists: for, according to Russell’s analysis, that would entail “There exists a pink elephant”. A brain in a vat could use demonstratives to refer to non-existent objects, but if it were to switch to definite descriptions it would incur the charge of making false statements (or neither true nor false). This is simply because demonstratives don’t literally say that they refer to existing entities, so semantically there is no contradiction between them and statements of non-existence. But descriptions do literally say that their reference exists, so there is a contradiction if we add, “However, there is no such thing as a pink elephant”.[2] Therefore it is not possible to analyze demonstratives by means of descriptions, as construed by Russell. They are too desiccated for that, too neutral, too non-committal. They don’t logically imply existence or uniqueness—not as a matter of their semantic content. But descriptions do, as is revealed in their explicitly containing the word “the”. That word is a quantifier, according to Russell, but there is no trace of it in names and demonstratives. This is why we are happy to use proper names for fictional characters—there is no logical implication of existence deriving from the semantics of the name. There is nothing that needs to be canceled in order not to be accused of error. But using descriptions for fictional entities invites criticism because it carries an implication of existence: you can’t just say “the detective who lives at 10 Baker Street” and expect to be taken to refer to a fictional entity. The word “the” actively quantifies, and so needs to have this implication canceled in some way, whereas the name “Sherlock Holmes” tells us nothing about the existential status of its bearer. Thus names can’t be analyzed by means of descriptions.

The intuition behind non-descriptive theories of names is that names are simply labels or tags with no internal semantic structure. Once we take on board Russell’s analysis of descriptions we see more clearly the kind of semantic structure that descriptions possess; in particular, we see that descriptions are not themselves name-like, i.e. labels or tags. They are loaded not desiccated. Descriptions contain a predicate, a quantifier, and a uniqueness operator. They are semantically intricate and referentially demanding. Names are not like this, which is why they can be used in the absence of an existence claim and without providing any individuating concept (ditto demonstratives). The description theory overstates their semantic commitments (or ambitions). It would be different if descriptions were simply denoting devices without significant internal semantic structure, but Russell’s theory shows us just how complex they really are—the word “the” packs a semantic punch. Thus it is not possible to analyze names and demonstratives by means of descriptions. This is an ironic result given that Russell himself held both a description theory of (ordinary) proper names and his three-clause analysis of descriptions, not seeing that the latter rules out the former. A natural alternative theory would anchor names in demonstratives not descriptions and reject a description theory of demonstratives (which looks pretty hopeless anyway). It is certainly very clear that a demonstrative doesn’t embed a uniquely identifying descriptive condition, relying instead on context to select a unique reference. We might indeed say that the whole point of demonstratives is to circumvent the need for individuating descriptions—instead we just point and say “that” (coupled with a suitable noun). And the reason a description must be loaded is that it can’t rely on context in this way: it needs to contain within itself the means and mechanism of reference.

The point I am making is distinct from the kind of point made by Kripke against the description theory: it is not a point about the rigidity of names or about the fallibility of our beliefs concerning the properties of the things we name. It is a point that only emerges once Russell’s theory of descriptions is properly absorbed: names would have to be equivalent to descriptions as so understood. According to that theory, a name would have to imply existence and uniqueness under a concept, but it is doubtful that this is the case with names—they make no such ambitious claims. The word “the”, as explicated by Russell, does not appear in their analysis. It is not just that names are not descriptive, i.e. predicative; they are also not in the business of asserting existence and uniqueness.[3]


[1] The example is from Saul Kripke’s Naming and Necessity, the locus classicus of anti-descriptivism about names.

[2] Quine’s doctrine of ontological commitment can be stated as the principle that only what you say there is do you say there is. Applied to the present question, this tells us that descriptions say that something exists (according to Russell) but names and demonstratives don’t say any such thing—though their use may conversationally imply or otherwise assume it. Descriptions are quantifiers, but names and demonstratives are not: hence the description theory is false.

[3] This is, of course, true of nearly all expressions: hardly any words contain an assertion of existence and uniqueness—predicates, connectives, prepositions, adverbs, etc. In fact, it looks as if only definite descriptions work this way. So it is not surprising that names and demonstratives (as well as indexical expressions generally) also fail to have this kind of semantic content. Descriptions are special.


Analysis of Analysis

The concept of analysis crops up in a variety of disciplines: we have chemical analysis, spectral analysis, psychoanalysis, linguistic analysis, anatomical analysis, analysis of political systems, market analysis, literary analysis, dream analysis, conceptual analysis, and so on. We can ask what all these types of analysis have in common, i.e. what the correct analysis of analysis is. The OED gives two definitions: “a detailed examination of something in order to interpret or explain it”, and “the process of separating something into its constituent elements”. Both definitions are evidently correct, the latter being more restrictive than the former. I would combine them into this: “a detailed examination of the constituent elements of something in order to interpret or explain it”. Such an examination is necessary because the constituent elements of something are not always (or often) evident or manifest: we need to work to reveal what they are. There is the surface appearance of the thing and there is its underlying constituent structure—analysis takes us from the former to the latter. It advances our knowledge by excavating what was hidden. It is informative. The paradigm is the analysis of light: light appears uniformly white under normal conditions but a prism can reveal it as consisting of a variety of different hues (as in a rainbow). We now know that this spectrum consists of varying wavelengths. There is not just a single unitary phenomenon here but a combination of distinguishable elements. Light may look simple but spectral analysis reveals it to be complex—a composite, a congeries. It consists of separable components. The same is true of a great many phenomena, which is why analysis is informative: things tend to be composed of more basic elements that combine in certain ways. Perhaps there are ultimate things that are simple and permit no analysis, but many things are susceptible to analysis. Accordingly, the different disciplines engage in analysis as part of their standard methodology. It would be odd if they didn’t. Isn’t this what the world expects of us? Certainly there is nothing suspect or illegitimate about the method of analysis.

It is worth bearing all this in mind when considering what is called conceptual analysis. For some reason, many philosophers have decided that conceptual analysis is not part of philosophical method.[1] They feel there is something misguided about it, perhaps unscientific. That would be very strange, given that analysis is part of virtually every other discipline, especially the sciences. Surely philosophy should expect that its subject matter is open to some sort of analysis—that the things it deals with exhibit constituent elements. For example, it is highly likely that knowledge consists of parts or aspects or properties in combination. The alternative is that it is a simple attribute or fact that admits of no analysis; and while that is not to be ruled out as a matter of logic, it seems pretty unlikely. We should at least attempt to provide an analysis of knowledge—as the physicist tries to provide an analysis of light or the chemist an analysis of water. It would be dogmatic to suppose that these things have no analysis, and they palpably do. Notice that I speak here of knowledge itself not of the concept of knowledge: it is the thing that has constituent structure (unless it is irreducibly simple) not the concept of it.[2] Of course, the concept too might have such structure, being made up of more elemental concepts, such as the concepts of belief and truth; but that is another question, given that we are interested in knowledge itself. You could believe in the analysis of knowledge but not believe in the analysis of the concept of knowledge, possibly because you don’t believe in concepts at all or believe that all concepts are simple. There is no contradiction in holding that knowledge is complex and analyzable while the concept of knowledge is simple and unanalyzable. Water is complex and analyzable even though the meaning of “water” may not be. There is the world on the one hand and our concepts of it on the other; and the twain might not meet, structurally speaking. In any case, it is entirely in line with other disciplines to expect that philosophy will include a substantial amount of analysis, either of things or concepts of things.

Sometimes people assume that the phrase “conceptual analysis” implies insulation from the world beyond concepts—that one who engages in conceptual analysis is concerned only with how we think of things not with things themselves, or worse with how we talk about things. They point out that chemistry and physics aren’t just concerned (or concerned at all) with concepts of chemical and physical things but with chemical and physical things themselves—that would make these subjects parts of psychology! Why then should philosophy concern itself only with our concepts? But that protest is confused—though the phrase “conceptual analysis” can certainly invite the confusion (which may be shared by those who applaud the activity of conceptual analysis). For the phrase is ambiguous: does it mean a method of analysis or does it mean an object of analysis? The former corresponds to the idea that we can analyze something X conceptually, i.e. by reference to our concepts of X—or as we might as well say, by a priori reflection. The latter is the idea that we can analyze concepts as such, whether a priori or a posteriori. So we might seek to analyze knowledge (the thing) by reflecting on our concept of knowledge, or we might analyze the concept of knowledge itself by whatever means we please (psychologically, neurologically). These are quite different ideas, one being a method, the other a subject matter. Thus there is no redundancy in the phrase “conceptual analysis of concepts”: this just means the a priori (conceptually based) analysis of concepts (a certain sort of mental entity). Likewise, there is no contradiction in the phrase “conceptual analysis of X”, where X is not a concept. And there is also “non-conceptual analysis of concepts”—the empirical investigation of concepts, such as might be undertaken by a cognitive psychologist. These would each be types of analysis, but employing different methods. One might naturally suppose that philosophers would focus on a priori analysis, given their interest in definition, though they might also take into account a posteriori types of analysis. In any case, they would be engaged on analysis, i.e. the discovery of constituent structure by means of detailed examination.

The only real alternative to this conception would be the view that nothing of interest to philosophers admits of analysis because everything of philosophical interest is simple. Not just concepts, but things—knowledge as well as the concept of knowledge. That seems highly implausible: why should philosophy alone concern itself with the logically simple? All the other disciplines deal in complex things that can be broken down into parts, so why should philosophy be any different? You might reply that philosophy is only interested in concepts, but (a) that is not true and (b) why should concepts alone be simple? Obviously there are complex concepts, typically expressed by complex phrases; and why shouldn’t simple words correspond to complex concepts, e.g. “knowledge”? Concepts don’t have to mirror words in their internal structure (especially words of the public language). So there is really no escaping analysis in philosophy: the goal of philosophical analysis is mandatory. Similarly, philosophy must be, in some departments at least, analytic philosophy—as chemistry must be analytic chemistry, literary studies analytic literary studies, etc. This simply reflects the fact that all disciplines deal in complex entities that can be broken into parts. Whether all philosophical analysis, properly so-called, is a priori analysis is another question, which I have not addressed (though I believe it is); what can’t be seriously denied is that philosophy is at least in part an analytical enterprise. There is room for something called “synthetic philosophy”, but there must also be a place for analytic philosophy, because philosophy discusses complex things with constituent structure. It may also identify unanalyzable elements as part of its analytical purpose, but it can’t avoid wholes and parts, because that is just how reality is constituted.[3]

How should the results of philosophical analysis be formulated? We have become accustomed to providing necessary and sufficient conditions, but that is not strictly entailed by the notion of analysis; and the former might carry commitments not integral to the latter. Thus we have become obsessed with the bogeyman of circularity. But a chemist doesn’t typically report his findings by saying things like, “x is water if and only if x contains hydrogen and x contains oxygen and there are two parts of hydrogen for every one part of oxygen”, and then anxiously waiting for any counterexamples to be produced. Rather, he says things like, “water is composed of hydrogen and oxygen in the ratio of two to one”. The epistemologist could mimic this mereological style by saying, “knowledge is composed of truth, belief, and justification”. Maybe it contains further ingredients (see Gettier) but at least it contains those ingredients. This is an interesting and informative statement, showing that philosophical analysis (of the a priori kind) can produce substantial results—and many other examples could be provided of informative philosophical analyses.[4] And it deemphasizes questions of circularity: it puts things in a way that is true to the spirit of the enterprise—the discovery of constituent elements. Non-circular definition becomes less central than illuminating constituent analysis. A literary scholar might analyze poems into stanzas, lines, phrases, and words, commenting on how the poem works, without undertaking to define the concept of a poem. Similarly for the atomic physicist, who might be hard-pressed to define the concept of matter. The philosopher can tell us how things are composed by the method of conceptual analysis without undertaking the arduous task of definition (though that is no doubt a worthy object). To change the example, we learn a lot about perception by discovering that perception has three constituent elements–a sense experience, a matching physical object, and a causal relation between the two—without troubling ourselves over whether this can be converted into a strict definition (which is difficult to do). Thus we could drop the habit of parading putative necessary and sufficient conditions and speak instead in mereological terms—the terms appropriate to the enterprise of analysis.

It is a question whether philosophy can do more than analyze, granted that it must at least analyze. I don’t propose to discuss this question here, except to observe that a broader conception of philosophy would inevitably call upon the results of analysis. Suppose we thought that philosophy also interprets and explains: it gives the meaning or significance of things and it produces explanatory theories. It doesn’t just analyze individual things but relates them together and elucidates their human importance. Thus it resembles other disciplines that also go beyond analysis—as chemistry and physics do in producing laws and theories, or as literary studies does. Philosophy can be theoretical and hermeneutic as well as analytic. Well and good: but these other dimensions are not unconnected to analysis—indeed, they feed off it. Analysis tells us what something is, its nature, its essence. There is no theoretical chemistry without analytical chemistry, and likewise there is no theoretical philosophy without analytic philosophy. An explanatory theory of knowledge, or an account of its human significance, needs an analysis of what knowledge consists in—of what constitutes it. Analysis is thus methodologically primary. So even if philosophy is not limited to analysis, it still depends upon it. Without conceptual analysis it is groping in the dark. Analytical philosophy must be the foundation of philosophy.


[1] I discuss this in detail in Truth by Analysis (2012); I won’t repeat that discussion now.

[2] Russell wrote two books entitled The Analysis of Mind and The Analysis of Matter, speaking of the things not the concepts. It was Ryle who introduced the analytic conceptual turn in The Concept of Mind. I self-consciously reversed the trend in The Character of Mind.

[3] A generalized primitivism might motivate resistance to analysis, but so might a generalized holism. This would be the idea that reality never divides into constituent parts but consists of “organic wholes” that go beyond their so-called constituents. Everything is an interconnected web that can never be constructed from more primitive constituents; the totality is the primary unit of reality. To break wholes into parts is always to falsify them—metaphysical holism rules. In fact, this type of metaphysics was what prompted by reaction the type of analytic philosophy favored by Russell and Moore. I take it that the opposition to analysis today is not motivated by this kind of metaphysics.

[4] Bernard Suits’ analysis of games is a noteworthy example: see The Grasshopper: Games, Life and Utopia (1978). I would also mention Grice’s work on meaning.


Experience Pluralism

We have five senses and five types of sensory experience. This is doubly contingent: we might have had fewer or more senses, and we might not have had a different phenomenological type of experience corresponding to each sense. The second claim is less obvious than the first, but evident on reflection. First, note that the relationship between stimulus-type and experience-type is contingent: the physical nature of the stimulus doesn’t entail the phenomenological nature of the perceptual response. Thus you can’t infer what visual experience is like from the physical nature of light or what auditory experience is like from the physical nature of sound waves (similarly for touch, smell, and taste). Nor can you infer the physics of the stimulus from the nature of the experience. There is a lawful correlation between stimulus and response, but there is no identity or metaphysically rigid relation between them. One could exist without the other. This lack of necessity underlies some familiar thought experiments: we can imagine rerouting the inputs from the ears into the visual cortex, so producing visual experiences from auditory stimuli, and vice versa. Or there could be beings initially set up to convert sound waves into visual experience and light into auditory experience. The stimulus contains information about the environment and the brain interprets this by using alternatives modes of phenomenological response. Isn’t this what the human senses already do to some extent? The same (distal) stimulus can be seen or touched or even heard, and smell and taste respond to the same molecular stimuli. There is also the phenomenon of synesthesia, in which the same stimulus produces a response in two sense modalities. How the brain codes sensory inputs is not dictated by the physical stimulus, distal or proximal; in principle, we could invert the relations that actually obtain. There are possible worlds in which light produces olfactory sensations and people taste visually.

But there is a thought experiment of this class that I have never seen mentioned: the idea that all the senses of a given creature might be served by the same phenomenological type.[1] For instance, our five senses might be centrally manifested in only visual experience—we only see things when stimuli impinge on the ears, skin, nose, and mouth. We reduce the phenomenological range to a single sensory type that is common to all the senses. For humans this set-up would require some major (futuristic) surgery, so let’s assume we are dealing with a Martian that is born this way. Symphonies are “heard” as complex patterns of shifting light, objects “feel” as they are seen, food “tastes” like a visual mosaic. For any type of input, there is just one mode of experiential response: instead of experience pluralism we have experience monism. Whether we should describe this situation as possessing a single sense that responds to a variety of stimuli or several senses that are mediated by a single type of experience is not critical to decide; what matters is that there is a leveling of phenomenology combined with the usual types of sensory impingements. The same variegated physical world is represented by a uniform type of phenomenal world. This seems like a logical possibility, not ruled out by the concepts or by some deep metaphysical necessity. Granted, we don’t find any actual instances of it on planet Earth, but there might be other planets that are home to beings like this.

This thought experiment raises interesting questions. First, is there some biological reason that we don’t find actual instances of creatures like this? On the face of it such a set-up is more parsimonious than the actual situation, and doesn’t nature prefer parsimony? The genes would only need to engineer a single type of central sensory nerve to handle input from all the senses—the visual type. This would serve as well, so why complicate the physiology? It may be retorted that representing all the senses in a single phenomenological type would be confusing for the organism, since it wouldn’t know whether it was tasting or seeing; but this could presumably be accommodated by assigning different visual types to the two sorts of stimulus. Isn’t this what we already do within the visual sense—as when we have distinct sensations for shape and color? Couldn’t the all-encompassing visual sense contain a reference to the part of the body being stimulated, so that it was clear what sense was being activated? Why should this be any more confusing than simultaneously receiving inputs from senses with different phenomenological character, since a central unit has to separate and integrate the inputs so received in this case too? The purely visual organism could be constructed so as to keep track of the origin of its visual experiences, in part by assigning different visual types to each type of input. Brighter colors might be assigned to one sense compared to another, or different colors entirely. Visual experience is already very various and dependent on varying aspects of the light stimulus, so there seems no problem of principle preventing a purely visual subject from existing (perhaps one that is perceptually simpler than us). More strongly, this might be a better way to increase sensory bandwidth: smell and taste might become more discriminating when mediated visually. To the objection that visual tasting wouldn’t have the motivational force of ordinary tasting, we could stipulate that gustatory visual sensations be genetically linked to the pleasure centers of the brain, so that certain visual arrays elicit pleasure in the hungry eater. Don’t some tastes become pleasurable to us that were once repugnant or bland? Why not have menus listing the particularly tasty color combinations on offer tonight? You bite into an oyster and your visual sensorium lights up with an accompanying rush of pleasure.

So parsimony recommends experience monism, but so do other aspects of the organism. Don’t we find a conspicuous absence of florid pluralism in the anatomy and physiology of the body? The bones are much the same in point of composition throughout the body, despite differences of function and structure—we don’t find different types of bone composition according to where the bone is located. What would be the point of that? It would just make ontogenesis more difficult. And the underlying physiology of the nervous system is likewise homogeneous: the nerves associated with the different senses are of basically the same type (a nucleus, axons, dendrites, and the same suite of chemical neurotransmitters); we don’t find radically different histological characteristics from sense to sense. Moreover, the distal stimulus is likewise uniform: the same physical world is present to each sense—consisting of atoms, forces, etc. But the sensory systems inject a marked heterogeneity into nature: they are more richly various from a phenomenological point of view than the external world or the physiology of the brain. They provide the pageantry and pizzazz. So we have a puzzle: why so much variety when parsimony and the general laws of nature recommend uniformity? Why make seeing so very different from hearing—or smelling so very different from touching? It seems like an act of generosity from nature to the experiencing organism—making life a little less boring and monotonous. But natural selection and the genes are not known for their generosity; they like things as simple as possible (such complexity as we find is forced on organisms by the rigors of survival). Our thoughts don’t exhibit as much phenomenological variety, no matter what their subject matter may be, so why do our senses insist on the gaudy plurality of our sensory experience? It seems surplus to requirements, a gratuitous gift, an unnecessary extravagance. What would you say if we had fifteen senses each equipped with its own distinctive phenomenology when far fewer would do just as well? That would seem like biological largesse above and beyond the cause of gene propagation; why not strip it down a bit? The natural thought is that the variety we experience must possess some hidden biological utility, but it is not clear what this utility is, given the informational powers of visual experience (or the other senses in their most advanced forms). The cell serves every biological purpose in the body, but it is fundamentally the same from organ to organ. To be sure, cells vary somewhat from heart to kidney, skin to brain, but no more than visual experiences differ among themselves. What we don’t find is organisms (or organs) made of completely different chemicals, or partly cell-based and partly continuous, or bones that are sometimes made of calcium and sometimes made of metal. We find variations on a theme: but sensory experience varies the theme. Seeing is really nothing like tasting. To lack a sense is to lack something sui generis, to miss out on something unique. A purely visual organism might go blind but still be replete with visual sensation; a blind man, however, can get at best hints of what vision might be like. Each type of sensory experience is, we might say, a world unto itself.

One possible view is that the present sensory set-up is temporary and the result of a dispensable holdover from earlier evolutionary times. The senses evolved separately as solutions to survival challenges and the senses that now populate the planet build upon these early forays (much the same is true of basic anatomy). This is not a matter of ideal optimality but of contingent evolutionary history. Conceivably, the process could have started with greater uniformity and stayed that way, or it might eventually work out the kinks and favor sensory homogeneity. If we were building sentient robots, we might be faced with a design decision—one type of central component that delivers only visual phenomenology or several types that afford sensory variety. The decision could affect future production whether or not we make the optimal decision. If reasons of economy favor the single-component approach, we might end up producing purely visual robots (though capable of responding to the full variety of physical inputs). This might correspond to life on other planets, depending on the actual course of evolutionary history. On our planet the earlier “decisions” favored distinct types of sensory experience, and thereafter organisms were stuck with them. This arrangement might be highly sub-optimal despite its universality in terrestrial life forms. If we imagine an early life form equipped only with visual sensations responsive to light, wondering how to expand into other stimulus fields, it would be intelligible if this form plumped for retention of its existing phenomenological capabilities extended to other types of stimulus. It could either devise new modes of sensory response to sound waves and other types of stimulus or stick with what it has onboard already. The latter choice might be preferable, given the engineering demands created by branching out. So we must not simply assume that experience pluralism is the biological ideal; it might just be an adventitious artifact of how evolution on earth has actually progressed. Aliens might view our mixed phenomenology as distinctly old-world, pre-self-technological, and recommend switching to a more streamlined approach (they promise it will not be boring compared to the cumbersome system we now employ). Or there might be hearing-obsessed aliens (of bat-like aspect) who urge the merits of their sensory world and disparage the purely visual species. After all, whoever said that we humans are biologically perfect? Surely pain is not the best possible way to cope with injury in every possible world, so why should sensory diversity be the best possible way to handle information in every possible world? Among the life forms of the universe it might be quite parochial. Certainly some life forms on earth manage quite well without the full panoply of the five human senses—bacteria, worms, and much marine life.

I will mention another possibility, if only for completeness. This is that our sensory phenomenology might be less various than we suppose. Obviously, introspection plays a determining role here—we experience ourselves as experientially plural. We seem to ourselves to contain phenomenological multitudes. But perhaps this appearance is misleading; perhaps we are more uniform than we think—as the external world is more uniform than we naively suppose given the way we experience it. From a more abstract or objective point of view we may be more uniform than we appear. We already accept that there are commonalities in perceptual experience—intentionality, spatial embedding, functionality—and it may be that there is a way of describing experience that will render it more unified than our current ways. A more objective phenomenology might be a more uniform phenomenology; there may be structural universals across sense modalities.[2] Synesthesia suggests as much. Just as science can reveal hidden universals, so a scientific phenomenology might reveal experiential universals beyond our current grasp. Then the variety of sense experience would be revealed as superficial. Chomsky sometimes suggests that there is really only one human language when you get right down to it, despite superficial appearances; well, is it ruled out that there might be just one type of human sense experience? Call this Universal Phenomenology (UP for short): just like Universal Grammar, Universal Phenomenology might unite all human experience and distinguish it from other possible types of sensory awareness (reptilian, Martian). If that were so, phenomenology might be as uniform as physiology at a deeper level. I don’t think we could ever conclude that really there is just the sense of vision, with every other sense a minor variation on it; but we might conclude that the deep structure of all sensory experience is common to every type—no more various than the cell types that correlate with experience in the brain. At any rate, this is a possibility to keep in mind, especially since otherwise we seem confronted by a genuine biological puzzle (the puzzle of excessive phenomenological variety, as we might call it).

Our language is hooked up to our senses, so that we can comment on what we see and hear (etc.), but we don’t have a separate language for each sense equipped with its own sound system, syntax, and semantics. That would be pointless and biologically redundant, as well as confusing and energy-consuming. So why do we have separate phenomenological systems hooked up to our senses instead of a single system? Why isn’t our sensory system more like our language system? The language system is a singular and separate module with its own distinctive internal structure; it is not divided into five different modules each with its own grammar and lexicon. Evidently, this kind of architecture could in principle characterize our sensory system—say, a single visual module hooked up to our several sense organs. Yet that is not what we find, but instead a diverse and divided set of systems that must all be integrated somehow. It seems unduly complicated and unwieldy, like speaking five languages when one would suffice. Why the difference? Why not speak a single phenomenological language?


[1] This thought experiment emerged during a conversation with Tom Nagel on October 10, 2019.

[2] Here we might be reminded of Nagel’s discussion of “objective phenomenology” in “What is it like to be a Bat?” The more a phenomenological description prescinds from the specifics of a given type of experience the more universal it is apt to be. Thus we might aspire to cross-modality phenomenology.


Explaining Mental Privacy

The privacy of the mind is generally treated as a platitude, but it is seldom (if ever) asked what explains this platitude. Privacy here is best understood perceptually: states of mind are not perceptible by means of the senses. It is not denied that they may be subjects of legitimate inference, or even of interpretative seeing, but they are not objects of perception in the way the body is. I can see your facial features in a way I can’t see your thoughts or feelings or sensations.[1] These are hidden from me, directly accessible only to you. If consciousness is a stream, it is an invisible stream. The senses are defeated by the mind; the two do not work well together. But why is this the case? What accounts for the invisibility of the mind?

First, let us develop a more articulate sense of the problem of privacy. Is the privacy of the mental a necessary truth? It is not an epistemic necessity: it could have turned out that the mind is not private. Picture the child’s inchoate understanding of the mind: it is by no means self-evident to her that the experiences she enjoys are perceptually inaccessible to others. She might be brainwashed into believing that others can see her thoughts (as we are brainwashed into believing that God can see them), and nothing in her understanding would enable her to rule out the truth of this proposition. It is not analytic or a priori that mental states are private. Nothing about how they appear to the child can logically exclude the possibility of perceptibility by others. This is something she must empirically discover. And likewise we adults cannot be certain that our minds are private—the claim is subject to Cartesian skepticism. I believe my mind is hidden from others, but perhaps I am wrong—perhaps I am surrounded by demigods that can literally see into my thoughts. I can’t be wrong that I have thoughts (the Cogito), but I may be wrong about their privacy: “I think, therefore something about me is invisible” is not a logical certainty.

So is the privacy point a contingent truth? It might not be, since epistemic contingency doesn’t entail metaphysical contingency, but for all we have said it might be. It might be that our senses, as they are actually constituted, don’t reveal other minds but that they could be altered in such a way as to overcome that limitation—or the senses of other beings might not be so limited. Or it may be that minds are necessarily inaccessible to our type of senses: nothing like vision, say, could ever reveal the mind of another—though some hard-to-imagine sense might not be so limited. The question is difficult, and it is compounded by an obvious fact: the mind is perceptible to its owner. I am aware of what I think and feel, though you are not. But then there is nothing about the mind itself that logically precludes perception of it; and if so, what obstacle is there to another subject having a similar kind of access? Couldn’t someone else have quasi-introspective access to my mind? It would be wrong to dismiss this point by insisting that introspection isn’t strictly a sense, so that its existence cannot be a reason to allow for a similar sense in others; for that would be equivalent to arguing that we can rule out the inner sense theory of introspection simply by observing that mental states are private. That would be an unconvincing argument, since privacy and the inner sense theory seem perfectly compatible—and indeed that theory has a lot to be said for it. So it is not at all obvious that privacy is a metaphysical necessity: there might well be possible worlds in which minds are perceptually accessible to others. Couldn’t there be reliable causal connections between states of mind and states of observers’ perceptual systems of such a kind that perceptions of the former by the latter were commonplace? Couldn’t there be possible beings that can see into the minds of others? At any rate, I will remain agnostic on the question here, since my question concerns the explanation of the privacy to which we are actually subject. Evidently this is an epistemic contingency, so it should be possible to explain why it holds. What, then, makes it the case that, in our world, minds are invisible?

It is important to appreciate how surprising this privacy really is, despite its familiarity. Not only is it not an a priori certainty; it runs counter to everything we know about the empirical world. It is hugely anomalous, quite bizarre, and very difficult to square with our general scientific picture of reality. Consider the matter biologically. Organisms have evolved a variety of traits useful to their survival. The visible body is the locus of those traits—a collection of functional organs than can be scrutinized with the naked eye (as well as under a microscope). Everything is open to view, public and perceptible. Well, almost everything, because among these evolved traits we have traits of mind—and these are private and imperceptible. This is true for us humans as well as for relatively simple organisms (it is not that the mental traits of reptiles are visible while mammals have evolved private mental traits). Why is this, and how is it possible? What is the point of evolving private mental traits—what does the privacy do for an organism in the way of survival? And how do the genes manufacture such traits? How is invisibility genetically coded? How does the nervous system, itself public and perceptible, generate inner mental states that resist observation? How does the private emerge from the public, both in phylogenesis and in ontogenesis? One would think that all the biological traits of organisms would be alike—all publicly visible—but some are the exact opposite. This cries out for explanation. What is the biological function of privacy, if any? What is it about the mind and the senses that renders the latter unsuitable for perceiving the former? What is the natural ground of privacy?

A first thought would be that it is in the very nature of the mind to be private: that is, we can derive privacy from other constitutive characteristics of mind. Thus we might consider whether intentionality or subjectivity or rationality lies behind privacy—is it because of the former that the latter holds? But this project is quickly dashed: privacy is a logically independent characteristic of mind. Nothing about those other properties of mind entails that mental states should be invisible. So there is no recognized intrinsic aspect of the mind that explains its privacy—nothing in our ordinary conception of it. In particular, nothing about consciousness as such (that we know of) leads inexorably to privacy; the mind could have been public and still had those other characteristics, so far as we can see.[2] Privacy is an add-on, a further trait of mind—as they are add-ons to each other. It begins to seem gratuitous that the mind should insist on invisibility—it lacks a clear rationale. Can we do better?

What about the power of deception? Organisms don’t always want their inner thoughts and intentions to be transparent, so there is survival value in keeping them hidden. Is this the reason the genes ensure that minds remain imperceptible? Hardly: the deception motive would only apply to a limited class of mental states; it wouldn’t apply to many organisms whose mental states are as private as those of professional deceivers; and there are also advantages to possessing public mental states, particularly where ease of communication is concerned.[3] So adaptive deception is not the ground of mental privacy. Nor are the usual physical explanations for invisibility applicable to the mind: optical transparency, camouflage, occlusion, size, speed, and glare. These factors can all result in a physical object being invisible, but they don’t apply to the mind. The mind is not a piece of glass or blends in with the environment or sits behind something or is too tiny or moves too quickly or gives off too much light.[4] And it is not as if it wears a magical cloak of invisibility that it can throw off to reveal its fleshly attributes. Its invisibility is more principled, more inherent. Or perhaps we should say that the limitations of perception with respect to the mind are more inherent to it. But this is puzzling for two reasons: first, mental states are inwardly perceptible, so it is not in their nature to resist all perceptual incursions; second, they have causal powers and exist in the natural world (unlike numbers), so there is no reason why they should preclude the evolution of a sense that can resonate to them.[5] In any case, we have not yet found a convincing explanation for such invisibility as we actually find.

An extravagant thought suggests itself: the reason the mind is invisible is that it is immaterial. We can only perceive material things in space, not immaterial things outside of space. So privacy entails dualism. We can deduce from the invisibility of the mind that the mind is not the body and is not dependent on it. There are a host of problems with this line of thought, which I won’t enumerate; let me make two quick points. The first is that it is not clear that immateriality entails invisibility (even assuming we know what “immateriality” means)—are angels and ghosts logically impossible? Is it our belief in immateriality that explains our commitment to privacy? Second, minds are not imperceptible tout court, since they are objects of introspection; and presumably this is compatible with dualism. Immateriality doesn’t preclude this kind of perception—so why the other kinds? Minds may be imperceptible in their very nature, but it is not because their nature is to be immaterial (whatever that might mean). Nor, to my knowledge, has any dualist attempted to prove that doctrine from the fact of privacy (Descartes was much too astute for that).

A more promising line of thought is that we are formulating the problem wrongly: we are presupposing that visibility is the norm and then fretting over why the mind doesn’t conform to the norm, but in fact, it is the other way about. Actually things are naturally invisible and what needs an explanation is why anything should be detectable by the senses. The question should be why the body is perceptually accessible. There is nothing surprising about the mind not being perceptible, since that is the default condition of the universe; what is surprising is local pockets of perceptibility—and even there the perceptibility may be glancing and superficial.[6] We can imagine a version of Kantian idealism behind this view, or reflections on current physics and “dark matter”. The thought is that reality is inherently removed from our senses and what we say we see is just our own mode of representing it—sense data, mental constructions, subjective phantasms. Then it is simply par for the course that minds are not perceptible—since nothing is! Reality is inherently not set up for the senses, or the senses are not set up for it.

This is a gratifyingly extreme position, but it won’t help us in the present connection, whatever may be said for it generally. For again, the fact is that mental states are perceptually available—to introspection. They are not naturally cut off from our epistemic faculties, like the most hidden of invisible particles or the constituents of dark matter—or even fields and forces. They are very proximate to us from an epistemic point of view. This is what is so puzzling about their third-person privacy: it is not that they are necessarily hidden to others as such that creates the puzzle but that they are selectively hidden. Why is that things that are so transparent from one point of view are so opaque from another?[7] This approaches the status of paradox, unlike generally hidden facets of the universe: so known and yet so unknown. It is as if God gave us one faculty with which to survey mental reality but refused to give us any more—introspection but not vision or hearing or touch or smell or taste. Mental states exist for the apprehension thereof, but the apprehension must only be from a single standpoint. Why not make the mind either universally unperceived or universally perceived? Why make us so acute in one way but so blind in another—and with respect to the very same objects?[8] Those who contend that reality at large eludes perception generally accept that the mind itself is open to immediate inspection—though they are insufficiently puzzled about its selective openness. We make no progress with our question by asserting that reality is generally cut off from our cognitive faculties, because that is signally not the case with the mind. Our explanandum is not that the mind is invisible tout court but that it is invisible to others while visible to oneself; or, as we might say, the fact that something as familiar as the mind is so removed from the world of the senses. One would think that it ought to be perfectly open to perception by others. Everything seems to point in that direction—introspection, biology, the general nature of reality—but for some reason, the mind refuses to yield itself up to the senses.[9]

Is it that the raw materials of mind are themselves imperceptible, so that evolution has no choice but to render the mind imperceptible? A panpsychist might contend that the proto-mental aspects of reality that form the basis of the mind are by nature imperceptible entities, more so than atoms and the like. So there is really no alternative to invisible minds—the properties of the parts transmit themselves to the whole. But this proposal also faces fatal objections. First, it can’t explain introspection: those primitive parts have to be capable of composing inwardly perceptible mental states–but how? Second, they raise the same question: why are they so radically imperceptible? Why does reality consist of both public and private objects? Could there be a sense that brought the proto-mental within its sights? The entities exist in nature, possess causal powers, and are aspects of public objects—and yet they are not perceptible, perhaps necessarily so. It is a puzzle; even a paradox, once we acknowledge introspection. Maybe there is an answer to the puzzle, but we have not yet found it.

What if the respiratory processes of the body were visible while the digestive processes were not? That would seem very odd. But why are the processes we designate as “mental” invisible while the rest are visible? Both are evolved biological traits arising from the same mechanisms and basic materials, so why the dramatic distinction? Suppose we were visited by aliens equipped with minds and senses that enable them to perceive each other’s mental states (perhaps ours too); to them, the idea of mental privacy would seem bizarre. They are surprised to learn that we are mentally blind (as they unkindly put it) and their philosopher-scientists get to work to explain this oddity in our makeup. To them there is nothing natural or inevitable about it, theoretically or practically; they read minds as they read texts. They have perceptual experiences of others being in pain and they think nothing of it. They think we are biological anomalies—they have seen nothing like us despite their extensive inter-galactic travels. Do we have some strange genetic defect? They refer to Earth as “the planet of the blind”, meaning mentally blind. They offer to engineer our deficit away, and are prepared to accept reasonable trading terms for this service (dancing lessons, massages). We accept their offer and wake up after the operation with vivid perceptions of other people’s states of mind—the phenomenology is extraordinary! You don’t feel other people’s pain exactly, but boy do you get a strong sense of it—nothing like that old insipid behavior-based belief that someone else is hurting. Your empathy really shoots up. You find yourself living in a whole new world. I am trying with this story to make strange the predicament in which we actually find ourselves: powerfully conscious of our own mind but only dimly aware of the minds of others. We stare at others and find nothing mental to look at, yet we know that behind the veil lies a mental life as rich as our own. How frustrating it is that we can’t penetrate the veil and observe other minds as they really are! Wouldn’t it be amazing if we could literally see the mental life of a bat? So near and yet so far! And we have no clear idea about why we are thus confined—no explanation of the fact of privacy. Why isn’t the opposite the case—the body is invisible but the mind is visible? Is that beyond God or nature to contrive?

There must be an explanation for privacy, despite its elusiveness, and whatever it is would surely shed considerable light on the nature of consciousness and the mind, as well as on the nature of our knowledge. But we haven’t been able to come up with one, not even an incomplete explanation. It is hard even to get one’s mind around the problem (what would it even be to see into the mind of another?): we have yet another mystery to add to the pile. We don’t know why our minds can’t be seen or heard or touched or smelled or tasted.


[1] I should say a word about those who believe that the mind is perceptible, as when we see a man’s suffering in his face. I don’t accept that type of description at face (!) value, but even if I did it would not affect the point of this paper, since it would be agreed that other minds are not perceivable except by virtue of expressive behavior. No one supposes that you can see the mind of another when there is no behavior to go on, say when the person is asleep and dreaming. The question then would be what explains this fact.

[2] The question is difficult but it doesn’t seem that consciousness itself is the root of privacy, its sine qua non, since the same point applies to the unconscious—it too is invisible.

[3] Would there be any need for speech if the mind were interpersonally transparent? Apparently not, so why isn’t there some selective pressure to evolve transparent minds, thus dispensing with the need for speech? Speech just seems like a rather cumbersome way to get your thoughts across.

[4] Note too that we cannot see the mind under ultraviolet light or take an X-ray of it (this inaccessibility applies equally to brain scanning machines).

[5] We could add that the close involvement of mind and brain also makes it surprising that the mind is invisible, since the brain is not. One would think that the mind would be as perceptible as the brain given their intimate connection. (To those who claim that the mind is perceptible because the mind is the brain, I make the obvious reply: even if it were true that there is de re perception of mental states in those states of the brain with which mental states are identical, it would not follow that the brain states are perceived as mental. That is, brain states do not appear to us as mental states.)

[6] We can easily envisage a form of materialism that accommodates the invisibility of the mental: just identify mental states with states of dark matter!

[7] Even if physical objects are not perceptually accessible, they are at least represented in perception—it seems to us as if we are seeing them. But that is not true of mental states—it does not seem to us that we see them. So there would still be an asymmetry between imperceptible physical objects and imperceptible mental states.

[8] Note that I can be experiencing a pain in my foot while being unable to perceive that very pain in myself: I introspect the pain in all its glory but no matter how hard I stare at my foot or my brain I can see nothing of its throbbing existence. I am perceptually cut off from a fact about myself that I know vividly from the inside. So the problem concerns not just my knowledge of other people’s sensations but also my knowledge of my own sensations.

[9] There is an analogous problem about introspection and the physical world: introspection is limited to the mental world, not extending to physical states of the organism. But why is this—why can’t we introspect our brain states, say? It doesn’t seem logically impossible and yet it never happens. I won’t discuss this problem here, limiting myself to the problem of why perception doesn’t extend to the mind.


What is Belief?



What is Belief?



For all the work that has been done on the topic of belief, do we really know what belief is?[1] What kind of state (if state it be) is the belief state? Two suggestions have been prominent: belief is a feeling and belief is a disposition. Either belief is a state of consciousness analogous to sensation (pain, seeing red, feeling sad) or it is a tendency to behave in a certain way (assenting to a proposition, combining with desire to produce action). The OED defines “believe” as “feel sure that (something) is true”, thus categorizing belief as a type of feeling: not “be sure” but “feelsure”. What that feeling might be is left undetermined, though the definition has the ring of truth. And indeed belief is connected to feeling: your feelings tend to change when you acquire a belief, and there is such a thing as feeling sure. But what about beliefs you hold without thinking about them– are those beliefs all associated with feelings? Do you feel sure that London is in England, for example, even when the thought has not crossed your mind in months? Here is where the dispositional theory suggests itself: belief isn’t an episodic state of consciousness but a readiness to act in a certain way—to respond “yes” when asked whether London is in England, say. Ramsey said belief is a “map by which we steer”, emphasizing that beliefs guide action (but do we inspect our beliefs as we inspect maps?). And certainly beliefs and dispositions are tightly connected (as are desires and dispositions): your dispositions change when you acquire a belief, and belief encourages assent behavior. But is this what a belief is? Isn’t it rather the mental state that gives rise to the disposition? What if you had a tendency to assent verbally to propositions not because you believe them but because you have been rigged up that way by a clever scientist intent on simulating the state of belief? In general, dispositional theories confound properties (states, facts) with their causal consequences; and we want to know what belief is not what it does. The OED also has this under “believe”: “accept the statement (of someone) as true”. But don’t we accept statements because of what we believe? It isn’t that the belief is the acceptance. It is hard to avoid the impression that the dictionary (and the usual philosophical theories) conflates the symptoms of belief—feelings and dispositions–with belief itself. But then what is belief itself exactly?

Are we acquainted with belief itself? We are acquainted with sensations and behavior, both signs of belief, but are we acquainted with beliefs? The answer is not obvious. If we are, it seems curious that we draw a blank when considering the nature of belief; but if we are not, why do we bandy the concept around with such confidence? Is it perhaps that the concept is logically primitive and hence admits of no explanation in other terms? But that can’t be the reason for our ignorance, because the same is true of many concepts and yet we are not blind to the nature of their reference (pain, seeing red, maybe moral goodness). Or is it that the felt ignorance is an illusion born of a mistaken assumption, namely that we only know what a mental phenomenon is if we can reduce it either to a feeling or to a disposition? Maybe we know exactly what belief is but we think we don’t because beliefs are not sensational or behavioral, these being our preferred touchstones of mental reality when thinking philosophically. But that approach, though not unsound in principle, is hard to square with an evident fact: we really don’t know what it is to believe something—we have no conception of what fact is at issue. Once belief is distinguished from its symptoms its elusiveness becomes evident (compare Hume on causation).

This leaves us with another possibility—that “believes” is really a name for an I-know-not-what that we introduce to denote something that we reasonably believe to exist but can’t properly conceptualize. Belief is thus that state, whatever it is, that has such and such symptoms and plays such and such a role but whose nature we find elusive. In short, “belief” is a theoretical term—not just in application to others but also in application to oneself. Our knowledge of belief operates at one remove from the thing itself, which is why we have such an indeterminate conception of it. A similar approach might be suggested for the concepts of meaning and the self: these too are not directly encountered constituents of consciousness, which is why we can’t reconstruct them in such terms, but they are real nonetheless, just at some epistemic distance from our cognitive faculties. That is, not all parts of what we think of as the mind exist at the same epistemic level (and not because of a detached Freudian unconscious); some are not objects of direct inspection (perceptual or introspective). The ontology of folk psychology is an amalgam of these two types of fact (and we can add desire to belief): the mind consists of directly known constituents and relatively unknown constituents. Differently stated, belief (desire, meaning, the self) is a state that we refer to but are not acquainted with; we know many of its properties, but not its intrinsic nature. We know it is a propositional attitude (but what is an attitude exactly?) and that it involves the exercise of concepts, as well as being a truth-bearer, subject to referential opacity, and capable of combining with desire to lead to action: but we don’t grasp what kind of state it is—not in its intrinsic nature. The state gives rise to inner feelings and to outer behavior, but we have no clear idea of what it is in itself. We experience shadows of it, fleeting intimations and glimpses, but we have no firm conception of the thing itself: it is just “that which gives rise to these symptoms”. Ask yourself what kind of mental state you are in when you are asleep: you have various beliefs, but what is their mode of existence exactly? You might be tempted to reach for the concept of a disposition, but we have been down that road before—what is the ground of such a disposition? Let’s face it: you don’t know what to say, and yet you don’t doubt that you are in some sort of mental state. You might sputter that you are in a “cognitive state”, but that raises the same question over again: what kind of state is that? Not a feeling state and not a disposition, but a sui generis state that confounds comprehension. As we might say, we have only a partial grasp of what belief is. And the part we don’t grasp intrigues us the most, i.e. the very being of belief.

I grant that this position might sound counterintuitive. Doesn’t the Cogito express certain knowledge (“I believe, therefore I am”)? But how can that be if we don’t know what thinking (believing) is? However, this is really not such a paradoxical position to be in: we know that we think and believe, and that this entails our existence, but it doesn’t follow that we know what thinking and believing are—or what the self is for that matter. And did Descartes ever claim anything to the contrary—did he suppose that the nature of thinking is totally transparent to us? Knowing that something exists is not the same as knowing its nature. If Descartes had claimed that thinking is processing sentences in the language of thought, he could have been wrong about that; but this wouldn’t undermine the Cogito. In fact, I would say that if you focus really hard on what is going on when you believe something you will see that nothing determinate comes into view—you never catch your belief in flagrante, as it were. And you have no clear conception of what it is that you attribute when you ascribe beliefs to others (beyond their conceptual content). Nor does knowledge of the brain help: identifying belief with neural excitation in the B-fibers, say, affords no knowledge of what belief is in the ordinary sense. The problem is that neither does anything else—crucially, not introspection. We didn’t come by the concept of belief by noticing feelings of belief in ourselves (where would those feelings to be located?), or by observing the operation of dispositions to behavior; rather, we introduced a term for a type of psychological state whose nature was not evident to us but which we were sure existed. I have evidence for my beliefs drawn from my experience (e.g. feelings of conviction), but I don’t believe in beliefs because I can grasp them whole. I see them through a glass darkly. I have a nebulous sense that certain propositions attract my assent, as if gravitationally, but what exactly my mind is up to I cannot tell. Even the strongest of our beliefs, say religious or moral or scientific beliefs, fail to disclose their inner nature—we just find ourselves filled with passionate conviction about certain things. It isn’t like feeling a headache or a hunger pang in the stomach. Nor is it like hearing a sentence in your head. It isn’t like anything.

Psychology used to be conceived as an introspective science, and then later as a science of observable behavior, but these ideas were predicated on a certain conception of the essence of the mind. Either the mind consists of inner episodes of consciousness of which we have immediate introspective awareness, or it consists of outer behavior that can be perceived externally. But the case of belief (also desire) shows that these alternatives are not exhaustive and are fundamentally on the wrong track. In so far as psychology is about belief and kindred states, it is not about feelings or behavioral dispositions, but about facts we find systematically elusive, which fit into neither category. Beliefs are not feelings and they are not dispositions to behavior, yet there are fully mental phenomena, paradigmatically so. As Hume would say, we have no impression of belief, yet belief is real and knowable (in some of its aspects).  Belief is yet another example of the limits of human cognition. Psychology thus has an elusive subject matter.[2]



[1] The background to this essay is scattered. The issues discussed bubble under the surface of Wittgenstein’s Philosophical Investigations and are explicitly posed in Kripke’s Wittgenstein on Rules and Private Language (as well as my Wittgenstein on Meaning). In addition, the emphasis on ignorance reflects my standing interest in human mysteries as they pertain to philosophy. Hume is hovering paternally in the wings. Russell makes a brief appearance.

[2] It might be said that belief is a computational state and that this gives its essential nature. There is a lot to be said about this suggestion; suffice it to remark that this doesn’t give us a conception of belief comparable to our intuitive notions of pain or seeing red. Belief may well have computational properties, but it is another thing to claim that this is what belief is (would it follow that computers believe?).