Jane Austen on Memory


Memory Illusions



In Jane Austen’s Mansfield Park, Fanny Price, a thoughtful and unassuming young woman, makes the following observations to a certain Miss Crawford: “If any one faculty of our nature may be called morewonderful than the rest, I do believe it is memory. There seems something more speakingly incomprehensible in the powers, the failures, the inequalities of memory, than in any other of our intelligences. The memory is sometimes so retentive, so serviceable, so obedient—at others, so bewildered and so weak—and at others again, so tyrannic, so beyond control!—We are to be sure a miracle every way—but our powers of recollecting and forgetting, do seem peculiarly past finding out.” (538) The author then reports Miss Crawford’s reaction: “Miss Crawford untouched and inattentive, had nothing to say; and Fanny perceiving it, brought back her own mind to what she thought must interest.” (539) Here we see Jane Austen at her most philosophical and, gratifyingly, mysterian. She clearly finds memory fascinating, but also “peculiarly past finding out”. She also realizes that such questions are not for the shallow of mind. The philosophical reader duly warms to Miss Price, who is already very much in our good books on account of her virtue and modesty (in her inventor we see yet another sign of genius).

Austen is comparing memory to our other “intelligences”, presumably including perception and thought. Her observation is that memory is more arbitraryin its powers than they are. To be sure, perception and thought have their failures and breakdowns—they are subject to error—but these weaknesses are fairly regular and predictable, while the powers of memory vary for no discernible reason. Sometimes we cannot for the life of us remember what we want to, but at other times memory pursues us relentlessly, refusing to relinquish its contents no matter how we may feel about them. Perception is not subject to the will, while thought is, but memory is partiallysubject to the will—sometimes under our control and sometimes very much not. The reasons for this are obscure: why do we so vividly remember some things, often seemingly insignificant, while others slip too easily from memory? I cannot ever choose what to see, while I can always choose what to think about, but I have partial choice when it comes to remembering things. Thus memory seems poised somewhere between perception and thought. It is quasi-perceptual and quasi-cognitive.

Then there is the question of its scope and limits. Perception is limited to the present moment and the impinging environment, though it handles an immense amount of information simultaneously. Memory has a broader scope, taking in large tracts of the past and being relatively independent of time of occurrence (you can often remember your childhood more vividly than last year). Perceptions rapidly come and go, while memories can linger indefinitely. Thought ranges more widely still, taking in the future as well the past and present, and including things not perceptible at all (you can think about atoms but you can’t remember what they did, not directly anyway). So memory falls between these two poles. My question is whether it is subject to illusions. There are clearly perceptual illusions, while thought is not vulnerable to illusions (though error is commonplace), but what about memory? Does memory sometimes give rise to illusions of the past? Errors, certainly, but are there also actual illusions? I don’t mean memories of past perceptual illusions, like remembering that Muller-Lyer illusion you saw yesterday; I mean specifically memory illusions, where a memory impression misrepresents a past event. If so, what are these illusions, what types do they fall into, what laws govern them? Are they analogous to perceptual illusions or are they sui generis? The question is not an easy one, despite its simplicity. On the one hand, memory has a sensory, particularly visual, dimension: we have sensory image-like impressions of the past. These can be inaccurate, as when one remembers a face with the wrong color eyes or a page with a word misplaced. It is tempting to describe this phenomenon as a visual illusion of the past not merely as a false belief. One can also have a complete hallucination of the past, as when one appears to recall a past event in full vividness that simply didn’t happen. These seem rather like mirage cases or seeing two equal lines as different in length: a sensory misrepresentation of the world, but relating to the past state of the world. The memory system has delivered up a mental representation that fails to fit the “stimulus”. False beliefs may be formed, as in the perceptual case.

Yet, on the other hand, what we don’t find are reliable, predictable illusions generated by certain types of configurations of objects—as in the moon illusion or the Muller-Lyer illusion. It is not that whenever one remembers the moon it always seems larger than it really is, or that memories of adjacent parallel lines always present them as unequal in length. Memory “illusions” are haphazard, unsystematic, and unrelated to the properties of the stimulus; they are more like errors of belief in this respect. There are apparently no laws of memory illusion analogous to the laws of perceptual illusion (faces with brown eyes don’t always produce memories of blue eyes).[1] Moreover, memory errors are correctable in the light of contradictory knowledge, unlike perceptual errors, which are incorrigible in the light of true belief.[2]Thus the moon illusion and the Muller-Lyer illusion persist even when one knows quite well that the facts are otherwise than they appear, but memory impressions are permeable by extraneous knowledge—you will not keep on remembering things a certain way once you have been enlightened (at least there is the possibility of a change of impression, unlike in the perceptual case). This makes sense, given that memory lies somewhere between the perceptual systems and the central cognitive system. Memory thus seems vulnerable to something likeperceptual illusion but also unlikeit. It is neither fully one thing nor the other (as Jane Austen intimates).

I think, then, that there is no clear answer to my question, because memory provides a counterexample to the dichotomy between perceptual and cognitive error. Perhaps we can say that it gives rise to quasi-illusions, where the qualifier simply indicates being betwixt and between. In other terminology, memory is neither an encapsulated module nor a general-purpose ratiocinator. On balance, I would say that it is not susceptible to illusion in the strict sense, but that it does give rise to sensory misrepresentation as a matter of course. In fact, it is moreprone to sensory error than the senses themselves, being less governed by psychophysical laws; but these errors are not as rooted in the architecture of the system as perceptual illusions. The senses systematically act in ways that can defy our reasoned view of things, as autonomous informational agents; but memory is not so cut off from reason, not so independent of cognition in its operations. It is puzzling and counterintuitive, quite exceptional in our psychological economy; whether it is “peculiarly past finding out” is another question, but not one to dismiss. It is noteworthy how little philosophical reflection has been devoted to it in comparison with perception and thought.[3]As Miss Price remarks, no doubt we are all “miracles”, but memory seems especially impenetrable even in its most quotidian operations.



[1]The closest thing I can think of concerns memory impressions of elapsed time. We remember a busy time as passing quickly while a boring time is remembered as passing slowly, and this seems somewhat lawlike. But even here the illusion has its source in the contemporaneous perception of time in the two cases not in an inherent tendency to illusion in memory as such.

[2]See Jerry Fodor, The Modularity of Mind.

[3]In contrast memory has been a staple of psychology since its inception, with much experimentation and theory devoted to it. I don’t recall ever hearing the question of memory illusions discussed, though failures of memory are routinely studied. Psychologists recognize how perplexing memory is, though I have not heard of one who takes Austen’s mysterian line.


Semantics Politicized


Multi-Dimensional (Inclusive) Semantics



I address you today in a spirit of inclusiveness and diversity. For too long semantics (theory of meaning) has been the confine of a single type of entity held to constitute all that meaning encompasses (or a couple of entities, closely related).  We must broaden our horizons and recognize that many kinds of entity contribute to the overall significance of an expression, often emanating from different traditions and regions. Above all it is referencethat has proved hegemonic, squeezing out other contenders for semantic acceptance. Whether that notion is phallocentric to boot I shall not venture to say[1]; what I shall say is that we need a far more inclusive and diversity-driven approach to semantics. Semantics correctly conceived is a rainbow.[2]

It used to be that only reference (denotation) was admitted into the semantic club: the meaning of an expression was its denotation. This was the view of Lord Bertrand Russell, English aristocrat and logic whiz (Western logic). Definite descriptions had to be distorted beyond recognition in order to fit them into this narrow picture (a form of linguistic colonialism perhaps). In any case, this approach, hailing from John Stuart Mill, another privileged upper class Englishman (and we duly note the gender) held sway until a rebellious German, a certain Gottlob Frege, added an extra element to the story—what he provocatively labeled sense. This was an improvement, breaking the stranglehold of the English referential aristocrats, but sense was conceived as the mode of presentation of the reference; so reference was still occupying center stage, with sense acting merely as its reflection or image, i.e. how we viewreference. (Can we say that while reference is the phallus sense is its codpiece?) Still, the basic monism is firmly in place: semantics remains one-dimensional, or at least one-and-a-half-dimensional. Not till Ludwig Wittgenstein arrived (also a white male aristocrat) was this monism seriously questioned and a certain kind of pluralism put in its place—with all the variety of language emphasized and celebrated. This was a welcome development in the openness of semantic studies, even allowing for the existence of actual workingmen (those builders of the early Philosophical Investigations—though again we must note the gender bias). But instead of embracing diversity the Austrian aristocrat insisted on imposing a new one-dimensional hegemony—all meaning is use. Reference drops out of the picture entirely, as if use has ousted it altogether. We don’t have use andreference but use and notreference. The old exclusiveness survives in a new form, less rigid perhaps, but with the same drive towards uniformity. One half expects the use to be restricted to only the most privileged of users! This entire trajectory then reaches its climax, i.e. nadir, in the person of Sir Michael Dummett, a white male Oxford philosopher, whose main mantra is that everything about meaning should be explained by one central concept—such as truth or verification. There could not be a more blatant hegemony! Nothing is to be included in meaning except what can be subsumed under a single conceptual category: you are welcome to join the semantic club, but only if you are properly related to the concept of truth (or verification). No diversity allowed!

At this point I shall drop the political backstory and proceed immediately to theoretical matters, though I trust my enlightened readers to keep that political context always in mind. And let me lay my cards on the table right away: I am all in for maximum semantic inclusiveness with as much diversity as possible (within reason of course). Not just two-dimensional semantics, or even three or four, but manydimensions, indefinitely many—as many as we can come up with. Fortunately, we have this diversity already lying around—it requires no strenuous inventing on our part. I have prepared a long list: reference, sense (mode of presentation), tone, character and content, intension and extension, grammatical mood, inferential role, rules, stereotype, mental image, individual and social understanding, ideas, brain states, use, conceptual analysis, truth conditions, criteria, causal chains, and whatever else comes to mind. For my contention is that allof these may be reckoned to the meaning of a word or sentence: not one of them and notthe others, but the whole lot. They don’t exclude each other but coexist peacefully. For example, a proper name, say “Aristotle”, has reference, sense, an intension and extension, a character (constant in all contexts), a role in inference, an associated stereotype (“bearded cogitating Greek man”), individual grasp and socially agreed grasp, a use, a contribution to truth conditions, criteria of application (see stereotype), a causal-historical chain, even a tone (vaguely distinguished and admirable). From among this variegated list we may pick out sense and intension for instructive contrast: the former is defined in epistemic terms (mode of presentation and interchangeability in belief contexts) while the latter is defined in modal terms (functions from worlds to extensions). These are by no means the same notion, but they equally belong to a single name, existing side by side in perfect harmony. There is no point in arguing that one is the realmeaning and the other a mere impostor: both belong to the overall semantic significance of the name. Both are attributes the name has, and they clearly flow from what it means (not what it sounds like). Meaning is multi-dimensional, diverse, and inclusive. No doubt there are interesting relations of dependency between these various elements, which may be studied, but the plurality is irreducible—part of meaning’s rich pageant. We can even throw in some Meinong-style ontology if that is to our taste, assigning to so-called empty names a subsistent entity as reference, or what is called an ”intentional object”. A committed Kantian might insist that reference be divided into phenomenal reference and noumenal reference. A follower of Sir Arthur Eddington might propose a double reference for “chair”: the commonsense chair and the chair of physics. The possibilities are endless, to be considered on their merits; but they should not be rejected simply because of some presumed one-dimensionality in meaning. In the theory of meaning our adage should be, “The more the merrier”. Plurality is a sign that we have not omitted anything not a symptom of conceptual chaos or indecision.

It may be remarked that the situation in other departments of linguistic theory is already happily pluralist. Consider the theory of syntax, taken to include the study of the sound system of a language. There is no one central concept here to which others must bow down; instead there are layers and dimensions. We can study speech as an acoustic phenomenon (as with a speech spectrograph), or as an articulatory system, or as embodied in the brain, or computationally. None of these competes with the others; all are legitimate and important. Syntax more narrowly conceived is typically understood as consisting of layers of rules, which may be viewed computationally or in terms of brain mechanisms. These are all aspects of the “formal” properties of language, and they all coexist—people don’t go around complaining that someone else’s pet theory isn’treallyabout syntax. Syntax isn’t one-dimensional. Similarly, in pragmatics there is room for a diversity of perspectives—not a single overarching concept. Thus there is no inconsistency between Gricean, Austinian, and Wittgensteinian approaches to (philosophical) pragmatics: all can be true and illuminating in their different ways. After all, there are many aspects to the employment of language by people, and we should not expect to be able to subsume all them neatly under a single heading. For example, an utterance of “Shut the door!” may be made with Gricean intentions, while having an Austinian perlocutionary effect, and occurring within a Wittgensteinian language game. Then too, we may approach pragmatics from an individual’s perspective, studying the way language is used as a tool of thought (say), or we can approach it socially, studying how language is used in interpersonal communication. There are indefinitely many possible ways to do pragmatics, as there are multiple ways to do syntax; and there is no reason semantics should be an exception. There are multiple components across the board. The fact is that the list of concepts I gave represents a variety of insights into meaning on the part of different thinkers, each valuable in its own way, and there no necessity to reject some in favor of others. I don’t mean to say that no semantic theory can conceivably be false, just that the fault is usually incompleteness not outright error. Apparent inconsistencies often melt under more tolerant investigation (as with Fregean versus Kaplanian approaches to indexicals). I used to be all in favor of “dual component” semantics, but really we should expand the dimensions dramatically to accommodate everything that characterizes meaning. The concept of meaning is a multi-dimensional concept incorporating a large variety of factors. It is not a simple thing like being square or red; it is more like the concepts of democracy or marriage or success. It contains multitudes.

Let me return to my political platform, because I was not being entirely frivolous (though mainly so). In ethics there has historically been a tendency towards monolithic theories, as with utilitarianism and Kantian ethics. It was left to more ecumenical ethicists like W.D. Ross to advocate a pluralist reconciliation between these apparently competing systems, thus producing a multi-dimensional ethical system. It is easy to see this development as an integration of different political perspectives—the pure will of the privileged autonomous agent versus the maximization of happiness in a suffering population. In the case of semantics we also have a politically contested domain, because language is spoken by diverse groups of people each with their purposes, positions, and ways of life. It would not be amazing if a certain kind of linguistic hegemony were in effect according to which only certain aspects of meaning are deemed “proper”, the rest consigned to illegitimacy and disdain. Hence we get the idea of the logically perfect language. The messy reality of meaning might not receive its due recognition because of an ingrained habit of favoring some things over others. There is always something evaluative in theories of meaning, as if only a certain dimension is deserving of respect. Why has tone not received the attention it deserves? Could it be that its prime examples are racial slurs and sexist language? Why would people want to explore the expression of their own prejudices and hostilities? Speaking very broadly, there is something democraticabout meaning: everyone speaks no matter his or her social class or place in society, and meaning itself combines disparate elements jostling together. Oversimplifying culture from political motives is not so far removed from oversimplifying language from similar motives. The habit of exclusivity is deeply rooted and ubiquitous. At the least it can operate as a factor in determining what theoretical options people tend to take seriously. Semantics is political too.[3]


Colin McGinn

[1]I have no wish to wax psychoanalytic, but isn’t the notion of reference suspiciously phallic (at least as phallic as some of Freud’s phallic symbols)? It seems to involve a kind of mental protrusion, as the act of reference extends outward to make contact with objects in the environment. People sometimes talk of reference as like tentacles reaching out to grasp, but other organs of the body can reach out and make contact too. And what about pointing? The pointing finger has a rigidity and angle not unlike… And then there is “rigid designation”, a phrase that trips suspiciously easily off the tongue. Just saying.

[2]Light can appear homogeneous, but the rainbow resolves it into an array of separable hues. Meaning can seem homogeneous too until we resolve it into its components.

[3]For all I know intellectual traditions from beyond the West have suggested aspects of meaning Western thinkers have missed. If so, I cordially invite them in.


Thoughts and Things



External Conditions of Thought



The idea of the singular proposition is that propositions can contain particulars as well as universals as their constituents. If I think that that bird is pretty, my thought’s content contains both a particular bird and the general property of being pretty. Thus a singular thought has conditions of identity and existence that depend on objective particulars—birds, cities, planets, other people, etc. No such particular, then no such thought; and thoughts are distinct in virtue of the distinct particulars they contain. It is not often remarked that the same thing is true of the properties that constitute the other half of the proposition (so to speak): they too supply the identity and existence conditions of the thought (or the meaning of the corresponding sentence). The property also sits inside the proposition, alongside its partner, the particular. Propositions offer hospitality to both sorts of entity.

Given this general picture, we can formulate a kind of transcendental argument for the existence of particulars, as follows. If we accept that singular thoughts exist, then the world must contain the particulars that form them—both those particulars in particular and also particulars as a category. That is, there is no possible world in which singular thoughts exist and particulars don’t. The particulars don’t have to be material objects or events but could also be mental particulars; so the transcendental argument doesn’t disprove idealism. What is required is just thatsomeparticulars exist—specifically those that form the singular propositions that constitute the content of singular thoughts (and meanings). We know, then, that the world cannot consist solely of universals. That would not follow if propositions were invariably general as to content; then thosepropositions could exist in the absence of particulars. If description theories of reference were true, perhaps accompanied by Russell’s analysis of descriptions, then the existence of particulars would not be a precondition for the existence of propositional contents; so the world could be void of particulars and those thoughts would still be available to be thought. This is a straight consequence of the theory of singular propositions: singular thought is impossible in the absence of particulars, so if there are singular thoughts there are particulars. The theory of thought thus implies a certain kind of metaphysics—one that accepts particulars as real. And it is certainly an interesting point that thought should be capable of having such metaphysical consequences. I might put it by saying that the distinctness of thoughts depends upon the distinctness of the particulars they concern.

Interesting, but perhaps not startling. More startling, however, is the analogous thesis in respect of universals: that there is a similar transcendental argument proving that universals exist. For thoughts also have general content, carried by concepts corresponding to properties, and this content depends for its existence on the existence of the properties it concerns. Without those properties general thoughts would not be possible. Just as the mind cannot from within its own resources generate singular thoughts—it needs the contribution of objective particulars—so the mind cannot from within its own resources generate general concepts—it needs the contribution of objective universals. Since the proposition contains properties, it depends on properties for its existence; but then there are no thoughts in a world without universals. Thus a certain type of metaphysics is implied: reality must contain universals, in addition to particulars, which serve to make thought possible. It is hard to see how these universals could be creatures of the mind, inventions of some sort, because invention depends upon thought, and hence presupposes the existence of general propositional content. Nor could properties reduce to sets of particulars, on pain of making thoughts about properties into thoughts about sets. I don’t have all these particular things in mind when I think that a certain bird is pretty; I simply have the property of being pretty in mind.[1]So we need to countenance a robust ontology of general properties (universals) given the nature of thought: no universals, no thoughts. Thus we can deduce ontology from psychology, world from mind.

The root reason for this dependence lies in an essential feature of universals: their ability to bring things together. They allow for similarity among diverse particulars. If particulars spread universals around, by giving them multiple instantiations, then universals round particulars up, by determining their similarities. So the following thesis sounds plausible: General concepts need objective universals in order to provide the groupings that general thought delivers. The mind could not manufacture the groupings that record similarities without the aid of objective universals that constitute these similarities. Picture the mind trying to find similarities among particulars without appeal to the objective basis of similarities—it would flounder in the dark. No, it needs to latch onto the external objective grounds of similarity, viz. universals. You can’t think about the class of square things without representing the objective property of being square (or if you do you will need some other subsuming concept). So the basis of mental classification is grasp of the objective respects of similarity, i.e. properties or universals. The mind can’t just make this stuff up itself—it needs outside help. It needs objective universals as much as (or more than) it needs objective particulars. That is, singular propositions really do contain particular objects and general properties of them—and this presupposes a certain kind of ontology. You can’t separate semantics from metaphysics, meaning from reality, thinking from being. In this respect, there is no logical gulf between the subjective and the objective, the inner and outer. Metaphysics shapes psychology. The external world of objects and properties is the foundation of the internal world of individual concepts and general concepts. It is not possible to hold the world of thought constant while varying what reality contains in the way of ontological categories. This is the most general lesson of what has come to be called “externalism”. There cannot be a thinking mind without an objective world to mirror.[2]


[1]This raises the question of whether direct reference theory applies to thoughts explicitly about sets: can propositions contain sets of particulars as well as particulars? That would seem to involve overpopulating the content of thought with all the members of a given set. The alternative would be to suppose that thoughts about sets have descriptive content, as in “the set consisting of all F’s”: here the thought would contain a general concept applicable to a given set and not the set itself as a particular object.

[2]According to traditional “Fregean” thinking, there is no overlap between thought (meaning) and reality, since propositions do not contain anything that belongs to the world of reference (particulars and their properties). But according to direct reference theory propositions contain worldly entities, so there is an overlap between world and mind: the constituents of the one are also constituents of the other. Thus it is that we can deduce ontology from psychology (and vice versa).


Tricky Cogito


Existence and theCogito



The Cogitostrikes most people as intuitively valid, but it has been trenchantly criticized. How exactly the inference is supposed to work still excites controversy. Here I will consider a line of objection that I have not seen pressed before. The natural way to interpret the inference is that it moves from a premise about instantiation to a conclusion about existence: I know with certainty that I instantiate the property of thinking, so I must exist as the subject of this property. We might expand the argument as follows: “I have the property of thinking; if something has a property, that thing must exist; therefore I exist”. My thoughts exist (as I know with certainty), and they must be instantiated in some object; this object is not identical to my thoughts; so we can infer that there exists an object (viz. myself) that is not identical to my thoughts. Thus we can move nontrivially from the existence of thinking to the existence of a thing that thinks. But consider the analogous argument concerning unicorns: “Unicorns are horses with one horn, so unicorns instantiate the property of having a horn; but there has to be an object that instantiates this property; therefore unicorns exist.” Or: “Santa Claus has a beard, so he instantiates the property of having a beard; but then he must be an object that instantiates a property; therefore Santa Claus exists”. The premises seem true but the conclusion is false, so the argument must be invalid—but where does it go wrong? Meinong would give the following answer: it does not follow from the fact that an object instantiates a property that the object exists—it might only subsist. In more recent terminology, these objects might be merely “intentional objects” not real existent objects; and so instantiation does not imply existence on the part of the instantiating object. Applying this point to the Cogito, what is to rule out the possibility that the self is a merely intentional object that instantiates the property of thinking but does not exist? We can see that objects are able instantiate properties without thereby existing, so why can’t the self be one of those? To be sure, the propertiesexist—they are real entities all right—but it doesn’t follow that anything that instantiatesthem is itself real. Fictional objects are a counterexample: the property of being a detective is a real property, but the fact that Sherlock Holmes is a detective doesn’t make himreal. Likewise, thinking is a real property that things can have, but it doesn’t follow that anything that instantiates this property is itself real—after all, Holmes also thinks. Maybe the self is like Holmes.

How might we respond to this objection? One possibility would be to appeal to the certaintyof the premise that I think, holding that this is what sets the Cogitoapart. But am I not also certain that unicorns have one horn, that Santa Claus has a beard, and that Holmes is a detective? The fact that an object certainly instantiates a property does not automatically confer existence on that object (it is certain that the Golden Mountain is a mountain). Another possible way out would be to question the whole ontology of subsistent or intentional objects, insisting that there are no objects but existent ones. This flies in the face of the seemingly obvious fact that some things don’t exist and yet have properties; but also, from the point of view of the Cogito, it gives up the certainty of the inference—for now we have to accept that the validity of the Cogitodepends on the rejection of Meinongian metaphysics. Maybe we should, maybe we shouldn’t, but we don’t want the fate of the Cogitoto be tied to that metaphysical issue. No one hearing the Cogitofor the first time thinks, “Well, it depends on your view of Meinong”. Descartes surely did not take a stand against Meinong-style ontology when he enunciated the Cogito. The validity of the argument should not depend on whether or not you entertain Meinongian predilections. It isn’t as if those with such predilections are prohibited from accepting the Cogito; at any rate, that’s not the way the question presents itself.

A third suggestion is that there is an asymmetry between unicorns (etc.) and thinking selves, namely that you can hallucinate unicorns but you can’t hallucinate thinking selves. Thus it can seem to you as if there is a unicorn in front of you without there being one, but you can’t hallucinate having thoughts without actually having them. That asymmetry must be conceded—beliefs in unicorns are not epistemically necessary but beliefs in thoughts are—but it doesn’t help to salvage the argument: for all we get from this is that the existence of thoughtsis certain, not that the existence of thinking selvesis. True, we can be certain that thoughts exist, but we could still be in error about the existence of a self that has them. There might be nothing except thoughts in the vicinity. We might be under an illusion about the self, as we might be under an illusion about unicorns. We might be misinterpreting a collection of thoughts as a self that hasthem. Compare seeing a swarm of bees in the distance and mistakenly thinking there is a single big organism there. Maybe we hallucinate a unitary self when we introspectively encounter a swarm of thoughts. Who knows what might be going on? We can’t hallucinate the thoughts, but we could be under an illusion about what they signify, i.e. an underlying unitary self. Similarly, when we hallucinate a unicorn we are not hallucinating its component properties—they are real enough quaproperties (though their present instantiation is illusory). Anyway, even if it is somehow impossible to hallucinate a self, why should the existence of a self be entailed by the existence of individual thoughts? We still haven’t justified the step from the existence of thoughts to the existence of a thing that has them (the Gassendi-Lichtenberg objection).

Fourth, we might hope to find something in the specific nature of thought that guarantees an existent thinker, where this something is not present for properties in general. Maybe having a single horn doesn’t prove the existence of what has the horn, but thought might be such as to necessitate an underlying existent thinker. This would be the analogue of the ontological argument: most properties indeed fail to guarantee existence, but the property of total perfection does guarantee it, because of its specific nature. Instantiation by thatproperty entails the existence of the instantiating object (according to the ontological argument). There has been a strong intuition (notably voiced by Frege) that mental states necessarily require a bearer—something that has them, a subject. But it is far from clear how this can help the Cogito: Meinong could agree with this point while insisting that the bearer is a merely subsistent entity. Fictional mental states logically require a bearer too, viz. a fictional character, but they lack existence. The notion of a subject of predication—an object of instantiation—is too weak to deliver the conclusion of the Cogito: we might be predicating thought of a non-existent object. But also: does the intuition hold for all kinds of mental states or only thoughts? And does it apply to unconscious mental states as well as conscious ones? Is it true of the bodily sensations of jellyfish or worms? And is the point any different from the claim that any kind of state needs something for it to be a state of, including the state of being electrically charged or the state of the weather? We really need an argument, analogous to the ontological argument, showing that thinking inherently and uniquely calls for an existent subject that can be the reference of “I”, but what this argument might be remains elusive. Could it be that the capacity to think requires the capacity for the self-attribution of thoughts, and hence for a self? But what could establish that, and how would it guarantee the reality of the self in question? Descartes never argued anything of the kind, and it would certainly undermine his claim that the Cogitois primitively compelling. So there is nothing comparable to the ontological argument showing that it is in the nature of thought to bring with it an existent bearer, let alone one with the characteristics of the self as normally understood.[1]

But perhaps there is an intuition lurking in this unsuccessful argument that might have more cogency: namely that there is a contradiction in the idea of existent thoughts occurring in a non-existent object. Fictional thoughts can occur in a fictional object, but non-fictional thoughts require a non-fictional object. Thoughts are particulars not universals (token not types) and existent particulars need existent objects to inhere in. Properties can exist and be properties of non-existent objects, but events and processes can’t both exist and also inhere in non-existent objects. What would it mean for Sherlock Holmes the fictional character to have an existent real thought? Wouldn’t that make him real? So there is a metaphysical assumption at the heart of the Cogito: those thoughts whose existence is evident to us must exist in a being that is itself existent, since real events need real objects as bearer. Real mental particulars need real mental substances to inhere in—they can’t exist in an unreal substance. In Meinong’s language, existent events cannot inhere in merely subsistent objects. If there is real thinking going on, then this requires a real entity to do the thinking; and we know for sure that real thinking is going on—hence we know for sure that we exist. If the thinking was fictional, the subject of it would or might be fictional too; but granted that the thinking is not fictional, neither can he thinker be. Thus I know that I am not a character in fiction (or a “logical fiction” or an hallucination). Compare the bodily counterpart to the Cogito: “I have a body, therefore I exist”. We could not object to this that the premise could be true and the conclusion false, because the body, not being fictional, cannot be had by something fictional: something must non-fictionally exist if my body non-fictionally exists (whether it is the reference of “I” is a further question). We can infer from the existence of the body that something exists. To be more precise, if I know that I have bodily states, then I know that something exists in which those states occur—for example, if I digest there must exist something that digests. The inference is solid because it is compelling to claim that physical states require a physical bearer—a physical thing that has them. If I know there are physical states, then I can deduce that there are physical objects, because states must be states ofsomething. And it would be absurd to suggest that existent physical states could exist in a non-existent object—a merely intentional object. If Holmes is in real physical states, then he must be real himself—fictional characters can’t have real indigestion! The difference from the classic Cogitois just that the premise here is not certain (not an epistemic necessity): I don’t know for certain that I am in physical states or even that I have a body. So this is no use for Descartes’s purposes, though the connection between premise and conclusion is the same in both cases, viz. a metaphysical principle precluding unreal bearers of real states. As we might put it: we can’t mix the existent with the subsistent, the real with the imaginary, the factual with the fictional.

Where does this leave the Cogito? It allows it to struggle on, to retain a semblance of cogency, but it leaves it vulnerable to skeptical doubt. Descartes was working with a scholastic metaphysics of substance and accident—his evil demon was not supposed to call thatinto question. Against this background the Cogitois relatively smooth sailing. But a dogged opponent might protest that this metaphysics is not immune to doubt, and if it is doubted the Cogitowill not go through. Why should we accept that there are substances at all, material or immaterial—why not make do with an ontology of states, events, and processes? Later philosophers indeed did advocate such a low-calorie ontology (e.g. Russell) and so there is no substance in the world for thoughts to inhere in anyway. Thoughts may form sets or aggregates, according to this point of view, thereby gaining a sort of collective unity, but they don’t inhere in substances. So there is no valid metaphysical principle licensing the move from the existence of states to the existence of substances that support them. That may be wrong as a piece of metaphysics, but in the context of the Cogitoit would need to be addressed.

More subtly, however, there is this problem: the prima faciepersuasiveness of the Cogitoseems not to be hostage to the kind of principle I have invoked to bolster it. Is this what someone is thinking who accepts the Cogito? Possibly, but it would have to be in some subliminal or tacit manner, since the requisite principle is hardly laid bare in the usual presentations of the Cogito. It takes work to come up with it and it is not entirely self-evident, even if you accept it. So it is not clear it can explain the felt cogency of the Cogito; indeed, it renders it much more wobbly than we might initially have supposed. Once it is fully articulated in this way its appearance of self-evidence starts to fade, and yet it is routinely hailed as the surest of philosophical theses. The resourceful Meinongian has produced a new threat to the Cogito, and the suggested repair to the argument lacks transparent cogency, even if it is ultimately correct as a piece of metaphysics. So we are left in a rather unsatisfactory position: the argument is not clearly valid but not clearly invalid either. I would say that it can be salvaged more effectively than might have seemed possible once the Meinongian objection has been formulated; for that objection looks formidable at first sight and it takes some ingenuity to forestall it. Hume’s kind of critique of the self, which takes the self as a type of fiction (at least on some interpretations of Hume), was evidently alien to Descartes and quite inimical to the Cogito(we can’t deduce the existence of a substantial self merely from the existence of ideas encountered internally). And yet it is hard to deny that the Cogitohas all the appearance of self-evidence, whatever its final analysis may turn out to be. This philosophical gem thus remains as tantalizing as ever.



[1]Not that the ontological argument is really any good in the end, but it is at least an argumentthat needs to be contended with.


Moral Distance





Moral Distance



We tend to think that our moral obligations fall off with distance: the closer someone is to us the greater is his moral claim on us, and the further away the less. Morality operates like gravity—it weakens with distance. True, morality is an expanding circle, but it is also a diminishing sphere. At the outer limits it hardly gets a grip at all. It would be a mistake to interpret this notion of distance as mere spatial separation, though that seems to be one component of it. In addition to distance in space there is also distance in time: the more temporally remote some future person is the less hold she seems to have on us morally. What obligations do I have to people in the 30thcentury? If I have any, they are not so strong as the obligations I have to people now. Whether this is rational or morally justifiable is a debatable question, but as a descriptive truth about our moral attitudes it is surely correct. Further, there is the dimension of personal contact or emotional proximity: the more intimate my relationship with someone the greater the obligation I feel. This applies to family, spouse, friends, colleagues, and so on. It may be that spatiotemporal distance is really just correlated with this dimension, which is the underlying factor; relationship-distance is the main consideration. We might also add psychological similarity: we tend to regard beings similar to us psychologically as deserving more of our moral concern—humanlike, mammalian, warm-blooded, non-alien. Thus our contemporaneous close relatives have a stronger claim on us than a jellyfish-type creature living in a remote galaxy two million years from now. Moral distance is multi-factorial and complex not just a matter of physical miles. It introduces degreesof obligation into moral duty instead of just the all-or-nothing binary opposition of duty and non-duty. It also introduces uncertainty and messiness into our moral calculations.

It is helpful to picture the diminishing moral sphere as follows. At the center lies the ever-present self: this is the being minimally distant from the moral agent (they are identical) and it has a uniquely strong influence over our decisions. Given that prudence is also moral concern for one sentient being among others—I am a valuable being just like everyone else—we can think of prudence as the basic case of moral obligation. I am obliged to be concerned about my own interests and I am extremely close to myself. I am the center of the sphere of my concerns and others radiate outwards from me. The next closest being is then a matter of individual variation: it could be my spouse or my parents or my children or my friends, depending on circumstances. Then we get to the much more extensive circle of my general acquaintance. After that we have members of my local community perhaps; then other countries; then other species; then the next generation; then more remote generations; then beings in other galaxies; finally completely alien life-forms in remote regions of space millennia hence. My own interests come first (other things being equal and given a degree of selfishness) and then the interests of others according to their place in the sphere. This is the whole sphere of my moral obligations, and it varies in degree of demandingness. Most obviously, there is variation in the strength with which I am obliged to reduce or prevent suffering.

Imagine if someone inverted this ordering of moral priorities: she treats the more remote beings as having a strongerhold on her moral concern. Creatures in the distant future on remote planets that are psychologically dissimilar to her occupy the center of her moral universe, while family and friends have merely marginal moral interest for her (she might even regard herselfas morally negligible). That would certainly strike us as bizarre, insane even, but it is not easy to see how we could persuade her that it is irrational or immoral (she might point out that they are suffering sentient beings too, equally deserving of respect and care). But by the same token it is hard to see how we could be deemed irrational for our ordering. Nor does it seem justifiable to insist that only equalconsideration is rational or moral, so that we must treat spatiotemporally remote beings as morally interchangeable with our nearest and dearest. In fact, the distribution that seemsthe most natural is precisely the one that we adopt—despite the fact that no obvious foundation for it can be produced. Perhaps it is just psychologically necessary for human beings or other evolved creatures, or even for all beings with emotions directed at others: no other moral psychology is feasible given the basic nature of sentient beings. Ought implies can, so there is no point in reprimanding us for favoring the more proximate beings. Not that we can or should have noconcern for the remote and alien, but it must of psychological necessity be diluted and relatively undemanding. If this means that we cannot occupy an entirely impartial and objective moral perspective, then so be it; at least the perspective we have is workable and not too destructive or callous. Brain surgery that changed our moral psychology so as not to discriminate against the distant and different might totally wreck everything that makes human life worthwhile, or even possible. How could you marry someone who systematically favored the remote over the proximate? What would happen to loyalty, trust, solidarity, etc.? What would happen to family life if parents treated every child in the world as deserving the same care and attention as their own?

It is not an easy matter deciding how robustly moral obligation extends to the distant objects of possible concern. Morality has not evolved with these quandaries in mind. For instance, we have not had to think about how our actions will affect the wellbeing of people in future generations, as in climate change; nor did our ancestors put much thought into our obligations towards animals. I don’t want to argue that our current distribution of moral concern is correct and beyond reproach, only that it is not irrational to treat distance (in the multi-factorial sense) as morally relevant (certainly we should be careful about trying to reconfigure our psychology to adopt a more impartial point of view). I would not, for example, be happy to see support for foreign aid curtailed in favor of a supposed more pressing need on the part of future generations, or the political plight of the Venusians. This is also not a point about favoring humans over non-humans: I am all for treating our own animals as having moral priority over more distant animals, because this exhibits the kind of relational closeness that confers moral priority (though I also think remote animals do deserve some moral consideration). Somesort of moral ordering seems inescapable, but whether we have it right now is another question. What we should not do is try to motivate concern for our fellow man (and other animals) by appeal to some perfectly general principle banning all forms of moral distancing, as if every sentient being in the universe had an exactly equal claim over us.[1]Things are more nuanced than that, and more difficult to resolve.



[1]People sometimes say that we should try to occupy a God’s-eye view of creation, morally speaking, treating all sentient beings equally. But God does not exist in space and time, and he has no selective emotional relations with human beings and other animals. We do, and it is folly to try to make us take up a Godlike moral perspective. We can take this perspective into account, but we shouldn’t be governed by it, on pain of possible psychological collapse. In any case, there is no demonstration that the diminishing sphere that we habitually operate with is irrational or immoral.