The World of Universals

Reading: Bertrand Russell, “The World of Universals” and “On Our Knowledge of Universals,” chapters from The Problems of Philosophy

Introducing Universals

What Are Universals, and Why Do They Matter?

The topic of universals is one of the oldest and most contested in the history of metaphysics, with serious philosophical debate stretching back at least to Plato in ancient Greece (c. 428–348 BCE). Before plunging into arguments, it is worth pausing over what philosophers actually mean when they invoke this term — not because it is obscure, but precisely because it emerges from the most commonplace kinds of observation.

Consider two ordinary objects: a piece of chalk and a blank sheet of paper. Place them side by side and ask what they have in common. The obvious answer is that both are white. Now consider a triangle drawn on a chalkboard alongside the stone facade of an Egyptian pyramid. Again, ask what they share: both are triangular. These are humble observations, but they carry a surprising philosophical weight.

When we say that the chalk and the paper both have whiteness in common, we seem to be implying that whiteness is something beyond either object taken on its own. The whiteness cannot be the chalk alone, since it is not exclusive to the chalk — the paper is white too. Nor can it be the paper alone, for the same reason. This reasoning points toward whiteness as a kind of third thing: something that belongs to neither object exclusively but is common to both. The same logic applies to triangularity shared between the chalkboard sketch and the pyramid face. If we take the language of “having something in common” seriously, we seem to be committed to the existence of such shared entities.

One might object here. Could we not say instead that the chalk has its own whiteness and the paper has its own whiteness, and leave it at that — two objects, each with a private version of the colour? This is a coherent position, but as the lectures point out, it comes at a cost: if each object just owns its own private version of whiteness, then they do not really have anything in common at all. The notion of genuine sharing, of real commonality, evaporates. Philosophers who accept universals are precisely those who insist on taking such commonality at face value.

A Provisional Definition

Though Bertrand Russell never formally defines “universal” in so many words in the assigned reading, we can reconstruct a working definition from his discussion: universals are qualities, characteristics, or natures that can be common to many particular objects, even as those objects are situated at different places and different times. The whiteness shared by the chalk in one room and the paper in another room persists regardless of the spatial distance between them. The triangularity that the chalkboard sketch shares with the ancient pyramid holds across centuries of time. Crucially, on standard theories of universals, these shared characteristics are entities distinct from the particular objects that instantiate them. Triangularity is not the same thing as this particular triangle, since triangularity can equally characterize innumerable other triangles.

Particulars, by contrast, are the individual objects and events we encounter in ordinary sensory experience — the sheet of paper on the desk, a specific act of someone repaying a debt, the pool ball sitting on the table. Universals are defined in opposition to particulars: they are whatever can be shared by many particulars, whereas a particular is something that is just where it is, just when it is, belonging to no one else.

Plato’s Theory of Forms

Universals have a long pedigree, and Russell traces much of the background to Plato. Plato was, first and foremost, a moral philosopher who inherited from Socrates the urgent task of arriving at clear definitions of ethical concepts like justice and piety. He noticed that you cannot define justice simply by pointing to a specific just act — say, one person repaying a debt to another. That single example cannot do the work of a general definition, because it tells you nothing about what all just actions have in common. To define justice adequately, you must identify some characteristic (or set of characteristics) shared across every possible instance of just action.

This insight led Plato to his Theory of Forms (sometimes misleadingly called the “theory of ideas” — misleadingly because Plato did not think of these as mental entities but as objective features of reality). Forms are the universals of Plato’s system: the Form of Justice is what all just acts instantiate; the Form of Beauty is what all beautiful things share. As Russell summarizes the Platonic insight, “a universal will be anything which may be shared by many particulars, and has those characteristics which, as we saw, distinguish justice and whiteness from just acts and white things.” Particulars — things “given in sensation” — contrast with universals, which are not given to us in the same direct sensory way.

Universals — Further Details and Arguments

Types of Universals

Having established the basic concept, Russell draws attention to the fact that universals are not all of one kind. First, there are qualities — features like whiteness, hardness, or triangularity that characterize individual objects. Second, there are kind universals (sometimes called sortal universals): terms like “human being” or “dog” refer not to any individual but to a whole kind or type of being, membership in which requires having the common nature of that kind. Third, and most importantly for Russell’s subsequent discussion, there are relational universals — universals that express relations between objects rather than intrinsic features of single objects. When we say that Edinburgh is to the north of London, or that one thing is taller than another, we invoke a relation that holds between the paired objects. Relations, too, qualify as universals, since any such relation can hold between many different pairs of objects and is therefore not tied to any one particular instance.

Philosophers frequently use the term property as roughly synonymous with “universal,” so when they say that whiteness is a property many objects share, they mean roughly what Russell means when he says whiteness is a universal. Relational universals may also be called relational properties.

The Curious Features of Universals

Having described what universals are supposed to be, Russell is candid about how strange they appear. He asserts that universals are “not perceivable by the senses” and do not “exist in the world of sense.” More dramatically, he notes of the relation “north of” that it is “neither in space nor in time, neither material nor mental; yet it is something.” Consider: if Edinburgh is north of London, you cannot point to a place where the northness of the relation sits — not in Edinburgh, not in London, not somewhere in between. And yet the relation must somehow be available to link Edinburgh to London and also to link every other place that stands north of some other place. For this reason, Russell concludes that universals cannot be spatially or temporally localized. They subsist in some mode of existence entirely different from the particulars we encounter in everyday life.

Given these curious features, why should we believe in universals at all? Russell offers two main lines of argument.

The Argument from Language

The first argument is linguistic. Some words in our language clearly refer to particulars: proper names like “Bertrand Russell,” pronouns like “I,” “he,” and “this,” and complex descriptive phrases that pick out single individuals. But many other words do not seem to refer to any particular object. Common nouns like “triangle” or “house,” adjectives like “blue” or “tall,” prepositions like “above” or “to the left of,” and even verbs like “like” — none of these obviously denotes any particular thing. They seem rather to denote types, characteristics, or relations, which apply across many particular cases. Russell’s claim is that these words refer to universals, and further, that “no sentence can be made up without at least one word which denotes a universal.” From this linguistic premise, he concludes that “all truths involve universals, and all knowledge of truth involves acquaintance with universals.”

The argument is suggestive, but one should evaluate it critically. The premise already presupposes a certain interpretation of general terms — namely that they refer to universals rather than to something else. A skeptic who had not already accepted universals might resist that interpretive step. The transition from a claim about language to a claim about the nature of truth also deserves scrutiny.

The Argument Against Nominalism

The second argument targets nominalism — the opposing view that there are no universals, only particular objects that resemble one another. On this view, the word “triangular” does not refer to a shared universal called triangularity; rather, it classifies all those objects that happen to resemble each other in a certain way. The appeal of nominalism is that it dispenses with the mysterious non-spatial, non-temporal entities that Russell posits, replacing them with nothing more than observable resemblances between ordinary things.

Russell’s counter-argument is elegant: the nominalist still cannot avoid universals, because the resemblance relation itself must be a universal. Consider: in order to group all triangular objects together on the basis of resemblance, you need a resemblance that holds not just between one pair of triangles but between any pair that falls in the class. The resemblance between triangle A and triangle B must be the same relation as the resemblance between triangle C and triangle D. That means resemblance is itself a common feature shared across many instances — precisely the profile of a universal. The nominalist is, as Russell puts it, “in for a dime, in for a dollar” when it comes to universals. Granting the reality of resemblance as a universal undermines the motivation for rejecting other universals.

Knowledge of Universals

Empirical Knowledge and the Puzzle It Creates

The third lecture shifts from metaphysics to epistemology, from asking what universals are to asking how we can know them. Russell’s position is that at least some universals are known empirically — that is, through experience. He uses the term “acquaintance” for this mode of knowing: we become acquainted with universals through our experience of particular things that instantiate them. This yields an immediate puzzle, however: Russell has already told us that universals are not perceptible by the senses and do not inhabit the empirical world of particulars. So how can our knowledge of them be empirical?

A posteriori knowledge (also called empirical knowledge) is knowledge gained through sensory experience of the world — knowledge of what is in our drawers, what the weather is like, who won a game. Russell wants to say that our knowledge of some universals is at least rooted in such experience, even if it ultimately goes beyond it.

Abstraction

Russell’s resolution to this puzzle is the notion of abstraction. When we see many white patches — the chalk, the paper, the snow, the paint — sensation gives us acquaintance with the particular patches themselves. But we are not yet acquainted with whiteness as a universal. To arrive at the universal, we must engage in a process of abstraction: a mental operation by which we attend to what is common across many particulars and set aside the features that differ. In Russell’s words, “in seeing many white patches we easily learn to abstract the whiteness which they all have in common, and in learning to do this we are learning to be acquainted with universals.” Sensation provides the starting point; abstraction carries us beyond the individual instances to the shared characteristic.

Russell extends this account to relational universals as well. We can abstract the relation of one thing being to the left of another, the relations of before and after in time, and the relation of resemblance, by observing many instances of these relations holding among particulars, and then attending to what is common across those instances.

Relations Among Universals

Beyond relations between particulars, Russell insists there are also relations between universals. The claim “grey is darker than white” does not mention any particular grey or white object; it asserts a relationship between the two universals, grey and white, themselves. More intricately, consider: “the resemblance between two shades of green is greater than the resemblance between a shade of green and a shade of red.” Here a relation (the greater than relation) holds between two resemblances, which are themselves relations. So we have relations holding between relations — a point that Russell, as a philosopher of mathematics, found particularly important, since mathematicians routinely work with relations of this higher order.

A Priori Knowledge and Relations Among Universals

A priori knowledge is knowledge that is not based on sensory experience. It typically involves necessary truths — truths that could not be otherwise, that would hold no matter how the empirical facts were shuffled. It also tends to involve generality: claims that apply universally and without exception. Consider two contrasting pairs:

The claim “many bachelors are unhappy” requires empirical investigation — you would need to survey bachelors to find out whether it is true. By contrast, “all bachelors are unmarried” requires no such legwork. You need only grasp the meaning of “bachelor” to recognize that it is necessarily and universally true. Similarly, “there are two pens in my desk drawer” is an empirical claim (I have to check), whereas “two and two are four” seems to be knowable independently of any particular inspection.

A priori truths seem to anticipate and control experience in Russell’s phrase: knowing that all bachelors are unmarried, you can know right now that any bachelor who exists a thousand years from now will also be unmarried. How is this possible?

Russell’s answer is that all a priori knowledge deals exclusively with relations of universals. To know that all bachelors are unmarried, you need to grasp the universal bachelor (a sortal universal), the universal unmarried (a characteristic), and — crucially — the relation between these two universals that makes being unmarried part of what it is to be a bachelor. No particular bachelor need be inspected, because the knowledge concerns the universals themselves and the relational structure among them. Similarly, “two and two are four” is, on Russell’s view, a statement about the universals two, four, and collection, and the relations that hold among them. We grasp its truth by grasping those universals and their relations, not by counting up actual pairs of things in the world.

Reflecting Critically on Universals

Russell’s Realism

Russell’s overall position is a form of realism about universals: he maintains that universals are objective elements of reality that exist independently of any mind that perceives or conceives them. Moreover, this world of universals is extensive. Any time you describe an object’s quality, refer to it by a generic kind-term, or assert a relation between things, you are — on Russell’s account — making reference to universals. The realm of universals is, accordingly, non-spatial, non-temporal, and inaccessible to direct sensory perception.

This extensiveness invites the application of a famous philosophical principle: Occam’s Razor, named after the medieval philosopher William of Occam. The principle states that one should never posit entities beyond what is necessary for explanation. Applied to the theory of universals, the question becomes: could we explain everything that Russell uses universals to explain — the phenomena of shared qualities, generality of language, a priori knowledge — without positing all these mysterious non-sensible entities?

Nominalism Revisited and Conceptualism

Nominalism, as we saw, attempts to eliminate universals by appealing to resemblance rather than shared properties. Russell’s counter that resemblance itself is a universal may or may not fully defeat nominalism; one might argue that even if the nominalist must accept one universal (resemblance), this is still less theoretically extravagant than Russell’s rich ontology of qualities, kinds, and relations.

A more sophisticated alternative is conceptualism. Conceptualism agrees with nominalism in rejecting real, mind-independent universals, but insists that the generality associated with general words is grounded in concepts — elements of our cognitive or mental makeup that enable us to recognize resemblances and classify things under general terms. Crucially, concepts on this view are not objective features of the external world but belong to the mental or cognitive furniture of the mind.

The appeal of conceptualism is that it retains much of what Russell’s theory was designed to explain while making a more modest metaphysical commitment. Consider the term “bachelor.” For Russell, the term refers to a real, atemporal universal of bachelorhood. For the conceptualist, it refers merely to a concept, defined as something like unmarried human male. And this concept is sufficient to ground a priori knowledge: we can know that no bachelor is married simply because the concept is defined to exclude married people. No mysterious universal existing outside time and space is required.

The concept of “snow” provides a harder test. When scientists give a precise definition of snow, are they simply articulating a concept, or are they describing an objective feature of the world? The realist insists on the latter; the conceptualist says we need only concepts, not real universals, to capture what all instances of snow have in common.

Russell’s Argument Against Conceptualism

Russell himself confronts conceptualism with an argument from geography. He asks us to consider the proposition “Edinburgh is north of London.” The part of the earth’s surface where Edinburgh stands would be north of the part where London stands, he argues, “even if there were no human being to know about north and south, and even if there were no minds at all in the universe.” If the fact is mind-independent, then the relation north of that it involves cannot be a mere concept in some mind. Concepts are, by definition, mind-dependent; but this geographical fact seems to obtain independently of any mind.

However, one can push back against Russell here. We might all agree that there is some objective arrangement of territory — Edinburgh’s tract of land is separated from London’s by a certain stretch of countryside. But does it follow that the conceptual framework of north and south would apply independently of any mind? North and south are orientation concepts; without minds that set up such orientational frameworks, one might argue there is no “north” or “south” at all, only an arrangement of landmasses that minds conceptualize in these terms. If that response is correct, the conceptualist can resist Russell’s argument.

A Middle Ground?

Perhaps the most honest conclusion is that the debate between realism and conceptualism does not have a clean resolution. Realism captures the objectivity of our best scientific descriptions of the world — when a physicist defines a property like mass, it seems to track something genuinely out there, not merely something in our heads. Conceptualism captures the cognitive dimension of our grasp of generality and offers a more parsimonious metaphysics. A middle ground might hold that even concepts, insofar as they are shared and applied with mutual understanding across a community of thinkers, have a kind of intersubjective objectivity — they are not merely private, even if they are not mind-independent in the full realist sense. Perhaps the work of science is precisely to develop and refine shared concepts that track the real structure of the world, without committing us to the full platonist picture of atemporal, non-spatial universals existing in a separate realm.

The Nature of Causation

Readings: David Hume, selections from Treatise of Human Nature (1738); G.E.M. (Elizabeth) Anscombe, “Causality and Determination”

What Is a Cause?

The Importance of Causality

Causality is one of the most pervasive concepts in human thought. We take it for granted that changes in the world do not just happen randomly — they occur as a result of prior events that bring them about. When we put a turkey in an oven, we infer a causal connection between the heat and the eventual cooking. When we see smoke on the horizon, we infer a fire as its cause. When we see animal tracks in the snow, we infer that an animal passed through. None of this is philosophically deep in itself, but causality becomes philosophically interesting when we ask what, exactly, we are implying when we make such inferences.

From a metaphysical point of view, causality is extraordinary in scope. Some philosophers have argued that the very concept of something happening implicates causality — that there can be no event without a cause. Causality also lies at the heart of science: much of what scientific laws do is describe causal regularities — the most fundamental and general patterns in which causes produce effects.

David Hume (1711–1776), writing in his Treatise of Human Nature (first published 1738), proposed an analysis of causality that remains one of the most influential and controversial in the history of philosophy. Hume was also famous as an essayist and historian, but his philosophical reputation rests substantially on this analysis.

Hume’s Methodology

Hume’s starting point is methodological: to analyze causality, we should first ask what is packed into the idea of causation as we actually use it, and then trace that idea back to its experiential origins. This reflects Hume’s broader empiricism: for Hume, all genuine concepts (which he calls “ideas”) must be traceable to sensory or inner experience (which he calls “impressions”). The question “where did we get the idea of causality?” is not a psychological curiosity but a philosophical tool for assessing whether a concept is genuinely informative or illusory.

Causality Is Not an Intrinsic Quality of Objects

Hume begins with a negative observation: the idea of causality does not derive from any single perceptible quality of individual objects. A red rubber ball, for instance, has many perceptible qualities — redness, roundness, a certain texture. It is the cause of many effects. But being a cause is not itself one of those perceptible qualities, because a white marble cube — lacking all those qualities — is equally capable of being a cause. What makes something a cause cannot be any intrinsic perceptible quality, since wildly different objects can all be causes of effects. As Hume puts it, causality must be “derived from some relation among objects” rather than being an intrinsic property of any one object.

Three Elements of the Causal Relation

Having established that causality is a relation rather than an intrinsic quality, Hume proceeds to identify the elements that compose it.

The first is contiguity: cause and effect must be spatially and temporally adjacent. Nothing can “operate in a time or place which is ever so little removed from those of its existence,” as Hume writes. If cause and effect seem to operate at a distance, careful analysis will reveal a chain of intermediate causal events that are themselves contiguous — the appearance of action at a distance is always an illusion.

The second is priority in time: the cause must precede the effect. Though cause and effect may overlap in time (fire and smoke often seem simultaneous), there is always a sense in which the cause must initiate the process before the effect fully unfolds. Hume offers a clever argument against strict simultaneity of cause and effect: if every cause were perfectly simultaneous with its effect, and every effect simultaneous with its effect in turn, there could be no such thing as a sequence of change in the world — everything would have to coexist at once. Since we clearly do observe temporal sequences, causes must precede their effects.

The third and most philosophically significant element is necessary connection: given the cause, the effect must follow. This is the element Hume calls “of much greater importance than any of the other two.” If we see the motion of one pool ball succeed the contact of another, and acknowledge that they were contiguous and that the first motion preceded the second, we might still wonder: was there anything that compelled the second ball to move? The idea of necessary connection is the idea that the effect was not merely contingently followed by the cause but was in some sense required by it. And it is this notion that creates the most philosophical difficulty.

The Search for Necessary Connection

The Empiricist Challenge

Hume’s empiricism creates a serious problem. If every idea must be traceable to an impression (a sensory or inner experience), then we must ask: where does the impression of necessary connection come from? We can see contiguity (the balls making contact) and priority (one motion preceding the other), but we cannot seem to see the necessity of the connection. No matter how carefully we observe a causal transaction, necessary connection does not appear as a distinct perceptible feature of the objects involved.

Hume establishes this by a kind of elimination. The idea of necessary connection does not come from impressions of the qualities of individual objects (their roundness, hardness, etc.). Nor does it come from impressions of the observable relations between objects (their spatial contact, temporal sequence). Even if you observe a fire producing smoke in perfect sensory detail, you observe the co-occurrence and sequence — not any thread of necessity binding them.

Hume makes a crucial further observation: if you were exposed to a single instance of two events occurring together — one fire and one puff of smoke, seen just once — you would never derive the idea that one necessarily produced the other. The two things would just be co-occurring events, for all you could tell. The idea of causal necessity, Hume argues, requires repeated instances of similar conjunctions before it begins to form in the mind.

The Surprising Answer

If repeated instances are what give rise to the idea of necessary connection, then why? What do repetitions add that a single observation lacks? Hume’s answer is striking: repetitions do not reveal anything new in the objects themselves. They do not expose some hidden feature of the fire or the smoke that makes the connection necessary. Rather, the repetitions condition the mind to form an expectation. After seeing fire regularly followed by smoke many times, the appearance of fire automatically produces in us an expectation of smoke. It is this habit of expectation — this “determination of the mind to pass from one object to its usual attendant” — that constitutes what we call necessary connection.

The idea of necessary connection, then, is ultimately an impression not of the external world but of our own mental tendency: “Necessity is nothing but an internal impression of the mind, or a determination to carry our thought from one object to another.”

Two Definitions of a Cause

From this analysis, Hume derives two definitions of causation. The first definition defines a cause in purely objective terms: a cause is “an object precedent and contiguous to another, and where all the objects resembling the former are placed in like relation of precedency and contiguity to those objects that resemble the latter.” In plain terms: causes and effects are objects that regularly and constantly go together in temporal succession, in the same spatiotemporal locale. This is the regularity (or constant conjunction) definition. It makes no mention of necessity at all, treating causality as nothing more than an exceptionless regularity of sequence.

The second definition incorporates Hume’s analysis of necessary connection: a cause is an object “so united with [another] that the idea of the one determines the mind to form the idea of the other, and the impression of the one to form a more lively idea of the latter.” This definition is explicitly psychological: it defines causal connection partly in terms of the mental habits of the perceiving subject. As Hume bluntly puts it, “necessity is something that exists in the mind, not in objects.”

Hume as Causal Anti-Realist

These two definitions position Hume in the broader metaphysical landscape in an interesting way. Realism about causality holds that causal necessity is a genuine objective feature of the world, existing independently of any mind. Anti-realism holds that causal necessity is not purely objective — that it somehow depends on the mind that perceives it. Hume’s second and more complete definition of causality is clearly anti-realist: necessary connection is a projection of our mental habits onto the world, not a discovery of something that was already out there. Objectively speaking, there are only regularities; subjectively, we dress those regularities up in the language of necessity.

Causal Laws and Causal Necessity

Hume’s Legacy

In the centuries since Hume, two aspects of his account have proved especially influential among philosophers. First, the idea that causality involves constant conjunction — that causal claims are ultimately underpinned by regularities relating causes to effects. Second, the idea that causality involves necessary connection of some kind, even if Hume himself treated necessity as mind-dependent.

From the regularity aspect of Hume’s account, philosophers have developed the idea of causal generalizations or causal laws: universal statements of the form “events of type A are always followed by events of type B.” The example “smoking causes cancer” is a causal generalization of everyday science. The physics claim “a net force causes acceleration in the object acted upon” is a more formal causal law. These generalizations register constant conjunctions between types of causes and types of effects, and where the conjunction is sufficiently exceptionless, we call it a law of nature.

If we add the notion of necessity to the regularity, we get a stronger formulation: “Events of type B necessarily follow from causal antecedents of type A.” This stronger picture — causality as both regular and necessary — is what many philosophers, inspired by Hume, have taken to be the proper analysis of causal relations in the world. Even if Hume himself was anti-realist about the necessity, others have argued that causal necessity is real and objective.

Anscombe’s Challenge

Elizabeth Anscombe enters this discussion with a sharp and deliberate challenge to the entire Humean picture. Her central claim is that the ordinary concept of causality implies neither necessity nor universality. Causality, properly understood, does not require that causes necessarily bring about their effects, and it does not require that there be any underlying universal generalization relating types of causes to types of effects.

Anscombe illustrates this with the example of contagious disease. Suppose you have contact with a carrier of a disease, and subsequently contract it. We would naturally say that contact with the carrier caused your illness. But now suppose you ask your doctor, before getting sick, whether you will get the disease. The doctor’s honest answer will be “I don’t know — maybe you will, maybe not.” Contact with a carrier does not always lead to getting the disease; it is a genuine cause in individual cases without being either universally or necessarily operative. If we accept this, then individual causal claims do not presuppose underlying universal regularities or necessary connections.

Anscombe presses the point with further examples: smoking causes heart disease and unprotected sex causes pregnancy are both true in many individual cases, yet neither is true as an exception-free universal generalization. Not all smokers develop heart disease; not every instance of unprotected sex results in pregnancy. But we do not hesitate to say that smoking causes heart disease or that unprotected sex causes pregnancy. If these causal claims are genuine and true, yet they lack universality and necessity, then universality and necessity are not part of the concept of causality as such.

Causality as Derivation

What, then, is causality, if we strip away necessity and universality? Anscombe’s answer is refreshingly simple: causality consists in the derivativeness of an effect from its cause. An effect comes from, arises out of, or derives from its cause. That is the core of the causal relation. In her words: “If A comes from B, this does not imply that every A-like thing comes from some B-like thing, or that every B-like thing has an A-like thing coming from it, or that given B, A had to come from it.” Any further features — universality, necessity — “will be an additional fact, not comprised in A’s coming from B.”

Anscombe also questions whether successful causal generalizations even have the universality that Humeans require. Take the billiard ball example: “Always, when one ball impacts another, the second moves in the direction of the impact.” This generalization fails if the second ball is pressed against a wall. To save the generalization you must add qualifications: “absent interfering factors,” “in open space,” “ceteris paribus.” But as Anscombe argues, the task of specifying and excluding all possible interfering circumstances is literally inexhaustible. The world is too complex for us to enumerate everything that might prevent the usual course of events from unfolding. If causal generalizations always require such open-ended caveats to be true, they do not possess the clean universality that the Humean picture demands.

Reflecting Critically on Causality

Sharpening the Debate

The fourth lecture invites us to step back and evaluate the disagreement between Hume and Anscombe with some philosophical care. The two positions can be stated crisply:

The Humean (or Neo-Humean) view holds that causality involves (1) constant conjunction — whenever A occurs, B follows — which entails universality; and (2) necessary connection between causes and effects, whether understood as an objective feature of the world or as a habit of mind projected onto it.

Anscombe’s view holds that causality involves only the derivation of one thing from another in individual instances, with no implication of universality or necessity. We observe causality in individual cases directly — scraping a knee, pushing a cart, rain wetting the streets — without having to await a long record of constant conjunctions.

The Case for Hume

One of the strongest considerations in favour of the Humean view is the intuitive connection between causality and regularity. Suppose that every time you applied a knife to an onion, something different happened: in the first application, the onion split; in the second, it flew away; in the third, the knife was somehow repelled. In this imagined scenario, you would be genuinely reluctant to say that the knife caused the onion to split in the first case. Your confidence in saying the knife caused the splitting in ordinary life derives, arguably, from the fact that knives regularly split onions. Take away the regularity and the causal judgment becomes uncertain. This suggests that our causal thinking is more deeply entangled with ideas of regularity than Anscombe allows.

Humeans can also respond to Anscombe’s disease example by invoking the notion of a full cause. When contact with a carrier only sometimes leads to disease, perhaps this is because contact is not the full cause — there are other causal factors (the state of the immune system, the dosage of the pathogen, genetic predispositions) that, together with contact, constitute the complete causal condition. If we knew the full cause, we would see that it is indeed necessarily and universally followed by the effect. Our ordinary causal talk is “loose,” referring to partial causes for convenience, but the full causal story always conforms to the Humean picture.

The Case for Anscombe

Anscombe anticipates and resists this “full cause” reply. She points out that it is an assumption that there is always such a thing as a full cause that makes the connection necessary. But we understand perfectly well what it means for contact with a carrier to cause a disease in an individual case, and this understanding is entirely independent of whether any hypothesis about a full, necessitating cause is true or even plausible. Our concept of causality does not wait upon the discovery of such full causes; we apply it readily in individual cases with incomplete information.

Moreover, Anscombe and her allies would argue that it is simply question-begging for the Humean to insist that genuine causal claims must always be backed by exceptionless universal generalizations. The correct procedure is to observe what kind of evidence actually convinces us that causality is present, and what Anscombe claims is that we do observe individual causal processes — we see the rain wet the street, we feel the knife cut the onion — without first requiring a catalog of constant conjunctions. If that is so, then universality and necessity are not built into our basic causal concept.

The debate ultimately turns on a deep question: do we perceive causality in individual instances, or only infer it from repeated patterns? Hume says the latter; Anscombe says the former. And settling this question has ramifications far beyond the philosophy of causation — it matters for how we understand scientific explanation, moral responsibility, and, as Week 3 shows, for the problem of free will.

Freedom and Causation

Reading: Richard Taylor, “Freedom and Determinism,” from Metaphysics

The Problem of Freedom and Determinism

What Is Determinism?

The problem of freedom and determinism is one of the oldest in philosophy, and Richard Taylor’s essay approaches it with characteristic clarity. The discussion begins with a statement of the central thesis at issue: causal determinism.

Determinism is a metaphysical thesis about the causal structure of the world. It holds that everything that occurs does so as a result of antecedent causal conditions that necessitate it — conditions such that, given their existence, the event could not have failed to occur. A useful image is that of a long row of dominoes: each domino falls because the previous one struck it, and once the sequence is set in motion, every subsequent falling is guaranteed by the one before it. Determinism applies this picture to everything that happens — not just falling dominoes but every physical event, every biological process, every thought, every decision, every action.

Taylor’s own statement is precise: in the case of everything that exists, “there are antecedent conditions, known or unknown, given which that thing could not be other than it is.” Note the modal weight: things could not be other than they are. This is not merely the claim that things happen to occur in regular sequences; it is the claim that those sequences are necessary, that given the causal antecedents, no other outcome was genuinely possible. This connects the determinism debate directly to the discussion of necessity in Week 2: determinism tends to rely on a Humean or Neo-Humean view that associates causality with necessity.

Why Believe in Determinism?

One reason to find determinism plausible is simply that the world is determinate: at this very moment, the world is exactly the way it is, not some other way. Something must explain why it is this way and not another — and the most natural explanation is that prior causal conditions brought it about. Taylor also notes that human beings seem to operate with an ingrained presupposition of determinism: “Men believe, or at least act as though they believe, that things have causes, without exception.” Whenever we hear an unfamiliar noise, we instinctively look for a cause; we are never satisfied with the explanation that a noise simply occurred with no causal background. This pervasive tendency to seek causes suggests that we already, at some practical level, accept something like the determinist thesis.

The Problem: Freedom and Determinism in Tension

The philosophical difficulty arises when we apply determinism to human action. If every event in the world — including every thought I think, every desire I feel, every decision I make — is necessitated by prior causal conditions, then in what sense am I free? As Taylor writes: “What am I but a helpless product of nature, destined by her to do whatever I do and to become whatever I become?” If my actions are merely the inevitable outputs of a causal machine that began running before I was born, it is hard to see in what sense I can be credited (or blamed) for what I do, or in what sense my actions are genuinely mine rather than just effects of antecedent forces.

Why not simply accept determinism and give up on the idea of freedom? Taylor resists this quick resolution by drawing attention to two features of our ordinary experience that are deeply familiar and that any account of human action must take seriously.

The Two Data: Deliberation and “It’s Up to Me”

The first datum is deliberation: the process of weighing options and making a decision about what to do. Deliberation is something we genuinely seem to engage in. Several important points can be made about it: (a) you can only deliberate about future actions, not past ones already done; (b) you can only deliberate if the outcome is genuinely open, not already decided; and (c) you can only deliberate about things that are, in some sense, within your control. If what you are going to do is already fixed by prior causal conditions, there is literally nothing left to deliberate about. Deliberation presupposes genuine openness.

The second datum is that some things are “up to me”. Taylor illustrates this with a simple case: right now, it seems genuinely open whether I move my finger to the left or to the right. Both options seem available, both under my direct control. More broadly, decisions about how to spend an afternoon, which path to take, which words to speak — these seem like things I am genuinely in control of, not things that happen to me as a boulder rolls down a hill. Taylor’s point is not that everything is up to us (many things clearly are not — reflex actions, heartbeats, involuntary responses), but that some things genuinely are.

These two data are so familiar, so woven into our ordinary practical self-understanding, that any philosophical account of human action must grapple with them. And crucially, neither of them seems comfortable alongside a strict determinist thesis.

Soft Determinism — A Compatibilist Approach

The Compatibilist Strategy

Soft determinism (also called compatibilism) is a philosophical position that accepts the determinist thesis yet maintains that human freedom is nonetheless possible. It aims to reconcile determinism with freedom rather than choosing one at the expense of the other.

Taylor identifies three core theses in soft determinism:

First, soft determinism simply accepts that determinism is true. Every human behaviour, voluntary or otherwise, “arises from antecedent conditions, given which no other behaviour is possible.” The soft determinist grants that causal necessity extends into the domain of human action without exception.

Second, soft determinism redefines freedom: a free action is one that is “not externally constrained or impeded.” If nothing external is preventing you from doing what you choose to do, then you are acting freely. The key word here is external: freedom, on this view, is the absence of external obstruction.

Third, free actions are specifically those whose causes lie in the agent’s own internal states — “their own acts of will, volitions, choices, decisions, desires.” If you go to the corner store because you desire chips and decide to go get them, and nothing external prevents you from doing so, then your going is a free action. If, by contrast, you are late to a meeting because a traffic jam held you up despite your best intentions, your lateness is not free — it was caused by an external constraint.

Illustrating Soft Determinist Freedom

A free action, on this picture, is one where the causal chain runs smoothly from internal states (desire, decision, volition) to action, without external interference. An un-free action is one where external circumstances break into that causal chain, preventing the internal states from producing their intended effects. Note that the soft determinist does not deny that the desires, decisions, and choices are themselves caused — they have causal histories stretching back before the agent was born, part of the whole causal fabric of the determinist universe. But this does not matter for freedom; what matters is that the immediate cause of the action is the agent’s own inner life, not some external force.

Taylor’s Critique of Soft Determinism

Taylor does not find this account of freedom satisfying, and his objections are worth tracing carefully.

His first objection is that the soft determinist’s concept of freedom is inadequate as a concept of freedom. Taylor asks us to imagine an ingenious physiologist who has devised a mechanism for implanting any volition he chooses in his experimental subject — any desire, any decision, any motivation. By manipulating buttons and wires, the physiologist causes the subject to want to raise their hand, and the subject raises it; causes them to want to kick their foot, and they kick. Crucially, these actions meet the soft determinist’s definition of free action: the actions are caused by the subject’s inner states (the induced desires and decisions), and nothing external is impeding their execution. Yet Taylor argues — persuasively — that this is not a description of a free and responsible agent. “It is the perfect description of a puppet.”

The same logic applies even without a literal physiologist: whenever your desires, decisions, and choices are causally determined by factors entirely beyond your control — even impersonal factors — the resulting actions do not seem genuinely free, even if they flow from “your” inner states in the causal chain.

Taylor’s second objection is that soft determinism fails to accommodate the two data. Since soft determinism is a form of determinism, it entails that whatever you do at any time is the only thing you could have done given the causal antecedents. This leaves no room for deliberation (which requires genuinely open options) and no room for things being “up to me” (which requires that alternatives be genuinely available). As Taylor puts it: “If I believe that what I am going to do has been rendered inevitable by conditions already existing, then I cannot try to decide whether to do it or not, for there is simply nothing left to decide.”

Soft Determinism vs. The Agency Theory

Taylor’s Agency Theory

In response to the failures of soft determinism, Taylor proposes what he calls the agency theory of free action. This is an incompatibilist view: it denies that freedom can be reconciled with determinism, and it concludes that free human action involves a kind of causality that is fundamentally different from the causality governing the rest of the natural world.

The core claim of the agency theory is that human beings are sometimes self-determining beings — beings who are “sometimes, but of course not always, causes of their own behaviour.” When a person acts freely, it is the person themselves, not some prior event or circumstance, who is the cause of the action. Taylor insists on this: “When I believe that I have done something, I do believe that it was I who caused it to be done, I who made something happen, and not merely something within me, such as one of my own subjective states, which is not identical with myself.” Free action is authored by the agent as such, not merely by some internal mental state that the agent happens to possess.

The Metaphysical Commitments

The agency theory comes with significant metaphysical presuppositions. First, it requires that there be a self or person who is not merely a collection of things or events, but a substance and a self-moving being. A person cannot be simply a bundle of physical parts obeying the same causal laws as the rest of matter, nor can they be simply a sequence of mental experiences. They must be something more: a genuine originator of action, capable of standing as the ultimate source of what they do.

Second, the agency theory requires a special kind of causality — what we might call agent-causation — that differs from ordinary event-causation in two important ways. (a) An agent can “cause an event to occur — namely some action of his own — without anything else causing him to do so.” This is causation without being caused, origination without prior necessitation. (b) Agent-causation is not a relation of sufficient conditions: when a person causes a free action, the person’s existence and character are not a sufficient condition for the action’s occurrence. The action happens, but it could have failed to happen — the person could have done otherwise. This is precisely what distinguishes agent-causation from ordinary causal necessity.

Taylor acknowledges the strangeness of this notion. The causation that associates with agents “is in fact so different from the usual philosophical conception of a cause that it should not even bear the same name.” He suggests it might be better to say that agents originate, initiate, or simply perform their actions, rather than saying they cause them in the ordinary sense.

Fitting the Two Data

The agency theory’s chief virtue is that it accommodates the two data naturally. Since free actions are not determined by any prior causal condition, alternatives genuinely remain open at the moment of action. The agent chooses without the choice being necessitated, and therefore there is real scope for deliberation between open options. Similarly, since nothing external to the agent determines the action, the action is genuinely “up to” the agent in the fullest sense.

Evaluating the Debate

What Can Be Said for the Agency Theory?

The agency theory has clear attractions. It fits with the two data that any theory of human action must address: it genuinely explains how deliberation is possible (by allowing that alternatives remain open) and how things can be genuinely “up to me” (by making the agent the non-necessitated originator of the action). It also provides a more robust account of freedom than soft determinism — one that matches ordinary intuitions about what it would mean to be truly in control of one’s actions.

The Problems with the Agency Theory

The difficulties are serious, however. The agency theory requires us to accept a notion of selfhood that is, as Taylor acknowledges, philosophically extraordinary. Persons, on this view, cannot be purely physical composites operating under the ordinary laws of nature. They must be self-moving substances of a fundamentally different kind. This seems to conflict with a broadly scientific worldview, which finds it hard to find room for non-physical agents who inject causation into the world from outside the ordinary causal order.

The concept of agent-causation is similarly problematic. Almost every other causal process we know of involves causes that are events or states sufficient to bring about their effects. Agent-causation involves a kind of causation that is neither an event nor a sufficient condition. Where else in nature do we see anything like this? If nowhere else, we must ask whether the agency theory is positing an entirely ad hoc and otherwise unobservable kind of causality, motivated solely by the desire to salvage free will.

One might also note, applying Occam’s Razor, that the agency theory introduces very costly metaphysical posits — a special kind of self, a special kind of causation — for the sole purpose of accommodating freedom. If the price of human freedom is accepting these extraordinary notions, one might reasonably wonder whether it would be more economical to deny that freedom, in the robust sense the agency theorist requires, actually exists.

Reconsidering Soft Determinism

Taylor’s critique of soft determinism is powerful, but it is worth asking whether the soft determinist’s picture is really as inadequate as he suggests. We know that human beings are, in part, physical organisms, and that our choices and actions are shaped by physiological and psychological causal factors. Psychological research has made genuine progress in identifying causal influences on human motivation. It does not seem obviously wrong to say that we live within a causal world and that our inner lives are part of that causal order.

Given this, perhaps the soft determinist definition of freedom — action caused by internal states, unimpeded by external constraints — is not so different from what we should expect freedom to mean. When we call someone “free,” in everyday life, we usually mean they are not in prison, not being coerced, not subject to compulsion. We do not normally mean that their actions have no causal history. On this reading, the soft determinist is capturing something real about the ordinary concept of freedom, even if it is not the metaphysically robust notion that Taylor seeks.

The challenge for the soft determinist remains Taylor’s physiologist thought experiment: if freedom means nothing more than “caused by internal states without external impediment,” then a puppet whose inner states are induced by an external agent would count as free. This seems wrong. But the soft determinist might respond that this objection applies only to far-fetched hypothetical cases, and that in the real world — where inner states are not induced by external manipulators but arise from the agent’s own developmental history — the soft determinist definition captures all the freedom that is practically and morally significant.

Anscombe’s View and the Debate

The final lecture encourages us to consider how Anscombe’s analysis of causality, developed in Week 2, bears on the freedom and determinism debate. Recall that for Anscombe, causality does not imply necessity. To say that one event caused another is simply to say that the second derived from or came from the first — it does not imply that the second had to follow. If this is right, then the determinist thesis — that everything that occurs is necessitated by prior conditions — depends on a conception of causality (causality as involving necessity) that Anscombe has given us reason to question.

If we adopt Anscombe’s view, the world might be full of causal relations among events while nonetheless not being deterministic in Taylor’s sense: things can come from prior conditions without being necessitated by them. This would potentially dissolve the tension between causality and freedom: we could say that our actions have causal histories without those histories necessitating the specific actions we take. Whether this constitutes a satisfying resolution — or whether it merely relocates the problem — is itself a live philosophical question, and one that the course invites us to take away and think through independently.

In the end, the debate between soft determinism and the agency theory illuminates a profound tension in how we understand ourselves: are we nodes in a causal network that extends infinitely back in time, or are we genuine originators capable of introducing something new into the world? The answer has implications not only for metaphysics but for ethics, law, and the entire framework within which we hold each other responsible for our actions.

The Search for Selves

Identity and Diversity

Setting Up the Problem

Before asking what it means to be a person, we need to understand a more fundamental concept: what it means for something to be numerically the same thing over time. This week’s discussion, grounded primarily in John Locke’s 1689 chapter “Of Identity and Diversity” (from An Essay Concerning Human Understanding), opens by drawing a distinction that will anchor everything that follows.

Qualitative identity refers to the relationship between two distinct things that share all (or nearly all) of the same intrinsic properties. Two sticks of gum from the same package, or two pens from the same box, are qualitatively identical — they look the same, feel the same, and behave the same way. Strictly speaking, any two things will have some minuscule difference if inspected closely enough, but the concept captures the idea of things being indiscernible in their qualities. The more philosophically significant concept, however, is numerical identity: being “one and the same” thing, rather than two similar things. The oak tree you pass every morning is numerically the same tree whether it is a sapling or a towering mature specimen. You as an adult are numerically identical to you as a child. Numerical identity persists through change, and it requires no sameness of qualities — only sameness of being.

What makes this interesting is that change makes numerical identity puzzling. If everything were changeless and immutable, there would be nothing to ask. But things do change, sometimes dramatically, and so we naturally wonder: at what point does change become so radical that we are dealing with a different thing rather than the same thing in a different condition?

The Ship of Theseus

To illustrate just how puzzling this can get, the lectures invoke one of philosophy’s most ancient thought experiments: the Ship of Theseus. Theseus, a figure from Greek mythology, captains a ship on a long voyage carrying a cargo of timber. As the journey progresses, damaged planks are replaced using timber from the cargo. By the time the ship arrives at its destination, every single plank of the original vessel has been swapped out for a new one.

The question that follows is deceptively simple: is the ship that arrives the same ship that set out? Two opposite intuitions pull with roughly equal force. On one side, since a ship is a material object and every bit of the original material has been replaced, it seems that the arriving vessel cannot be identical to the departing one — it is, in effect, a brand new ship. On the other side, a ship is the kind of thing that undertakes voyages, and no two ships can have undertaken exactly the same voyage at exactly the same time, which suggests that it must be the same ship. The puzzle has no obvious resolution at the outset, and that is precisely its philosophical value.

The Ship of Theseus is not an antiquarian curiosity. A living plant sheds old cells and grows new ones; an animal undergoes dramatic physical transformation from birth to old age. Most acutely for our purposes, human beings are the same way. The atoms constituting your body today are different from those that constituted it years ago. If we are inclined to say Theseus’s ship is a different ship once every plank has changed, should we not say the same of ourselves?

Locke’s General Framework: Origin and Category-Relativity

Locke’s approach to these puzzles is organized around two key ideas. First, he holds that judgments of numerical identity involve comparing a thing with itself at different stages of its existence, and that a thing’s origin provides the standard against which later stages are measured. In Locke’s own words: “That, therefore, that had one beginning, is the same thing; and that which had a different beginning in time and place from that, is not the same, but diverse.” Two things that trace back to the same origin, in the same time and place, must be one and the same thing, because origins are unique.

Second — and this is crucial — Locke holds that identity is category-relative: the criterion for numerical identity differs depending on the kind of thing under consideration. The same physical matter can be described under different categories, and judgments of numerical identity will vary accordingly. Locke distinguishes at least three fundamental kinds of substance: finite spirits (minds or souls), bodies (any material object), and God (divine substance, eternal and unchanging). For each, the origin criterion applies straightforwardly: a given soul is numerically the same soul over time because it traces back to a unique origin; the same holds for an individual atom of matter.

The Problem with Complex Bodies

Things become complicated when we move from simple atoms to complex bodies. Most physical objects — and all living ones — are massive collections of atoms, and those collections are always in flux. Atoms are shed, new ones incorporated, with the net result that the collection today differs from the collection yesterday. Locke’s verdict on this case is stark: if even one atom is added or removed from a complex body, it is strictly speaking no longer the same body. “If one of these atoms be taken away, or one new one added, it is no longer the same mass or the same body.”

This principle, applied consistently, leads to an uncomfortable conclusion: since human bodies shed and gain atoms constantly, my body today is numerically different from my body yesterday. Does this mean I am a numerically different person every day? This is the cliff-edge Locke walks up to at the end of Lecture 1 — and the resolution comes in the lectures that follow, because Locke’s point is precisely that being a person is not the same as being a body.

Personal Identity

Category-Relativity in Practice

The resolution to the “many Me’s” puzzle lies in Locke’s insistence that the category under which you describe something determines the criterion of identity you apply. To illustrate this, consider a deck of cards before and after shuffling. Is it the same thing? If you consider it as a deck — a collection of fifty-two cards — the shuffle changes nothing and the deck is numerically the same. If you consider it as a sequenced collection, the shuffle has produced something numerically different, because the sequence has been altered. The answer depends entirely on the category you apply.

Locke applies the same logic to living things. A growing tree is, considered merely as a mass of atoms, numerically different from day to day. But Locke proposes a different criterion for plants. The identity of a plant consists not in the sameness of its constituent atoms but in something he calls the organization of parts in sustaining a common life. In his words: “that being then one plant which has such an organization of parts in one coherent body, partaking of one common life, it continues to be the same plant as long as it partakes of the same life, even if that life be communicated to new particles of matter.” What matters is not which atoms are there, but whether the same life-sustaining organization persists through the change of parts.

The criterion for animals is essentially the same: a “fit organization or construction of parts to a certain end” — the end of sustaining a biological life. Locke draws an illuminating comparison to a wristwatch: a watch is an assembly of parts working together to sustain a particular function, namely telling time. If you replace a broken part, the watch remains the same watch because the function is being sustained by the new collection of parts. Similarly, an animal remains numerically the same animal through changes of cell and tissue, because the life-function of the organism as a whole continues uninterrupted.

The category of “man” — meaning, in Locke’s (admittedly sexist) terminology, human being understood biologically — follows the same pattern. A man’s identity over time consists in “a participation of the same continued life, by constantly fleeting particles of matter, in succession vitally united to the same organized body.” This is why your body’s daily atomic turnover does not make you a different human being: the biological life continues, and it is that continuity that matters for the identity of human organisms.

What Is a Person?

But for Locke, being a person is something over and above being a human animal. Persons, for Locke, are a distinct category altogether, with a distinct criterion of identity. He defines a person as “a thinking intelligent being, that has reason and reflection, and can consider itself as itself, the same thinking thing, in different times and places; which it does only by that consciousness which is inseparable from thinking.” The defining feature is reflective consciousness — not merely the having of thoughts and experiences, but the capacity to be aware of one’s own thoughts and experiences, both present and past.

This capacity to locate oneself at different times and places through consciousness is what makes personhood distinctive. Locke uses the word “consciousness” to mean something like awareness of one’s own mental states, the ability to focus on and reflect upon the details of one’s experience.

The Memory Criterion

Given this definition of personhood, Locke’s criterion of personal identity follows naturally. He writes: “As far as this consciousness can be extended backwards to any past action or thought, so far reaches the identity of that person; it is the same self now as it was then.” What makes you the same person you were as a child is that you can remember the experiences, thoughts, and actions of that child. The reach of conscious memory is what binds a person together across time.

This is the memory criterion of personal identity, and it has several striking consequences. First, sameness of body and sameness of soul are irrelevant to personal identity on this view — consciousness alone is what matters. Locke illustrates this with the thought experiment of a prince’s consciousness passing into the body of a cobbler upon the cobbler’s death. The resulting individual would be the prince, not the cobbler, because the prince’s memories are intact, regardless of the body they now inhabit. Locke’s concern here may partly be theological — if bodily resurrection involves a different kind of body than the earthly one, it can still be the same person so long as the same consciousness persists.

Second, experiences that are irretrievably lost to memory are literally not yours. If you cannot remember an experience, then on Locke’s account you cannot have been the person who had it. The person who reveled at last night’s party (if you cannot recall the evening) must, strictly speaking, be a different person from you today.

Locke is not entirely troubled by these consequences. He notes that personhood is a forensic concept — it relates to moral and legal responsibility. Praising or blaming someone for actions they have no memory of seems arbitrary, because the memory link is what makes the actions genuinely theirs. The memory criterion thus has a kind of ethical attraction: we are owners of precisely those past actions we can bring to conscious recollection.

Memory and the Underlying Self

Reid’s Critique of Locke

The Scottish philosopher Thomas Reid, writing in 1785 in Essays on the Intellectual Powers of Man, leveled one of the most celebrated objections against Locke’s memory criterion. Reid’s charge is that Locke’s view is not merely false but internally contradictory — that is, the view leads to a reductio ad absurdum.

The argument runs by way of an example involving a brave officer at three stages of his life. At age ten, he was flogged as a schoolboy. At age twenty-five, he captured the enemy’s standard in his first battle. At age sixty-five, he was promoted to General. The additional facts are these: the twenty-five-year-old officer remembers being flogged as a boy; the sixty-five-year-old General remembers capturing the standard as a young soldier; but the sixty-five-year-old General does not remember being flogged as a ten-year-old.

Applying Locke’s memory criterion: the officer at twenty-five is the same person as the officer at ten (because he remembers the flogging). The General at sixty-five is the same person as the officer at twenty-five (because he remembers taking the standard). But the General at sixty-five is not the same person as the boy at ten (because he has no memory of the flogging).

Now apply a logical principle that seems unassailable: the transitivity of identity, which states that if x is identical to y, and y is identical to z, then x is identical to z. By transitivity, since the ten-year-old is identical to the twenty-five-year-old, and the twenty-five-year-old is identical to the sixty-five-year-old, the ten-year-old must be identical to the sixty-five-year-old. But Locke’s criterion directly entails the opposite. The result is a contradiction: on Locke’s own view, the General both is and is not the same person as the schoolboy. Something in the view must be rejected, and Reid is clear that the culprit is the memory criterion.

What Memory Actually Shows, According to Reid

Reid does not think memory is irrelevant to personal identity — he thinks Locke drew the wrong conclusion from the right observation. For Reid, conscious memory provides us with evidence that there is a continuing self, but it does not constitute that self. Memory gives us good reason to believe a self was there at earlier times, but the self that was there is not made of memory.

What is the self, then, on Reid’s account? It is “a permanent being” that underlies all of our changing thoughts, feelings, and actions. Reid writes: “My thoughts, and actions, and feelings change every moment; they have no continued, but successive existence; but that self, or I, to which they belong, is permanent, and has the same relation to all the succeeding thoughts, actions, and feelings, which I call mine.” The self is the subject of all these processes — the entity that has the thoughts — not the collection of thoughts itself. It is what Reid calls a monad: indivisible, admitting of no degrees, changeless in its essence. The self thinks, reasons, deliberates, acts, and suffers — but it is none of these things; it is that which does them.

Critical Reflections on Both Views

Reid’s view has its own vulnerabilities. If the self is truly changeless and indivisible, how can it be the same self that also thinks, reasons, and suffers — processes that seem to involve change? Reid seems to want a self that is both immutable and active, and it is not obvious that these demands are compatible. Moreover, Reid claims people have an “unfailing conviction” in such a being, but it is far from clear that ordinary people believe in an unchanging metaphysical monad — they believe in themselves, but perhaps not in a self of quite that description.

A possible modification of Locke’s view that avoids Reid’s specific objection is to think of memory not as a flashlight that illuminates our past (illuminating only what it directly reaches), but as a chain of links: I remember yesterday’s self, yesterday’s self remembered the day before, and so on, creating an indirect but unbroken connection to even the most distant past. On this model, I am linked to my five-year-old self not by a direct memory reaching all the way back, but by a continuous chain of overlapping memories. This “chain-link” model might be developed to answer Reid’s transitivity objection while preserving memory’s central role.

The Relational Self

Baier’s Naturalist (Relational) View

The debate between Locke and Reid shares a fundamental assumption: both philosophers ground personhood in something internal to the individual. For Locke, it is the internal reach of reflective consciousness; for Reid, it is an internal underlying metaphysical entity. Neither account makes any essential reference to the social or interpersonal world in which persons actually live. This is precisely the assumption that Annette Baier challenges in her essay “A Naturalist View of Persons.”

Baier opens her distinctive perspective with the observation that “our natural habitat, as persons, is among other persons. Our personhood shows in the way we are responsive to one another, responsive to earlier and later generations, responsive to the presence of other groups of persons.” On Baier’s relational view, you cannot fully understand personhood by examining the individual in isolation; persons are constituted in and through their relations with other persons.

What, concretely, does Baier take to be constitutive of personhood? Three kinds of features stand out. First, you were born of specific parents — your biological origin from particular people is an important aspect of who you are as a person, something the Lockean account conspicuously ignores (since your memory does not reach back to birth, your origin is irrelevant to your personal identity on Locke’s view). Second, persons develop the use of language, and in particular the use of language that distinguishes self from other — pronouns like “me,” “we,” “she,” and “they” are markers of a developing person who recognizes themselves as situated in a world of other persons. Third, persons acquire the competencies to participate in communal practices — the “games” we play, from literal games to the social games of making deals, seeking truth, participating in religious rituals, or even conducting conflicts.

All of these hallmarks of personhood involve relationships with other persons. There is no sense in which personhood can be extracted from the broader matrix of social relations and examined in isolation.

Anti-Metaphysical and Constructivist Implications

Baier’s most provocative claim is that “person” is a status term, and crucially, it is our term. She writes: “It is we who have to decide what that status is, and whether we give it to a human fetus, to other animals, to corporations….” For Baier, there is no metaphysical fact of the matter — no hidden feature of reality — that determines which beings count as persons. Instead, personhood is a kind of social construction: a status that human communities bestow on certain beings based on social decisions about which beings deserve the protections and responsibilities that go with personhood. The application of the concept is governed by collective social decisions, not by objective features of the beings in question.

This makes Baier’s view both anti-realist and anti-metaphysical about persons. Where Locke grounds personhood in the metaphysical fact of a certain reach of consciousness, and Reid grounds it in the metaphysical fact of an underlying monadic self, Baier denies that any such facts are doing the real work. The word “naturalist” in her essay’s title signals this anti-metaphysical stance — it is a view that takes persons as they are found in the natural and social world, rather than positing invisible metaphysical underpinnings.

Critical Questions for Baier

Baier’s view raises pressing questions. One concerns the special intimacy each of us has with our own self — a relationship no one else can replicate. My relationship to myself is singular and private in a way that seems to go beyond any social relation. Does a purely relational account of personhood have the resources to explain this intimacy? It is not obvious that it does.

Another concerns the comparison with Locke. Locke too treats personhood as a status (a forensic concept tied to moral and legal responsibility), yet this is fully compatible with grounding that status in metaphysical facts about consciousness. The mere fact that “person” is a status term does not, by itself, show that the status lacks an objective metaphysical basis. Baier’s inference from “status term” to “purely socially constructed” may be too quick.

Finally, we might ask: are you satisfied that your personhood is mainly constituted by your position within social practices and your responsiveness to other persons? Or does there seem to be something more to it — something interior and individual that the relational account leaves out?

Minds and Bodies

The Mind-Body Problem

What the Problem Is

Human beings seem to be, in some irreducible sense, mixed entities: we have bodies that undergo physical processes, and we have minds that undergo mental processes — thought, perception, emotion, decision. The mind-body problem is, at its most general, the metaphysical challenge of understanding the relationship between these two aspects of our nature. A central component of that problem is explaining how mental and physical events causally interact. When I perceive a red light and halt my steps, a perceptual state (mental) causes a bodily movement (physical). When I decide I want ice cream and walk to the freezer, an intention (mental) produces action (physical). These interactions seem obvious enough — but explaining how they occur, given what mind and body each are, turns out to be enormously difficult.

Ontology — the branch of metaphysics concerned with cataloguing what kinds of things exist — is the appropriate framework here. Different answers to the mind-body problem correspond to different ontological commitments.

Dualism holds that the mental and physical are genuinely distinct realms of being. There are (at least) two irreducibly different kinds of thing in the world: minds and bodies, or mental properties and physical properties.

Idealism holds that, despite appearances, there is ultimately only one kind of thing — and it is mental. All of reality is constituted by consciousness or perception. George Berkeley and G.F. Hegel are historical examples; this view is a minority position and will not be pursued here.

Materialism (or physicalism) holds that, despite appearances, there is ultimately only one kind of thing — and it is physical. All of reality, including the mind, falls under the physical category. Mental states, on this view, are in some sense physical states. This is the dominant view in contemporary philosophy of mind and the main alternative to dualism.

Within dualism, a further distinction matters. Substance dualism holds that minds and bodies are literally two different kinds of entity — there are mental substances and physical substances, and human beings are composites of both. Property dualism holds that there may be only one kind of substance (e.g., the human organism is a single physical entity), but that this substance has two irreducibly different ranges of properties — some physical, some mental.

Descartes’s Method of Doubt

The most celebrated defense of substance dualism comes from René Descartes’s Meditations on First Philosophy (1641). Descartes’s project is ambitious: he wants to place all philosophical and scientific knowledge on an unshakable foundation of certainty. The method he adopts — the method of doubt — proceeds by first entertaining every conceivable source of doubt about things previously taken for granted, and then identifying what, if anything, survives that onslaught of doubt.

Three main sources of doubt figure in the Meditations. First, the argument from sensory error: the senses sometimes deceive us (objects at a distance appear distorted, colors shift in poor light, and so on). A source that can deceive should not be unconditionally trusted. Second, the dream argument: when we dream, our experiences feel every bit as real as waking ones, and there are no certain marks by which we can infallibly distinguish the two from the inside. If I might be dreaming right now and not know it, I cannot be certain that my current sensory beliefs are true. Third, the evil genius hypothesis: how do I know I am not permanently deceived by some supremely powerful being — a god or evil spirit — who has engineered my cognitive faculties to go wrong at every step? Even mathematics and geometry, the most seemingly certain disciplines, might be corrupted if such a deceiver exists.

The Cogito and the Essence of Mind

After this sweep of systematic doubt, Descartes finds one proposition that survives: “I think, therefore I am” (cogito ergo sum). Even the most radical doubt requires a doubter. I cannot doubt my own existence without thereby existing as the thing doing the doubting. The act of reflection itself proves the reality of the reflector. And what am I, the certain thing? Descartes answers: “A thing that thinks. What is that? A thing that doubts, understands, affirms, denies, wills, refuses, and that also imagines and senses.” The essence of mind is thinking — thought, in all its varieties, is the defining characteristic of the mental substance.

By the end of the Meditations (after additional arguments, including a proof of God’s existence and God’s non-deceptive nature), Descartes also establishes the existence of physical bodies. The essence of body, he argues, is spatial extension — everything and only physical things occupy space. A mind is an unextended thinking thing; a body is a non-thinking extended thing.

The Argument for Substance Dualism

Descartes’s formal argument for substance dualism runs as follows:

1. I have a clear and distinct conception of myself as a thinking thing.
2. I have a clear and distinct conception of my body as an extended thing.
3. I can conceive of each without conceiving of the other — there is no conceptual entanglement between them.
4. Whatever I can clearly and distinctly conceive as separable must actually be separable, at least by God’s omnipotent power.
5. Therefore, mind and body are really distinct kinds of substance, and I (as a mind) can in principle exist without my body.

This argument has been deeply influential, but it invites a pointed objection: does the fact that we can conceive of two things as separate show that they are separate? Conceivability does not obviously imply real possibility. There may be dependencies between mind and body — causal ones, for instance — that we are simply unaware of and that therefore do not figure in our conceptions of them. Descartes appears to be inferring an ontological conclusion from epistemological premises, and that inferential step is contestable.

Dualist Puzzles and Mind-Brain Identity

The Interaction Problem

Even granting Cartesian dualism, a further problem immediately arises: how can a non-physical mind causally interact with a physical body? Descartes himself acknowledges that they do interact — he writes that hunger and thirst are “confused modes of thinking arising from the union and, as it were, commingling of the mind with the body” — but acknowledgment is not explanation. If minds have no physical properties and bodies have no mental properties, how can one produce effects in the other?

The problem of mind-body interaction is sharpened by the context of modern science. Physics, chemistry, and biology explain physical processes entirely in physical terms. The laws of these sciences make no mention of non-physical properties. Yet dualism requires that a non-physical mind can initiate physical processes in the body. This would seem to violate the causal closure of the physical — the principle that every physical event has a sufficient physical cause. Dualist interaction thus looks less like a scientific phenomenon and more like a miracle: an unexplained intrusion of the non-physical into the physical domain. That is a deeply unsatisfying conclusion for any theory of mind.

Armstrong and the Case for Materialism

D.M. Armstrong’s 1980 paper “The Nature of Mind” argues that if dualist views conflict with the scientific picture of nature, we should simply abandon dualism and seek a materialist alternative. For Armstrong, science is the most successful method of inquiry humans have, producing genuine consensus among practitioners, and any account of mind that contradicts its conclusions is unlikely to be correct.

Armstrong surveys the materialist tradition, beginning with behaviourism. In its simplest form, behaviourism holds that mental states just are overt behaviours — my anger at someone just is the harsh words I direct toward them. But this is obviously too narrow: I can be angry without showing it, think without speaking, fear without fleeing. Mental states can be “occurrent” without manifesting in any observable behavior.

A more sophisticated version, dispositional behaviourism, identifies mental states not with actual behaviors but with dispositions to behave in certain ways under certain conditions. Being angry at someone is having a tendency to act aggressively toward them if appropriate conditions obtain. Like the brittleness of a glass — its disposition to shatter when struck, which exists even when the glass is not being struck — anger can be present even when no aggressive behavior is occurring. This overcomes the simple behaviourist objection.

Armstrong is sympathetic to this dispositional move, but he finds the dispositional behaviourist’s understanding of dispositions itself mistaken. Gilbert Ryle, a leading behaviourist, defined dispositions purely in terms of behavioral tendencies, without positing any underlying state that produces the behavior. For Armstrong, this is exactly backwards. Dispositions are underlying causal states — not mere summaries of behavioral tendencies, but actual internal states of a system that causally give rise to the behaviors in question.

Mind-Brain Identity Theory

Updating behaviourism with the correct understanding of dispositions yields mind-brain identity theory: the view that mental states are physical states of the brain and central nervous system. This is not a denial of mental states but an identification of what they are. To say someone is in pain is to say their brain is in a certain physical state — specifically, whichever brain state causally underlies the behavior characteristic of pain (avoidance, wincing, verbal report of hurt, and so on). To say someone perceives the green traffic light is to say their brain is in whichever state causally underlies their stopping at red and going at green.

This view is a form of reductionism: mental states reduce to physical states. Talking about the mind is, ultimately, talking about the brain — in a vocabulary that tracks the functional causal roles of those brain states in producing behavior.

Pros and Cons of Mind-Brain Identity

Solving the Interaction Problem

One of mind-brain identity theory’s most elegant features is how it dissolves the interaction problem that plagued dualism. If mental states just are brain states, then mind-body interaction is simply a subset of ordinary physical interaction — brain states causing other physical events in the body and vice versa. There is no mysterious boundary between the mental and physical to cross, because there is no such boundary. Human psychology becomes, in principle, continuous with the physical sciences — the approach taken in neuroscience, for instance, is perfectly at home in the identity theorist’s world.

Armstrong on Consciousness

Identity theory faces an obvious worry: if mental states are just brain states underlying behavior, where does consciousness fit in? Armstrong addresses this with a distinctive account. His key example is automatic driving — driving a car without any conscious attention to what you are doing, on “autopilot.” You are executing the mental states required for competent driving, but you are not aware of them. When you suddenly “come to” — perhaps because a horn sounds behind you — you become conscious of your driving. The difference between automatic and non-automatic driving is a matter of the degree of perceptual awareness of one’s own mental states.

For Armstrong, consciousness is a form of inner perception — perception of our own mental states rather than of the external environment. Ordinary perception is “inner states or events apt for the production of certain sorts of selective behaviour towards our environment.” Consciousness is an atypical case of perception in which the object of perception is not the external world but one’s own inner states. When I am conscious of my anger, I am perceiving an internal state of mine, and this perception in turn enables certain further behaviors: I can choose to express the anger, describe it to someone, or suppress it.

Crucially, Armstrong insists that this inner perception is itself a physical state: “consciousness of our own mental state becomes simply the scanning of one part of our central nervous system by another.” The apparent mysteriousness of consciousness dissolves, on his view, once we recognize it as just another variety of physical information processing.

The Chauvinism Objection and Type vs. Token Physicalism

Despite its attractions, mind-brain identity theory in Armstrong’s form — sometimes called type physicalism — faces a serious objection called the chauvinism objection. Type physicalism equates types of mental state with types of brain state: pain = c-fiber stimulation, perception = such-and-such neural activity, and so on. The problem is that this makes it definitionally impossible for any creature lacking human-like brain physiology to have the mental states we have. An octopus, lacking c-fibers, could never be in pain on this view. An alien with a radically different physiology could never think or perceive. Even brain-damaged human beings, who may recover functions by recruiting different brain regions, might fail to count as having the relevant mental states — because the type of brain state in question has changed, even if the function has been restored.

This is “chauvinist” because it displays an irrational prejudice in favour of human neural architecture, treating our particular physiology as constitutively necessary for mentality.

The response to this objection is token physicalism: rather than mapping types of mental state onto types of brain state, we say only that each particular (or token) mental state is identical to some particular physical state, without specifying that all instances of a given mental type must be realized in the same physical type. My pain today is identical to a particular brain state in my nervous system; a Martian’s pain is identical to some particular state of its quite different nervous system; an octopus’s response to noxious stimuli is identical to some state of its neural ganglion. They are all pains, but they are realized differently.

Functionalism

Token physicalism raises a new question: if pain is not defined by any particular brain type, how do we identify mental types at all? The answer is functionalism: mental states are typed by the functional roles they play — by their causal relations to sensory inputs, behavioral outputs, and other mental states. Pain, for instance, has the functional role of being caused by tissue damage and of causing aversion behavior. Whatever state of a physical system plays that functional role, in that system, counts as pain.

The analogy to money is instructive: money cannot be defined by the material it is made of (paper, metal, stone), because many different materials have functioned as money across cultures. Money is whatever plays the economic role of regulating exchanges of goods and services in a community. Similarly, mental states are whatever plays the relevant causal-functional roles, regardless of the physical substrate. This makes functionalism nicely ecumenical: octopi, Martians, and brain-damaged humans can all have mental states if they have states that play the right functional roles.

The Problem of Consciousness

Functionalism and Artificial Intelligence

Functionalism’s indifference to the physical substrate of mental states opens up a philosophically intriguing possibility. If what matters is not the hardware but the functional input-output structure, then in principle a computer could think. If a machine’s hardware states realize the same functional roles we associate with intelligence, perception, and belief, then that machine would possess those mental states. The analogy: brain physiology is to thought as computer hardware is to running a program. Different hardware architectures can run the same program; different physical substrates can realize the same mental function.

This philosophical framework is part of what licenses the research program of artificial intelligence: the attempt to design systems whose hardware states realize the functional roles characteristic of human intelligence. Of course, whether existing or foreseeable computers actually achieve this is a separate empirical question.

Jackson’s Knowledge Argument

The most provocative challenge to all forms of physicalism comes from Frank Jackson’s 1982 paper “What Mary Didn’t Know,” which presents the famous knowledge argument. Jackson asks us to consider a thought experiment:

Mary is a scientist who has spent her entire life confined to a black-and-white room. She is educated through black-and-white books and a black-and-white television screen. Through this education, she learns every physical fact about the world — all the physics, chemistry, neurophysiology, and functional relationships governing color perception, including precisely which neural events occur when people see red, what wavelengths correspond to which colors, what functional roles color perception plays in human behavior. In other words, Mary is physically omniscient about color.

The argument proceeds as follows:

1. Physicalism is the thesis that the actual world is entirely physical.
2. Mary knows everything there is to know about the physical world.
3. Therefore, if physicalism is true, Mary knows everything.
4. But when Mary is finally released and sees color for the first time, she learns something new — she comes to know what it is like to see red. She acquires knowledge she did not have before.
5. Therefore, there must be non-physical facts about the world that her complete physical knowledge failed to capture.
6. Therefore, physicalism is false.

The key premise is (4): that Mary genuinely learns something new upon seeing color. Jackson insists she does — she comes to know the phenomenal character of color experience, a fact about what red looks like from the inside that cannot be captured by any amount of physical description of wavelengths and neural firing rates.

Property Dualism and Qualia

If Jackson is right, the argument points toward property dualism: though the world contains only physical substances, some of the properties of those substances are non-physical. Specifically, the experiential or phenomenal properties of conscious experience — the “what it is like” aspect — seem to be over and above what any physical description captures. These properties are sometimes called qualia: the qualitative, subjective character of experience. The redness of red-as-seen, the painfulness of pain-as-felt — these are qualia.

A common way of accommodating qualia within a broadly physicalist framework is epiphenomenalism: the view that phenomenal properties are causally produced by physical brain processes but have no causal powers of their own. They are by-products of the physical, not themselves part of the physical causal order. But this raises a new version of the interaction problem: if phenomenal properties are caused by physical processes, we face the puzzling question of how something physical produces something non-physical, even if the non-physical thing has no further causal effects.

Critical Questions About the Knowledge Argument

Jackson’s argument is not without vulnerability. One response: perhaps Mary does not learn any new facts when she leaves the room. What she acquires is a new ability — the ability to discriminate and recognize colors experientially — but not any fact that was absent from her physical education. On this view, she gains new knowledge-how (how to recognize red) but no new knowledge-that (no new propositional fact). Another response: we simply do not know whether truly omniscient physical knowledge would leave Mary in the dark about color experience. Given that our actual physical science is nowhere near complete, it is presumptuous to assume that even the most complete physical account must leave out the phenomenal facts. Perhaps, had Mary’s knowledge been complete enough, she could have predicted exactly what it would feel like to see red.

The debate remains open, but it forces physicalists to confront what has become known as the hard problem of consciousness: the difficulty of explaining why there is any subjective experience at all, and not merely functional processing in the dark.

Skepticism and Perceptual Knowledge

The Problem of Perceptual Knowledge

What Is Epistemology?

The transition from metaphysics to epistemology — the theory of knowledge — is marked by a shift in the central question. Where metaphysics asks “what exists?”, epistemology asks “what can we know, and how?” Epistemologists investigate questions such as: What kinds of knowledge do we have, if any? What distinguishes genuine knowledge from mere belief? How can our beliefs be justified? Knowledge, in the philosophical sense, is not just any belief but a belief with an especially strong backing — one that is justified, and (traditionally) true.

Skepticism represents the most radical answer to these questions: the denial that we have any knowledge at all (in the global case), or the denial that we have knowledge of some particular domain (in localized cases). The skeptic holds that our beliefs are never justified in the way knowledge requires, that the gap between appearance and reality can never be closed with certainty.

One important variant is methodological skepticism: the adoption of skeptical arguments not as a final position but as a tool — a rigorous standard of criticism whose point is to identify what, if anything, survives it. Descartes’s Meditations is the paradigm case.

Descartes’s Method and Its Standards

Descartes opens Meditation One by recounting his discovery that much of what he had taken for knowledge was, upon reflection, doubtful. He resolves to raze his belief system to the foundations and rebuild it on certainty alone. His methodological principle: “I should withhold my assent no less carefully from opinions that are not completely certain and indubitable than I would from those that are patently false.” Any belief subject to even the least doubt is treated as false, for the purposes of the project. Only once the dust clears and certainty remains can anything be counted as genuine knowledge.

The argument from sensory error is the first skeptical move. The senses sometimes deceive us — objects at a distance look different from what they are, colors appear different in dim light — and “it is a mark of prudence never to place our complete trust in those who have deceived us even once.” If the senses can err at all, every belief derived from them is potentially suspect. But this argument has a ready response: sensory errors tend to occur in suboptimal conditions (distance, poor light, tiny objects), and we can identify those conditions and correct our errors through further sensory inspection. The senses themselves provide the correction; their occasional unreliability under adverse conditions does not impugn their general trustworthiness under good conditions.

The Dream Argument

Descartes is aware of this response, which is why he escalates to the dream argument — a far more challenging and historically influential skeptical move. He writes: “How often does my evening slumber persuade me of such ordinary things as these: that I am here, clothed in my dressing gown, seated next to the fireplace — when in fact I am lying undressed in bed!… As if I did not recall having been deceived on other occasions even by similar thoughts in my dreams! As I consider these matters more carefully, I see so plainly that there are no definitive signs by which to distinguish being awake from being asleep.”

The crucial point is not that dreams sometimes deceive us (which is trivially true) but that we cannot reliably distinguish waking experience from dreaming experience from the inside. If our waking sensory experiences can be perfectly replicated in a dream — and they can — then for any belief formed on the basis of sensory experience, we cannot be certain it is a waking belief rather than a dream belief. Since dream beliefs are typically false, and since we cannot tell which category a given belief falls into, we cannot be certain of the truth of any sensorily derived belief.

Objections to this argument invite the response that even the tests we apply to determine whether we are dreaming — pinching ourselves, checking for logical coherence — could themselves be part of the dream. There are, as Descartes says, no “definitive signs” by which to settle the question with certainty. Note, importantly, that Descartes is not claiming our whole life might be a dream; he is claiming that even waking beliefs are cast into doubt by our inability to infallibly distinguish them from dream beliefs.

Whether this argument is decisive turns on whether we accept Descartes’s standard: that knowledge requires certainty in the sense of immunity from all possible doubt. If we are willing to accept something less than that — if we think knowledge can coexist with remote or speculative possibilities of error — then the dream argument may lose much of its force.

Skepticism About the External World

The Evil Genius Argument

Descartes does not stop at the dream argument. He notes that even if sensory beliefs are compromised, perhaps certain a priori beliefs — particularly those of arithmetic and geometry — are immune to the dream argument. Mathematical truths seem to hold whether we are asleep or awake. But Descartes devises a more radical skeptical hypothesis to threaten even these: the evil genius argument (also called the evil demon argument).

Suppose there is a supremely powerful being — not a good God but a malevolent genius — who has dedicated all his power to deceiving me in everything I believe. “I will suppose… an evil genius, supremely powerful and clever, who has directed his entire effort at deceiving me. I will regard the heavens, the air, the earth, colours, shapes, sounds, and all external things as nothing but bedeviling hoaxes of my dreams, with which he lays snares for my credulity.” Under the hypothesis of such a deceiver, even 2+2=4 might be something I am programmed to believe falsely.

Descartes is not claiming the evil genius exists. His point is that until we can rule out the hypothesis, no belief is beyond doubt. And since knowledge requires indubitability, we cannot claim knowledge until the evil genius has been definitively defeated.

Descartes’s Solution (Outline)

Descartes’s path out of this skeptical pit is well-known, even if the details are complex: he establishes his own existence (cogito ergo sum), then argues for God’s existence, and then argues that a perfectly good God would not deceive him systematically. If there is a non-deceptive God who created our faculties, we can trust those faculties — at least when we use them carefully, attending only to clear and distinct perceptions. The evil genius is thereby exorcised.

This is a demanding project. The steps Descartes takes are necessary only if one accepts his high standard — that knowledge requires absolute certainty. A less demanding standard might allow us to claim knowledge without having to first prove God’s existence and non-deceptive nature. As the lectures observe, Descartes’s procedure seems excessive to most of us, yet it reflects a principled commitment to not settling for anything less than certainty as the foundation of knowledge.

Hume’s Skepticism: A Different Route to the Same Destination

David Hume’s skepticism about the external world, found in A Treatise of Human Nature (Book I, Part IV, Section ii), takes a strikingly different approach. Where Descartes’s arguments rely on extravagant hypotheses — dreams, evil geniuses — Hume builds his case on an analysis of what the senses ordinarily provide and what our ordinary conception of external objects requires. No fantastical scenarios are needed.

Hume begins by analyzing our ordinary conception of an external object or “body.” Such an object is characterized by two features. First, continued existence: we ordinarily believe that objects continue to exist even when we are not perceiving them. When you look away from the pencil on your desk, you assume it is still there. Second, distinct existence (or mind-independence): we believe that objects exist independently of any consciousness perceiving them — the pencil’s existence does not depend on your seeing it.

The senses, Hume argues, provide us with sensory impressions — sounds, tastes, smells, tactile sensations, visual appearances — that are fleeting and intermittent. Your impression of the pencil lasts only as long as you are looking at it. When you look away, the impression ceases.

The problem, then, is a gap: our ordinary conception of objects attributes to them continued and distinct (mind-independent) existence, but our sensory impressions are intermittent and mind-dependent (they exist only in consciousness). How could the evidence of the senses ever justify the belief that there are objects with properties the senses themselves lack?

More Skepticism and an Answer to the Skeptic

Hume’s Arguments Against External Objects

Hume’s argument concerning continued existence is swift: it would be a “contradiction in terms” for the senses to persuade us of continued existence, because the senses “cannot operate beyond the extent in which they really operate.” Sensation has no content beyond the moment of sensing; it cannot reach beyond itself to disclose a world that persists in its absence.

Regarding distinct existence, Hume considers and rejects two possibilities. The first is representationalism: perhaps sensory impressions function as inner representations of external objects that exist independently of them. On this view, my visual impression of a pencil is a kind of inner picture of an outer pencil, and the outer pencil is what causes the inner picture. But Hume objects: “the senses offer not their impressions as the images of something distinct, or independent, and external, is evident: because they convey to us nothing but a single perception, and never give us the least intimation of anything beyond.” We can judge one thing to be a representation of another only when we can access both the representation and the thing represented. But in perception, we only ever access the impressions — we never get an independent view of the objects they are supposed to represent. All we have are the pictures, never the things being pictured.

The second possibility Hume considers is that sensory impressions themselves somehow present themselves as external to consciousness — that they have a built-in externality-presenting quality. Hume finds this no more satisfactory. For a sensory impression to present itself as external to me, I would need some additional impression of myself to establish the contrast. But as far as the senses are concerned, my body too is just another range of sensory impressions. To appeal to impressions of my body in explaining how impressions present themselves as external would be to argue in a vicious circle. As Hume puts it: “it is not our body we perceive, when we regard our limbs and members, but certain impressions, which enter by the senses; so that the ascribing a real and corporeal existence to these impressions, or to their objects, is an act of mind as difficult to explain as that which we examine at present.”

Hume’s conclusion is total: “the senses give us no notion of continu’d existence, because they cannot operate beyond the extent in which they really operate. They as little produce the opinion of a distinct existence, because they neither can offer it to the mind as represented, nor as original.” The gap between our sensory impressions and our beliefs about external objects is unbridgeable by the senses alone.

What explains our belief in external objects, then? Hume’s answer is the imagination: our minds take the fleeting, gappy impressions the senses provide and project upon them a world of stable, continuing, mind-independent objects. This projection is natural and unavoidable — Hume acknowledges that “we must take for granted” the existence of bodies in all our reasonings — but it is not rationally justified. We believe in external objects as a function of how our minds work, not because our senses have given us rational warrant to do so.

Russell’s Anti-Skeptical Response

Bertrand Russell, in “The Existence of Matter” (from The Problems of Philosophy, 1912), agrees with Hume on several points but draws a very different conclusion. Like Hume, Russell holds that we never directly perceive external objects; what we directly perceive are sense-data — sensory appearances, impressions, qualities as they seem to us. Like Descartes, Russell acknowledges that dreams constitute a source of doubt. But Russell firmly rejects Hume’s skeptical conclusion.

Russell’s strategy is to frame the issue as a choice between two competing hypotheses:

1. Our sense-data generally derive from external physical objects.
2. Only I and my private sense-data exist.

Both are metaphysical hypotheses. The question is which is better rationally supported. Russell considers the appeal to other people’s experiences as evidence for external objects, but rejects it as circular: we know about other people only through our sense-data of them, so we cannot use our knowledge of other people to justify the hypothesis that there are external objects causing our sense-data without already presupposing what we are trying to prove.

Instead, Russell starts from within his own private experience and looks for features of that experience that favor the external objects hypothesis. The key move is inference to the best explanation. Consider sense-data that seem to represent a cat in one corner of the room, followed (with a gap) by sense-data representing the same cat in another corner. The simplest, most economical explanation of this pattern of sense-data is that there is an actual cat that moved from one corner to the other while I was not watching. Without the cat hypothesis, the orderliness of the sense-data would be inexplicable — a mysterious coincidence. Similar reasoning applies to auditory and visual impressions of other people apparently speaking and thinking: the best explanation is that there really are other people there, thinking and speaking.

Even dreams, Russell argues, are best explained on the hypothesis of an external world: “dreams are more or less suggested by what we call waking life, and are capable of being more or less accounted for on scientific principles if we assume that there really is a physical world.” The very content and structure of our dreams presupposes and is shaped by a waking world; the simplest total account of all our experience — waking and dreaming alike — is one that posits an external physical world.

Russell concludes: “Since this belief does not lead to any difficulties, but on the contrary tends to simplify and systematize our account of our experiences, there seems no good reason for rejecting it.”

Evaluating the Debate

Russell’s Partial Agreement and Divergence

Russell’s position is carefully calibrated. He shares Hume’s claim that we do not directly perceive external objects — sense-data are the immediate objects of perception. He agrees that our belief in external objects is an instinctive belief, one we acquire naturally, prior to any explicit reasoning about justification. What he denies is Hume’s inference from “the senses alone cannot prove external objects” to “belief in external objects is irrational.”

For Russell, the senses provide the starting material for a rational inference, even if they do not provide a proof on their own. Once we adopt the approach of inference to the best explanation, our sense-data give us very good reason to believe in external objects — not a deductively certain proof, but a rational inference that the external world hypothesis is more warranted than the solipsistic alternative.

Contrasting Hume and Russell

The deepest contrast between Hume and Russell is not about what the senses do or do not provide — they largely agree on that. The contrast is about what rationality requires.

For Hume, it seems that unless the senses provide something approaching decisive proof of external objects, the belief in them must be irrational. Hume is a demanding critic: he will not call a belief rationally justified unless its warrant is tight and explicit. Since no tight explicit warrant can be drawn from sensory impressions alone, external world beliefs fail this test.

For Russell, the standard of justification is more pragmatic. We are justified in holding a belief if it forms part of the best overall explanatory account of our experience. Justification is not proof; it is a comparative matter — our beliefs about external objects are more justified than the competing skeptical hypothesis, even if neither can be demonstrated with certainty. As Russell puts it: “There can never be any reason for rejecting one instinctive belief except that it clashes with others; thus, if they are found to harmonize, the whole system becomes worthy of acceptance.”

This difference in approach reflects deeper disagreements about the role of philosophy itself. For Hume, philosophy serves as a skeptical critic: it identifies the rational shortcomings of our beliefs, even where it cannot eliminate them (since we are psychologically incapable of abandoning belief in external objects). For Russell, philosophy’s job is to systematize and rationalize our most fundamental beliefs — not to undermine them, but to show how they cohere and what grounds we have for holding them. Philosophy, on this view, works in concert with science and common sense, not against them.

The Question of Standards

A residual question underlies the entire week’s discussion: how demanding should our standards for knowledge be? Descartes holds that knowledge requires absolute certainty — immunity from all possible doubt, however remote. Hume, while not framing his standard quite this way, seems to require something close to demonstrative proof from the evidence of the senses before beliefs can count as rationally justified. Russell is more permissive: good explanatory coherence, systematic harmony with other beliefs, and superiority over competing hypotheses are enough.

The choice of standard is not arbitrary. But it is worth reflecting that Descartes’s standard makes knowledge nearly unattainable (requiring us to first prove God’s existence to establish any knowledge at all), while Russell’s standard makes knowledge more like the best-justified belief system available — fallible, revisable, but genuinely responsive to evidence. That our ordinary knowledge-claims are fallible and revisable, yet still rational and warranted, may be the most defensible middle ground between the Cartesian demand for certainty and the Humean insistence that our beliefs in external objects are nothing more than projections of the imagination.

End of Week 4–6 Notes PHIL 251: Metaphysics and Epistemology, University of Waterloo, Spring 2021

Knowledge Beyond Perception

Knowing What We Don’t Perceive

From Perception to Inference

Earlier in the course, the discussion of perceptual knowledge established that at least some of what we know about the world arrives through the senses — through seeing, hearing, smelling, and feeling the objects around us. That discussion also revealed how philosophically complicated even this apparently simple foundation becomes. Skeptics like Hume and Descartes illuminate the puzzles lurking beneath ordinary perceptual claims, while philosophers like Russell defend more optimistic accounts of how perceptual experience can justify beliefs about an external world.

In Week 7, that perceptual question is set aside — simply granted for the sake of argument — in order to face a different and arguably even more pressing problem: how can knowledge extend beyond what we actually perceive? A great deal of what we ordinarily take ourselves to know consists of things we have never directly observed. This raises a central epistemological puzzle about what entitles us to believe anything beyond the immediate reports of our senses, and the answer involves examining two distinct kinds of inference.

Deductive inference is the more familiar kind. In a valid deductive argument, the premises guarantee the truth of the conclusion — if the premises are true, the conclusion cannot fail to be true. A toy example illustrates this: given that my cat is bigger than my mouse, and that my dog is bigger than my cat, we can infer with absolute certainty that my dog is bigger than my mouse. The key insight is that the conclusion of a deductive argument does not actually say anything beyond what is already packed into the premises; it simply repacks the same information in a new form. This is precisely why deductive inference cannot extend our knowledge to unobserved cases. If the premises of a deductive argument contain only information about observed cases, any conclusion about unobserved cases would require additional information not present in the premises — which would violate the very feature that makes deduction so strong.

Inductive inference, by contrast, is the mode of reasoning that actually carries us from observed to unobserved cases. An inductive argument’s conclusion is not guaranteed by its premises; rather, the premises confer some degree of probability on the conclusion. The classic form runs from observed regularities to broader generalizations or predictions. Every salad I have eaten has nourished me; therefore, the salad I will eat tomorrow will nourish me too. Every material object ever observed has behaved gravitationally; therefore, all material objects behave gravitationally. In each case, the conclusion extrapolates beyond what the premises directly report, and this extrapolation is precisely what makes inductive conclusions less than certain, and precisely what makes them philosophically interesting.

Hume on Causal Reasoning

David Hume, in his Enquiry Concerning Human Understanding (written some ten years after the skepticism-laden Treatise of Human Nature), frames the problem of extending knowledge beyond perception in terms of causal reasoning. For Hume, when we move from observed cases to unobserved ones, we are almost always relying on beliefs about causal relations. The classic example: seeing smoke billowing over a tree line leads us to infer a fire we cannot see, because we believe that smoke of that kind is caused by burning. Similarly, having eaten bread many times in the past and found it nourishing, we infer that the bread on today’s table will nourish us as well.

Hume’s formulation is instructive: he asks what is the nature of all our reasonings concerning “matters of fact” — by which he means unobserved matters — and answers that they are “founded on the relation of cause and effect.” The bread example is particularly illustrative. Our senses tell us about the perceptible qualities of bread — its color, weight, texture. Past experience tells us that bread of those qualities has always nourished us. From this, we confidently expect that today’s bread will do the same: we “foresee, with certainty, like nourishment and support.” But what exactly justifies that step from past cases to future ones?

The Problem of Induction

This question is the problem of induction: the problem of how to justify conclusions about unobserved cases on the basis of cases we have observed. Hume is at pains to stress that he does not dispute the practical indispensability of inductive reasoning — “none but a fool or a mad man will ever pretend to dispute the authority of experience.” The problem is specifically philosophical: what rational justification can be given for the inferential leap from observed to unobserved?

Hume’s Problem of Induction

Causation and Experience

Hume insists that our causal beliefs — our beliefs that fire causes smoke, that bread causes nourishment — are grounded not in any perception of a mysterious “necessary connection” between cause and effect, but solely in past experience of constant conjunction. We have always seen smoke accompany fire; we have always seen bread nourish us. This repeated conjunction is both the metaphysical basis and the epistemological basis for our causal beliefs. Yet this immediately generates a problem: our experience is limited, covering only cases we have actually encountered. The conclusion we wish to draw, however, extends far beyond those cases. Hume puts the difficulty precisely: “As to past Experience, it can be allowed to give direct and certain information of those precise objects only, and that precise period of time, which fell under its cognizance: but why this experience should be extended to future times, and to other objects, which for aught we know, may be only in appearance similar; this is the main question on which I would insist.”

Consider pool balls. Having observed how billiard balls interact on past tables, we expect that balls on the next table — similar in shape, weight, and texture — will behave the same way. But how do we assure ourselves that apparent similarities in sensible qualities guarantee similarities in causal powers? Hume’s point is that causal powers are not among the properties we directly perceive; we perceive only what he calls “sensible qualities.” The uniformity of underlying causal powers over time and across instances is just the thing that needs to be justified, not something we can read off from observed similarities.

The Uniformity Principle and Its Circularity

Hume identifies the implicit premise underlying all inductive inference: what he calls (and what is generally known in the literature as) the principle of the uniformity of nature — roughly, the claim that “the future will be conformable to the past,” or that regularities observed in the past will persist into the future. If we could legitimately add this principle as a premise to any inductive argument, we could, in principle, turn the inductive argument into a deductive one. The general argument would then run: all observed A’s are B’s (from experience); the future will resemble the past (the uniformity principle); therefore, all A’s are B’s, including future ones (by deduction from those two premises).

But this tidy solution merely relocates the problem. Now we must ask: what justifies the principle of uniformity itself? It is not a necessary truth — denying it involves no contradiction, unlike denying “all bachelors are unmarried.” Nor can it be known a priori, since how we could know, purely from conceptual analysis, that nature will remain uniform? If the principle is to be justified at all, it must be on the basis of past experience — but then we would be arguing inductively for the very principle that is supposed to justify inductive inference. This is the circularity objection, or what Hume regards as vicious circularity: we cannot justify induction by appealing to induction without begging the question.

Even the attempt to appeal to causal powers — inferring that fire has the power to produce smoke, from the repeated conjunction of fire and smoke — runs into the same wall. As Hume himself says, “all inferences from experience suppose, as their foundation, that the future will resemble the past, and that similar powers will be conjoined with similar sensible qualities.” Any such inference already assumes what it needs to prove.

Hume’s Skeptical Conclusion and Positive Account

Hume’s conclusion is clear-eyed and, for most people, deeply unsettling: there is no rational justification for inductive inference. We cannot, as philosophers, provide any non-circular grounds for believing that the future will resemble the past. And yet, practically speaking, we have no choice but to make inductive inferences — “none but a fool or a mad man” would refuse. Hume’s positive account of why we do in fact make such inferences is not a justification but an explanation. Repeated experience of past regularities breeds in us a certain expectation — a “habit” or “custom” of mind — that the regularities will persist. This expectation is a stable feature of human psychology, not a product of reasoning. The sun has risen every day; we expect it to rise tomorrow, not because we can justify this expectation rationally, but because habit has conditioned us to form it.

This move is significant: Hume effectively replaces the traditional philosophical ambition of showing how our beliefs are rationally justified with the more modest psychological question of how we come to form the beliefs that we do. As the course will show in the following week, this Humean “turn” has a twentieth-century descendant in what Quine calls naturalized epistemology.

Foster’s Response — The Nomological-Explanatory Solution

Setting Up the Solution

John Foster, in his paper “Induction, Explanation and Natural Necessity,” disagrees strongly with Hume’s pessimistic conclusion and sets out to provide a genuine solution to the problem of induction. The key move is to argue that extrapolative induction — the kind of inference Hume was analyzing, where observed regularities are simply projected onto unobserved cases — is not the fundamental form of empirical inference at all. For Foster, a more basic form of inference underlies our reasoning from experience: inference to the best explanation (IBE).

Foster’s approach can be outlined as follows. The observed regularities of nature call for explanation. When we observe that every material object we have ever encountered behaves gravitationally, that smoke always accompanies fire, and that water consistently has the chemical composition H₂O, these regularities are striking enough to demand an explanation. There are two broad kinds of explanation on offer: either the regularities are purely accidental — just coincidences, with nothing underlying them — or they express a form of natural necessity, a law of nature that makes the regularity not merely actual but inevitable. Foster argues that the latter explanation is invariably better, and that we are rationally entitled to infer it.

Two Illustrative Examples

The water example shows how Foster’s account differs from the Humean one. A Humean account would say scientists simply observed many water samples and extrapolated inductively to the generalization that all water is H₂O. Foster argues that what actually happened was more sophisticated: scientists subjected samples to chemical tests and asked what hypothesis, in conjunction with the known laws of chemistry, best explained how those samples responded. The best explanation was the general hypothesis that water’s composition is H₂O — not a mere inductive extrapolation, but an inference to an explanatory hypothesis grounded in scientific laws.

The gravitational example is even more illuminating. Every material body ever observed has behaved gravitationally: objects fall, tides flow, planets orbit. Why? The accidental explanation says this is just how things happened to be, with no deeper ground. The law-based explanation says it is a law of nature that bodies behave gravitationally — a law that expresses natural necessity. Foster argues that the accidental hypothesis is “quite literally incredible,” because the regularity is so sweeping, so invariable, and covers so many independent instances that its being a mere coincidence would be, as he puts it, “an astonishing coincidence.” The explanatory inference to a natural law is therefore more rational, and once the law is accepted as expressing natural necessity, the conclusion that bodies will behave gravitationally in the future follows deductively: if they must behave gravitationally, then they will.

The Nomological-Explanatory Solution (NES)

Foster calls his full account the nomological-explanatory solution (NES) — “nomological” meaning “pertaining to law.” The structure of the solution has three components. First, the only primitive rational form of empirical inference is inference to the best explanation; extrapolative induction is not fundamental. Second, what appears to be an inductive inference is actually reducible to two more basic steps: (a) an inference from an observed regularity to its best explanation (typically given in terms of laws of nature), followed by (b) a deductive inference from that explanation to the conclusion that the regularity will persist in future cases. Third, the laws that figure in best explanations are not mere generalizations — they express natural necessity, and it is this necessity that licenses the deductive step from law to future persistence.

The crucial advantage over Hume is that, on Foster’s account, the step from past regularities to future ones is not a mere inductive extrapolation that requires the uniformity principle as an unsupported premise. Instead, it is grounded in a deductive inference from a law of nature — and the law is justified by inference to the best explanation, not by circular appeal to induction.

Evaluating the Induction Debate

Ayer’s Objection and Foster’s Reply

Foster opens his paper by addressing an objection from A.J. Ayer. Ayer argues that a generalization involving natural necessity — a claim that all A’s must be B’s — is harder to support on the evidence than a mere inductive generalization that all A’s are B’s. If the evidence makes it difficult to justify even the modest inductive generalization, it must be all the harder to justify the stronger claim that involves necessity. Therefore, appealing to laws of nature embodying natural necessity is not a solution to the problem of induction; it is an even greater problem.

Foster’s reply is that this objection holds only if the kind of inference available to us is extrapolative induction. If all we can do is extrapolate from observed to unobserved cases, then yes, adding necessity makes the claim harder to support. But inference to the best explanation is a different kind of inferential move altogether. IBE takes observed regularities as evidence for an underlying explanatory hypothesis, and such explanations typically do invoke laws and natural necessity. The transition from evidence to explanation in IBE is not constrained in the way Ayer’s criticism assumes.

The Scope Objection

Even granting IBE, a further question arises: why should a law inferred to explain some limited range of observations be understood as applying universally — at all times and places — rather than only within the observational range in which the regularity was detected? Foster frames this himself in terms of gravity: why suppose the law of gravity covers all bodies at all times, rather than all observed bodies up to some particular time t? On the restricted reading, there would be no basis for predicting gravitational behavior beyond time t.

Foster’s response is that a law restricted to some particular time t would introduce a new explanatory mystery: why should natural necessity operate only up to t and then stop? The very point of explanation in terms of natural laws is that it removes the appearance of coincidence and replaces it with something systematic. Artificially restricting a law to a particular time or set of objects reintroduces arbitrariness into the explanation, weakening its explanatory force. Scientific practice bears this out: when laws appear limited in scope, scientists search for deeper, more general principles that explain both the original regularities and the apparent limitations. The more general law is invariably the better explanation.

Does Foster Answer Hume?

There are reasons for skepticism here. Foster’s account presupposes a realist conception of natural necessity — the idea that there genuinely are necessities in nature that enforce regularities. Hume, of course, famously denies this. For Hume, the very idea of necessary connection “exists in the mind, not in objects.” What we call necessity is just a projection of our habitual expectations — feelings of anticipation bred by repeated experience — onto the world. If Hume is right about this, Foster cannot simply invoke natural necessity as though it were an independently available explanatory resource; he would first have to refute Hume’s anti-realist account of necessity, and that is a substantial metaphysical task.

Moreover, Foster’s endorsement of the tendency of science toward greater and greater generality might simply presuppose, rather than prove, that the fundamental regularities of nature are genuinely universal and time-invariant. Hume’s original challenge — how can we be justified in supposing that the future will be conformable to the past? — can be reformulated as: how do we know that the laws we infer from past observations really are genuinely universal laws, rather than generalizations that happen to have held in our limited experience so far? Foster’s answer seems to rely on the explanatory superiority of universal laws over restricted ones, but that superiority may itself depend on an assumption that nature is fundamentally orderly — an assumption that seems to beg the very question Hume raised.

Despite these concerns, Foster’s view captures something genuinely important: it models actual scientific practice far better than a naive extrapolative inductivism does. Scientists do not simply enumerate cases; they build theoretical structures that explain regularities, and those structures do license predictions far beyond observed cases. Whether this vindicates Foster against Hume, or merely describes our actual epistemic practices without fully justifying them, is a question the lectures leave productively open.

Naturalizing Epistemology

Logical Empiricism — Bridging the Gap

The Justification Gap

A recurring structural pattern runs through the skeptical arguments considered in earlier weeks: first, note the limited character of our experiential evidence; then argue that the beliefs and theories we endorse go well beyond what that evidence supports; conclude that those beliefs are unjustified. This is what the lectures call the justification gap — the apparent disproportion between our meager experiential evidence and the expansive theories we build upon it. Descartes saw the gap with respect to our beliefs in external objects. Hume saw it with respect to inductive generalizations and predictions about the future. The same gap is visible wherever a philosophy of science tries to understand the relationship between observation and theory.

Different philosophical traditions propose different strategies for bridging this gap. Rationalism, associated with Descartes and others in the seventeenth and eighteenth centuries, holds that the mind is supplied with innate ideas and principles of reasoning that supplement the deficiencies of sensory experience. If God, who is perfect, would not have created us to be deceived in our most fundamental beliefs, then at least some of our natural convictions about the world may be trusted. But rationalism has declined in influence since the rise of empirical science, which has pushed epistemology decisively in an empiricist direction — toward the view that all knowledge ultimately originates in experience. Inference to the best explanation, as seen in Russell and Foster, offers another strategy: theories about material objects or laws of nature are inferred as the best explanatory accounts of our experience, even if experience cannot directly verify them.

The Logical Empiricist Program

The third strategy — the one featured in Week 8 — is logical empiricism, or logical positivism. A.J. Ayer’s Language, Truth and Logic (1936) is the principal English-language exposition of this movement, which originated among German-speaking philosophers in Vienna during the early twentieth century. The central idea of logical empiricism is strikingly bold: any meaningful theoretical statement about the world can be reduced to — that is, translated into — statements about direct sensory experience. If this reduction is successful, there is no gap between theory and experience, because theorizing about the world just turns out to be a refined and systematic way of talking about experience. Material objects, on this view, are what Ayer calls logical constructions out of sense-contents: the statements we make about tables and chairs are logically equivalent to complex statements we could make about the patterns of visual, tactile, and other sensory experiences that characterize our interactions with those objects.

The philosophical technique Ayer employs is what he calls definitions in use (and what Quine calls contextual definition). Unlike an explicit definition — which defines a term by providing synonymous terms — a definition in use defines a term by specifying equivalent sentences that, however, do not themselves contain the term. The statement “there is a brown table in the living room” is translated into a (potentially very long) set of statements about sense-contents: a certain brown visual patch occurring in a certain spatial configuration, associated with certain tactile sensations of solidity, and so on across the range of sensory modalities that a table-encounter involves. The translation, however impractical to state in full, conveys the logical meaning of the original claim.

The broader philosophical context of this program is the Verificationist theory of meaning: the meaning of any empirical statement just is its verification conditions — the range of observations that would confirm or disconfirm it. Theoretical claims in physics, chemistry, or ordinary life are meaningful precisely because and insofar as they can be connected to specific observational evidence. The definitional reduction is what accomplishes this connection. If logical empiricism succeeds, the justification gap simply disappears: we do not need to explain how experience justifies theory, because theory just is experience, re-described in a more systematic vocabulary.

Quine’s Naturalized Epistemology

Quine’s Critique of Logical Empiricism

W.V. Quine’s “Epistemology Naturalized” opens with a critical history of the logical empiricist program. Quine identifies two sides to the reductionist enterprise — a conceptual side (clarifying what our claims in mathematics and empirical science mean) and a doctrinal side (establishing how we can know those claims to be true) — and argues that the program failed on both.

In the mathematical branch, the hope was to reduce all mathematical truths to logical truths, thereby grounding mathematical certainty in the undeniable certainty of pure logic. The project came close, but could not complete the reduction without supplementing pure logic with set theory — which is itself a mathematical theory, and which is no more obviously certain than ordinary arithmetic. As Quine dryly notes, “the firmness and obviousness that we associate with logic cannot be claimed for set theory.” The reduction lost the epistemological advantage it was supposed to provide.

The failure in the empirical branch was, if anything, more dramatic. The attempt to translate theoretical claims about the physical world into claims purely about sense-data ran into at least two serious obstacles. On the conceptual side, the translations required not just sense-data but sets upon sets of sense-data, and they could not adequately handle the spatial-temporal indexing that theoretical claims involve. On the doctrinal side — and here Quine invokes a recognizably Humean point — even a perfectly successful translation of an empirical claim into the language of sense-data would not suffice to show that we can know the claim to be true. Generalizations, however reduced, always cover more cases than any finite range of observations can verify. “The Humean predicament is the human predicament,” Quine remarks. The whole Cartesian project of grounding knowledge in an unshakeable sensory foundation turns out to be, in Quine’s words, “a lost cause.”

Naturalized Epistemology as a Replacement

Having declared the old epistemology a failure, Quine proposes something genuinely radical: rather than persisting in the attempt to show how experience can justify our theories, epistemology should simply study, as a branch of natural science, how experience causally produces our theories. “The stimulation of his sensory receptors is all the evidence anybody has had to go on, ultimately, in arriving at his picture of the world,” Quine observes. “Why not see how this construction actually proceeds? Why not settle for psychology?”

On this view, epistemology becomes a chapter of empirical psychology. The human subject receives sensory inputs at the sensory surfaces — the eyes, ears, skin — and produces theoretical outputs: beliefs about objects, scientific theories, a “picture of the world.” What naturalized epistemology investigates is the psychological and neurological mechanisms that mediate between this “meager input” and the “torrential output” of our theories. The relationship between evidence and theory remains an object of study, but the question has shifted from “how does evidence justify theory?” to “how does evidence cause theory?”

The relationship between this Quinean approach and Hume’s positive account of induction is not coincidental. Hume, recall, ultimately explained our tendency to make inductive inferences not by showing that they are rationally justified, but by appealing to habit and custom — the psychological mechanisms by which repeated experience breeds expectation. Quine, in effect, proposes to pursue that explanatory, naturalizing approach across the whole of epistemology.

The Circularity Objection and Quine’s Response

Quine himself raises the most pressing objection to his program: if the goal of epistemology is to validate or justify empirical science, then using psychology — itself an empirical science — as the medium of that validation is viciously circular. As Quine puts it: “If the epistemologist’s goal is validation of the grounds of empirical science, he defeats his purpose by using psychology or other empirical science in the validation.”

His response is to concede the circularity while denying that it is vicious. The circularity would be problematic only if one were still pursuing the goal of the old epistemology — showing how experience deductively validates or guarantees our theories. But Quine’s naturalized epistemology has abandoned that goal. If we are “out simply to understand the link between observation and science,” rather than to justify science from some external standpoint, we are free to use any information available, including information provided by science itself. Since justification is no longer the aim, the circle is not vicious. Science investigates itself, as it were, from within, and there is nothing inherently problematic about this, any more than there is something problematic about a biologist using biological knowledge to study the biological processes of biological knowing.

Kim’s Critique of Naturalized Epistemology

The Centrality of Justification

Jaegwon Kim’s “What is ‘Naturalized Epistemology’?” mounts a pointed critique of Quine’s program. Kim begins by characterizing what traditional epistemology was actually about. Its central agenda, he argues, involved two interrelated questions: what conditions must a belief satisfy in order to be justified?, and what specific beliefs are we in fact justified in holding? Both questions centrally feature justification — and that, Kim insists, is no accident.

Justification’s centrality to epistemology is grounded in the standard tripartite analysis of knowledge, which was touched on briefly in Week 8 and will be explored in depth in Week 9. On this analysis, a person S knows that P if and only if (1) S believes that P, (2) P is true, and (3) S’s belief that P is justified. Kim observes that the first two conditions — belief and truth — are not distinctively epistemic notions. Belief is a psychological concept: I can believe all manner of things that I do not know. Truth is a semantic or metaphysical concept: a proposition can be true whether or not anyone knows or believes it. Justification is the only component of the tripartite analysis that is intrinsically epistemic — it is what transforms the combination of true belief into genuine knowledge, by requiring that the belief be appropriately supported rather than merely accidentally correct.

Epistemology as a Normative Discipline

Kim further characterizes traditional epistemology as a normative discipline. To call a discipline normative is to say that it concerns standards of correctness — not merely descriptions of what happens, but evaluations of whether what happens meets appropriate criteria. Normative ethics asks not merely what people do but what they morally ought to do; similarly, epistemology asks not merely what people believe but what they ought to believe, given appropriate evidential standards. Kim puts it directly: “If a belief is justified for us, then it is permissible and reasonable, from the epistemic point of view, for us to hold it, and it would be epistemically irresponsible for us to hold beliefs that contradict it.”

This normative dimension — the dimension of justification — is precisely what Kim accuses Quine’s naturalism of abandoning. Quine’s proposal to replace epistemology with a psychological study of the causal transition from sensory inputs to theoretical outputs is, for Kim, simply a change of subject. A descriptive causal science tells us how beliefs arise, not whether they are worth holding. As Kim memorably puts it: “For epistemology to go out of the business of justification is for it to go out of business.”

A related complaint concerns the concept of an evidential relation. Kim grants that naturalized epistemology studies a relation between evidence (sensory input) and theory (output). But the relation it studies is a causal one — a question of how sensory stimulation causes belief-formation. And a causal relation between evidence and theory is not the same as an evidential relation. Understanding that a certain type of sensory input causally produces a certain type of belief tells us nothing about the extent to which the input rationally supports the belief. The normative, justificatory dimension of the evidence-theory relationship disappears entirely on a purely causal reading.

Evaluating Naturalized Epistemology

Advantages and Disadvantages

The debate between Quinean naturalism and traditional epistemology is not straightforwardly one-sided. Both approaches have genuine merits and genuine liabilities.

Naturalized epistemology’s chief advantage is its immunity from skeptical arguments. Traditional epistemology, committed to showing how experience justifies theory, must answer the skeptic who denies that it does. This is a demanding requirement, and as the weeks preceding this module showed, meeting it is genuinely difficult. Naturalized epistemology simply sidesteps the demand: if justification is not the goal, there is no obligation to refute the skeptic. This means naturalized epistemology is free to work from within our ordinary and scientific assumptions about the world, employing those assumptions in the study of how sensory experience gives rise to belief — without first having to certify those assumptions from some skeptic-proof external standpoint.

Traditional epistemology’s chief advantage, however, is its capacity for normative evaluation of beliefs. We make distinctions between well-supported scientific beliefs and groundless superstitions, between careful reasoning and wishful thinking, between knowledge and lucky guessing. These distinctions matter, and they are normative: to say one belief is better supported than another is to apply evaluative standards. Traditional epistemology, with its focus on justification, has the resources to honor these distinctions. If Kim’s reading of Quine is right, naturalized epistemology does not.

The lectures end the week with a genuinely open question: might there be a way to preserve the best of both approaches — the freedom from skeptical pressure that naturalism offers, while retaining some normative conception of justification? This question points directly toward the work of Alvin Goldman in Week 9, whose causal theory of knowledge can be understood as one attempt to combine a naturalistically respectable account of knowledge with a normative dimension — the normative dimension being grounded not in internally accessible justificatory grounds, but in objective causal connections between belief and fact.

The Analysis of Knowledge

The Traditional Analysis of Knowledge

What a Philosophical Analysis Does

Week 9 opens by returning to the tripartite analysis of knowledge — already invoked in Week 8 in Kim’s discussion of Quine — and examining it on its own terms. This analysis, which Jaegwon Kim calls the “traditional tripartite analysis,” holds that a person S knows that P if and only if: (1) S believes that P, (2) P is true, and (3) S’s belief that P is justified. It is important to understand what kind of claim this analysis makes and what it does not make.

A philosophical analysis of this kind is not a dictionary definition. Dictionaries define “knowledge” in terms of familiarity, learning, or understanding — useful for everyday linguistic purposes but far too imprecise for philosophical inquiry. A philosophical analysis, by contrast, specifies the necessary and sufficient conditions for knowledge. Each condition is necessary in the sense that if any one of the three is not met, the person does not have knowledge. Together, they are sufficient in the sense that if all three are met, there is nothing further that knowledge requires — the person genuinely knows. The analysis is deliberately abstract: it applies to any genuine case of propositional knowledge, without specifying the details of any particular situation.

The kind of knowledge the analysis targets is propositional knowledge — knowledge that something is the case, expressed as “I know that the cat is on the mat,” “I know that 2+2=4,” and so on. This is to be distinguished from know-how (knowing how to ride a bike) and knowledge by acquaintance (knowing a person or a place). Epistemology has historically focused on propositional knowledge, partly because it is the kind most clearly subject to questions of justification and evidence.

Ayer’s Version

The A.J. Ayer who appears in Week 9 is a notably different philosopher from the Ayer of Week 8. The earlier Ayer wrote Language, Truth and Logic in 1936, a manifesto for logical empiricism. The later Ayer, writing The Problem of Knowledge some thirty years later, has abandoned that program and approaches knowledge more pragmatically. His version of the tripartite analysis modifies two of the three conditions. Where the standard version speaks of “belief,” Ayer speaks of being “sure” — because knowledge requires not merely any degree of belief but a sufficiently confident one. And where the standard version speaks of “justification,” Ayer speaks of the right to be sure — a phrase that captures both the normative character of the condition (being sure is one thing; being entitled to be sure is another) and its connection to practice.

The question of what confers the right to be sure connects the analysis to the ordinary practice of answering “How do you know?” When asked this question about empirical claims, we typically cite perception, memory, testimony, scientific laws, or historical records. These are the common bases on which we ordinarily claim the right to be sure. Ayer’s point is that in specific cases — a sharp memory of a recent event, testimony from a known expert — we can generally assess whether someone has the right to be sure. What we cannot do is enumerate a complete, general list of conditions covering every case: “to say in general how strong it has to be would require us to draw up a list of conditions under which perception, or memory, or testimony, or other forms of evidence are reliable. And this would be a very complicated matter, if indeed it could be done at all.”

This is a crucial limitation. The tripartite analysis does not by itself resolve every question about knowledge — it is a general framework, not a decision procedure. Ayer illustrates this with a puzzle: what about the person who consistently and successfully predicts the outcomes of lotteries, without being able to explain their own predictive success? We might grant that person knowledge (on the grounds that their track record establishes the right to be sure), or deny it (on the grounds that the predictions lack any of the recognized bases for justified belief). The analysis does not settle this, and Ayer does not think it needs to: the point is that a general analysis provides the right framework for epistemological inquiry even if it does not mechanically answer every individual question.

Gettier’s Challenge to the Traditional Analysis

A Short Paper, a Major Revolution

Edmund Gettier’s 1963 paper “Is Justified True Belief Knowledge?” is among the most influential short papers in twentieth-century philosophy. It amounts to a direct challenge to the traditional tripartite analysis: Gettier’s thesis is that the three conditions are not jointly sufficient for knowledge. That is, there can be cases where someone has a justified true belief and yet clearly does not know the proposition in question. Gettier is careful to state the limits of his challenge: he is not denying that the three conditions are individually necessary for knowledge. He is specifically targeting the sufficiency claim — the idea that meeting all three conditions is enough to count as knowing.

His strategy is to construct scenarios — Gettier cases or Gettier counter-examples — in which all three conditions are satisfied, but which are clearly not cases of knowledge. The paper prompted an entire industry of attempts to revise or replace the traditional analysis, making the “Gettier problem” one of the central topics of epistemology in the decades following its publication.

A First Gettier Case

Consider a simplified Gettier case not from the paper itself, but which captures the spirit of it. Smith has strong evidence for the proposition that Jones owns a Ford: his friend Brown, who is generally reliable and honest, has told him that Jones has always owned a Ford. Smith justifiably believes “Jones owns a Ford” on the strength of this testimony. But on this particular occasion, Brown has confused Jones with someone else. Jones has never previously owned a Ford. However, entirely coincidentally, Jones purchased his very first Ford just yesterday.

Let us check the three conditions. Smith believes that Jones owns a Ford: yes. “Jones owns a Ford” is true: yes, because Jones bought one yesterday. Smith’s belief is justified: yes, since testimony from a generally reliable source is a standard basis for justification. All three conditions are satisfied. And yet it seems entirely wrong to say that Smith knows that Jones owns a Ford. He arrived at the right belief, but for the wrong reasons: his evidence was based on false information about Jones’s past ownership. The fact that his belief is true is a lucky coincidence that has nothing to do with the evidence that led him to hold it. When the justification leads one to a true belief through a false intermediate step, the connection between justification and truth is severed in a way that seems incompatible with genuine knowledge.

Gettier’s Second Case

Gettier’s second counter-example (Case II in the paper) is more elaborate and makes use of the logic of disjunctive propositions. Smith again has strong evidence for believing that Jones owns a Ford — this time because, as far back as he can remember, Jones has always owned Fords, and Jones recently gave him a ride in a Ford. From this justified belief, Smith makes a deductive inference to the disjunctive proposition: “Either Jones owns a Ford, or Brown is in Barcelona.” This inference is valid: in formal logic, if you know P, you may validly infer “P or Q” for any Q whatsoever, since a disjunction is true whenever at least one of its disjuncts is true.

Now it turns out that the Ford Jones was driving was a rental, and Jones does not in fact own a Ford. But Smith has a friend named Brown, of whose whereabouts he is “totally ignorant” — and it happens that Brown is in fact in Barcelona. Therefore, the disjunctive proposition “Either Jones owns a Ford, or Brown is in Barcelona” is true, because Brown is in Barcelona. Smith believes it (he inferred it), it is true (because of Brown’s location), and he is justified in believing it (by the standards of his justified belief in the first disjunct, and the logical inference). All three conditions are satisfied. But again, clearly Smith does not know this disjunctive proposition. The fact that makes it true — Brown being in Barcelona — played no role whatsoever in Smith’s coming to believe it. He believed it on the basis of a false belief about Jones, and it just happened to be true by coincidence. The accidental character of the truth renders the case one of luck, not knowledge.

These cases reveal a structural problem with the traditional analysis: it does not adequately capture the requirement that the justification should be connected to the truth in the right way. Having justification and having a true belief are not enough if the truth is, so to speak, floating free of the justification.

Goldman’s Causal Theory of Knowing

Diagnosing the Gettier Problem

Before turning to Goldman’s own account, it is worth noting an initial diagnosis of the Gettier cases that, while tempting, ultimately fails. One might argue that the problem in each case is that Smith’s evidence is itself false: Brown’s testimony about Jones’s prior ownership is false; Smith’s belief that Jones owns the Ford he was driving is false. If we require that a genuine justification involves only true evidence — no false lemmas — then the Gettier cases would not satisfy the justification condition after all, and the traditional analysis would be saved.

The difficulty is that requiring all evidence to be true sets the bar for justification implausibly high. Smith is not in a position to verify the truth of his evidence independently of the evidence itself. If we could always access the truth directly, we would not need evidence in the first place. Evidence is valuable precisely because the truth is not transparent to us; we use evidence as a guide. To demand that the guide be infallible is to demand something beyond what the concept of justification can realistically require.

Goldman’s Diagnosis

Alvin Goldman’s “A Causal Theory of Knowing” agrees with Gettier that the traditional analysis fails and accepts the counter-examples as genuine cases of justified true belief that are not knowledge. Goldman’s own diagnosis fastens on what is missing: an appropriate causal connection between the belief and the fact that makes the belief true.

In the second Gettier case, Smith believes “Either Jones owns a Ford, or Brown is in Barcelona,” and what makes this true is the fact that Brown is in Barcelona. But as Goldman observes, “there is no causal connection between the fact that Brown is in Barcelona and Smith’s believing P.” Smith’s belief arose from misleading evidence about Jones; Brown’s location played no causal role in producing Smith’s belief. The fact that makes the belief true and the belief itself are causally unconnected — and Goldman argues it is this disconnection that disqualifies the case as knowledge.

Goldman’s proposed revision to the traditional analysis is accordingly: S knows that P if and only if the fact that P is causally connected in an appropriate way with S’s believing that P. This replaces the justification condition with a causal condition, or more precisely, adds the causal requirement as what justification was always supposed to track.

Types of Appropriate Causal Connection

Goldman recognizes that appropriate causal connections come in various forms and can be quite complex. The clearest case is perceptual knowledge: when I see the cat on the mat, the fact of the cat’s being on the mat causes a series of physical and perceptual processes that terminate in my believing that the cat is on the mat. The belief is causally tied to the very fact that makes it true, through the perceptual mechanism.

Memorial knowledge involves a similar structure, extended across time. If at time t₁ I perceive that the cat is on the mat, and at a later time t₂ I remember that the cat was on the mat, there is a causal chain from the original fact (the cat’s presence at t₁), through the perceptual event, through the memory mechanism, to my current belief. The causal connection between fact and belief is intact, if more complex than in the purely perceptual case.

Inferential knowledge is illustrated by Goldman’s volcano example. Suppose S perceives solidified lava in the countryside and, using background knowledge about how lava is produced, infers that a nearby mountain must have erupted long ago. Whether this constitutes knowledge depends on whether S’s belief is causally connected to the fact of the eruption. If the lava S perceives was actually produced by that eruption, then there is a causal chain from the eruption, through the lava, through S’s perception, through S’s inference, to S’s belief — and S knows. But if the original lava was cleaned away and replaced by a man who wanted to simulate the appearance of an eruption, the causal chain between the actual eruption and S’s belief is broken; S’s belief happens to be true but is not causally connected to the relevant fact — and S does not know. The case is structurally analogous to a Gettier case: true, justified belief, but without the right causal connection.

Goldman also accommodates what he calls Pattern II cases, where the fact that makes the belief true is not the cause of the belief, but both the belief and the fact share a common cause. A future-directed example: my friend T intends to go downtown tomorrow and tells me of his intention today, causing me to believe he will go downtown. T’s intention is also what will cause him to actually go downtown. So both my belief and the fact that makes it true (T’s going downtown) are caused by T’s intention; they are causally connected through this common cause. Goldman argues that such cases qualify as knowledge, even though the future event that makes the belief true obviously cannot cause a present belief about it.

Externalism, Naturalism, and the Causal Theory

Internalism and Externalism

Goldman’s causal theory exemplifies a general position in epistemology known as externalism. The distinction between internalist and externalist theories of knowledge runs as follows. Internalist theories hold that the grounds that make a belief qualify as knowledge must be internally accessible to the knower: the knower must be able to identify, state, or mentally access the perceptions, memories, or reasons that justify the belief. On a strict internalist reading of the traditional analysis, knowing P requires not just having justification for P but being able to present that justification. Externalist theories, by contrast, hold that at least some of the grounds that make a belief knowledge can be external to the knower — objective features of the relationship between belief and world that the knower need not be aware of.

Goldman’s view is plainly externalist. On his account, the knower does not have to be able to state or even recognize the causal connections that make a belief knowledge. The person who once learned Abraham Lincoln’s birth year from a history book, then forgot the source but retained the belief, can still be said to know Lincoln’s birth year — provided the causal chain from the fact of Lincoln’s birth to the belief remains intact. The inability to rehearse one’s grounds is no obstacle to knowledge on this view.

Goldman makes this point explicitly in contrasting his view with an internalist reading of the traditional analysis: “In at least one popular interpretation of the traditional analysis, a knower must be able to justify or give evidence for any proposition he knows… My analysis makes no such requirement.” And he takes this to be a strength of his view, not a weakness: requiring the ability to state one’s justification would wrongly exclude genuine cases of knowledge — like the forgotten-source case — where the underlying causal structure is perfectly in order.

Externalism and Skepticism

One significant upshot of externalism is that it sidesteps rather than answers the skeptic. The traditional epistemologist, committed to showing that beliefs meet internal standards of justification, must face the skeptic’s challenge: can you justify your belief that you are not dreaming? That an evil demon is not deceiving you? Goldman declines this challenge not because he has answered it, but because his theory does not require that it be answered. Knowledge is a matter of how a belief is objectively situated with respect to the relevant facts, not a matter of what the knower can prove. As Goldman writes: “I have made no attempt to answer skeptical problems. My analysis gives no answer to the skeptic who asks that I start from the content of my own experience, then prove that I know that there is a material world, a past, etc.”

The externalist can simply begin from within our ordinary framework — assuming that there is a physical world, that perception reliably causes true beliefs about it, that memory preserves the causal chains from past facts to present beliefs — and investigate what makes beliefs count as knowledge within that framework. The skeptic’s demand for a proof of the framework itself is not recognized as a legitimate epistemological obligation.

Goldman and Quine: Comparison

Goldman’s externalist causal theory has interesting affinities with Quine’s naturalism. Both views sidestep skeptical arguments. Both are free to appeal to the natural world and causal processes in analyzing knowledge without first proving that such an appeal is legitimate. Both reject the requirement that knowledge be vindicated from some external standpoint immune to skeptical challenge. And both look to objective, science-friendly relations between knower and world — rather than to internally certifiable justifications — as the ground of epistemic success.

But there are differences. Quine restricts his attention to the causal pathway from sensory stimulation to theoretical output, framing epistemology as literally a chapter of psychology. Goldman’s causal connections are more various: they include perception, memory, inference, common causes, and potentially other mechanisms, and they connect beliefs to specific facts rather than to general theoretical outputs. More importantly, Goldman’s view appears to recover some of the normative dimension that Kim accused Quine of abandoning. Even if Goldman’s theory is externalist — even if it does not appeal to internally accessible justifications — it still distinguishes between beliefs that meet the standard for knowledge (those causally connected to relevant facts) and beliefs that do not (those accidentally true, or based on broken causal chains). This is a normative distinction: some beliefs are worthy of the title of knowledge; others are not.

As Goldman himself reflects at the close of his paper: “The question of whether someone knows a certain proposition is, in part, a causal question.” This challenges the traditional epistemological assumption that the relevant questions are purely about logic or justification — a challenge that opens a new way of thinking about what epistemology is, one in which the normative and the naturalistic are not simply opposed. Whether causal connection can fully carry the normative weight that justification was supposed to bear, or whether something important is still lost in the transition from justification to causation, is a question the course leaves as an open and productive line of inquiry.

End of Week 7–9 Notes

The Structure of Knowledge

Two Pictures of How Knowledge is Structured

Why Structure Matters

A recurring theme in epistemology is that knowledge is not simply a collection of isolated facts held by a solitary mind. The course readings up to this point — dealing with inference, skepticism, the relationship between experience and belief — all quietly presuppose something important: that our justified beliefs are bound together by relations of dependence. Whenever we justify one belief by appeal to another, we are acknowledging a structural link. Deductive inferences, inductive inferences, inferences to the best explanation — all of these move justification from one belief to another, and it is this movement that gives knowledge its architecture.

Even the skeptic, who denies that we know anything, is trafficking in structural claims. The Cartesian skeptic insists that the very structures our beliefs depend on for justification are shakier than we assume. This is not a rejection of structure; it is a critique of a particular structure’s reliability. The upshot is that thinking about knowledge forces us to think about it as a structured system, not merely as a set of individually certified truths. Epistemic structure refers to the pattern of justificatory dependencies that holds among our beliefs — the way in which some beliefs support others, and how the whole system of knowledge hangs together.

This insight leads directly to a key principle: knowledge has structure because justification has structure. If knowing something requires being justified in believing it — a claim most epistemologists accept — then the shape of knowledge will mirror the shape of justification. Even reliabilist accounts like Goldman’s causal theory, which treat inference as part of a broader causal chain between beliefs and worldly facts, must acknowledge these webs of dependency.

Linear versus Circular Justificatory Structure

Confining attention to how beliefs depend on other beliefs for their justification, there turn out to be only two coherent possibilities for how that dependence is arranged. The first is linear structure: belief D is supported by belief C, which is supported by belief B, which rests on belief A, and so on in a directed chain or branching tree. Even a complex, branching web of dependencies still counts as “linear” in the relevant sense, because justification always flows in one direction — from supporting beliefs toward the belief being supported — without looping back on itself.

The second possibility is circular structure: belief D is supported by C, C by B, B by A, and A, in turn, by D. The chain of justification folds back on itself. While circular reasoning has traditionally been regarded as a fatal flaw in any argument, circular justificatory structure is the position that coherentists embrace — though, as we shall see, in a far more sophisticated way than a simple diagram conveys.

Foundationalism and the Regress Argument

The linear picture of justification leads naturally to foundationalism — the view that any chain of justification must terminate in a class of basic beliefs that do not depend on any further beliefs for their justification. The vivid image here is a house of cards: most cards rest on others, but the bottom layer rests directly on the table. Basic beliefs are the table-level cards of the epistemic structure.

The driving argument for this conclusion is the celebrated regress argument. If justification is linear and no beliefs are basic, then the justification of any particular belief B would require an infinite series of further justifying beliefs — each belief in the chain depending on the one behind it, without end. Since finite human minds cannot entertain an infinite series of justifying beliefs (let alone deploy such a series in any actual act of reasoning), this scenario is epistemically hopeless. The only escape — short of skepticism — is to posit that some beliefs are basic: they are justified, but their justification does not derive from any further belief.

It is worth noting that the regress argument does not logically compel one to accept basic beliefs. A determined skeptic could accept the regress and conclude that knowledge is simply impossible. But most epistemologists find this unacceptable, and so the argument functions, for them, as a motivation for foundationalism.

What beliefs might qualify as basic? Historically, candidates have included a priori beliefs (logical and mathematical truths), beliefs about our own sense-data or sensory impressions, and perceptual beliefs — beliefs about objects and events in our immediate environment. The problem with relying solely on a priori beliefs is that they tell us nothing about the empirical world. Foundationalism about empirical knowledge therefore requires basic empirical beliefs as well.

Foundationalists also disagree about how secure basic beliefs need to be. Strong foundationalism holds that basic beliefs must possess some special property rendering them especially secure — they must be indubitable, infallible, or incorrigible (immune to correction). Moderate (minimal) foundationalism requires only that a basic belief’s justification not depend on any other beliefs. It need not be certain or immune to error; it merely must stand on its own, epistemically speaking.

Bonjour’s Coherentism: A Preview

Laurence Bonjour, whose The Structure of Empirical Knowledge is the week’s primary reading, accepts that empirical knowledge has structure, but argues that the structure is not foundationalist. He defends instead what is called coherentism (or a coherence theory of empirical justification). On Bonjour’s view, the justificatory structure of empirical belief is not linear but, in some sense, circular. No beliefs are basic. Instead, as he puts it, beliefs are justified by being inferentially related to other beliefs in the overall context of a coherent system. What this means in detail — and why the circular diagram is itself somewhat misleading — will be elaborated in the next lecture.

Coherentism versus Foundationalism

Bonjour’s Critique of Moderate Foundationalism

Bonjour’s first move is to argue that moderate foundationalism is ultimately indefensible. He begins with a fundamental observation about justification itself: the role of epistemic justification is to serve as a means to truth. We require that knowledge be justified precisely because we want our beliefs to be true, and justification is supposed to give us good reasons for thinking that our beliefs are true. This is not a controversial claim — it is built into the very concept of what it means for a belief to be justified.

Bonjour then applies this observation to basic beliefs with considerable argumentative force. If basic beliefs are justified (which foundationalists insist they are), then there must be some characteristic that explains why they are justified. And because justification is a means to truth, that characteristic must also constitute a good reason for thinking the basic belief is likely to be true. This much is simply an application of the general observation about justification to the special case of basic beliefs.

The reductio argument proceeds from these premises. Suppose that some belief B is a basic empirical belief. Since B is justified, there must be a good reason for thinking B is likely to be true. For the believer to be justified in holding B, the believer must have cognitive possession of that reason — they must be aware of why B is likely to be true. The only way to have such cognitive possession, however, is to have further justified beliefs that serve as premises in an argument showing B to be likely true. And since B is an empirical belief (not an a priori one), those supporting premises cannot be entirely a priori — at least one further empirical belief must be among them. But this means B’s justification depends on at least one further empirical belief, which directly contradicts the supposition that B is basic. The concept of a basic empirical belief, Bonjour concludes, is self-contradictory.

In short: trying to explain what makes an empirical belief basic always leads us back to further empirical beliefs. Basic empirical beliefs cannot be justified as basic.

A notable feature of Bonjour’s argument, which will become important shortly, is that it assumes an internalist conception of justification: it requires that the believer actually have cognitive access to the reasons that justify their belief. This internalist assumption is precisely what Alston will later contest.

The Coherence Theory of Justification

Bonjour’s alternative is a coherence theory, which he develops carefully to avoid the obvious objection that circular justification is no justification at all. He grants that a circular diagram — where A is justified by B, B by C, C by D, and D circles back to A — makes the whole structure look viciously circular. The same belief appears both as conclusion and as premise in its own justification.

But Bonjour’s real position is more sophisticated. For him, justification is not linear at all — not even in a circular way. It is holistic or systematic. Justification is generated not by tracing individual chains of inference, but by the global coherence of an entire belief system. The mutual relations of support among all the beliefs in the system together produce justification for each component belief. No single strand of inference does the justifying work; it is the web as a whole that matters.

A useful image — not Bonjour’s own, but illuminating — is a spider’s web. No single thread holds the web together; rather, every thread is supported by, and in turn supports, every other. The web’s integrity is a property of the system, not of any individual strand. Similarly, on Bonjour’s view, the justification of any particular empirical belief is a function of the system’s overall coherence.

This holistic picture explains why Bonjour is not troubled by the circularity objection. The “circle” of justification is not a genuine circle in the logical sense — it is not the case that B is justified by an argument whose premises require B to already be justified. Rather, each belief is justified by its membership in a coherent system, not by any one linear path back to itself. As Bonjour puts it: “the justification of a particular empirical belief depends, not on other particular empirical beliefs as the linear conception of justification would have it, but instead on the overall system and its coherence.”

Bonjour also distinguishes between local and global justification. At the local level — when we assess a specific argument in a specific context — justification does appear linear, as we trace premises to conclusions. But this local linearity is, for Bonjour, something of an illusion. Local reasoning always takes place against the background of a larger, global system of mutually coherent beliefs. Only by zooming out to the global level do we see that justification is actually a property of the whole system, not a chain of linear inferences.

A Defense of Foundationalism — Perceptual Beliefs and Alston

The Special Status of Perceptual Beliefs

Perceptual beliefs occupy a naturally privileged place in any account of empirical knowledge. When I see a cat on the mat and come to believe that there is a cat on the mat, the belief seems to derive from the experience in a peculiarly direct way — not inferred from other beliefs, but simply arising from contact with the perceptual environment. Moreover, perceptual beliefs seem to provide the bedrock for our broader empirical theories: both common-sense beliefs and scientific theories are, at some level, grounded in what we perceive. This is part of the foundationalist’s attraction to treating perceptual beliefs as basic.

Different philosophers have understood perceptual beliefs differently. Some identify them as beliefs about our own sense-data — the immediate contents of sensory experience (as Ayer and the logical empiricists tended to do). Others identify them as beliefs about what we perceive in our environment — the objects and states of affairs we see, hear, and touch. These differing interpretations lead to different versions of foundationalism with correspondingly different candidates for basic beliefs.

Bonjour on Perceptual Beliefs: Cognitive Spontaneity

Even a coherentist like Bonjour must say something about why perceptual beliefs seem special. His account centers on the notion of cognitive spontaneity. When Bonjour comes to believe that there is a red book on his desk while sitting at it, the belief does not arise through inference or deliberate reasoning. It simply appears, as a spontaneous result of his perceptual interaction with the environment. It is this spontaneous character — the fact that the belief is not derived from other beliefs — that marks it as a perceptual belief.

Yet Bonjour insists that cognitive spontaneity in the acquisition of a belief is quite different from independence in the justification of that belief. A belief can arrive without inference while still requiring inferential support in order to be justified. For Bonjour, the perceptual belief that there is a red book on the desk is justified only if the believer also holds a set of supporting beliefs: that the perceptual belief is cognitively spontaneous in the relevant way (call this belief-type K₁), that standard perceptual conditions obtain (conditions C₁ — normal lighting, appropriate distance, and so on), and that spontaneous visual beliefs of type K₁ occurring in conditions C₁ are highly reliable indicators of truth. These three supporting beliefs form the premises of a justifying argument whose conclusion is the perceptual belief itself. Bonjour’s coherentism thus explains the apparent directness of perceptual belief while insisting that justification is always inferential and always systemic.

Alston’s Defense: Will’s Critique and Minimal Foundationalism

William Alston takes the opposite side of the debate, arguing that critics of foundationalism are too hasty. Alston focuses his response on Frederick Will, who argues that there are no cognitions meeting the foundationalist’s specifications. Will’s characterization of foundationalism targets the strong version: basic cognitions, on his reading of foundationalism, must be certain, incorrigible, infallible, and logically independent of all other claims — capable of being established entirely in isolation from any other beliefs.

Alston’s core response is to shift the debate from strong foundationalism to minimal (moderate) foundationalism. Minimal foundationalism does not require that basic beliefs be certain, infallible, or incorrigible. It requires only that whatever justification a basic belief has does not derive from other beliefs. So if Will’s objection is that no beliefs can be certain or infallible — a plausible objection — it simply does not apply to minimal foundationalism, which makes no such demand.

Alston offers an example: the belief that one is feeling depressed. What justifies this belief is not any other belief one holds, but simply being aware of feeling depressed — a kind of direct, non-inferential awareness of one’s mental state. One’s being awake, alert, and in possession of one’s faculties may also be part of what makes the belief justified, but these are conditions one is in, not beliefs one holds and reasons from.

Being Justified versus Establishing Justification

This example leads Alston to a crucial distinction that he develops in the following lecture: the difference between a belief’s being justified and establishing or showing that the belief is justified. This distinction will prove to be the linchpin of his defense of foundationalism.

Evaluating the Debate — Externalism and the Regress

The Alston–Bonjour Contrast

Alston’s key distinction is between two things that are easy to conflate: (1) a belief’s being justified, and (2) showing or establishing that the belief is justified. He grants, generously, that showing a belief to be justified always requires adducing further grounds — some belief q different from the belief p in question. One cannot establish p’s justification without pointing to something beyond p. In this much, the coherentist intuition is correct.

But Alston’s crucial point is that a belief’s being justified does not depend on our ability to show that it is. To say that a belief is immediately justified (basic) is just to say that there are conditions sufficient for its justification that do not involve any other justified beliefs. Even if you have further beliefs that could serve as premises in an argument to show the basic belief is justified, the basic belief’s actual justification does not depend on those further beliefs. The justification is there whether or not the believer marshals it into an explicit argument.

Returning to Bonjour’s perceptual example, Alston can concede that the elaborate argument Bonjour constructs — involving premises about K₁ beliefs, C₁ conditions, and reliability — successfully establishes the justification of the perceptual belief. But Alston denies that this establishing is necessary for the belief to be justified. The belief can be justified by the perceptual situation itself, without the believer needing to possess all those supporting premises.

The root of the disagreement is that Bonjour’s conception of justification is internalist: it requires that the believer have cognitive possession of the reasons that justify their belief. Alston rejects this internalist requirement, at least for basic beliefs.

Foundational Justification and Externalism

If basic beliefs can be justified without the believer being in cognitive possession of justifying premises, then what does justify them? Here externalism offers resources. Consider Alvin Goldman’s causal theory of knowledge: a belief counts as knowledge when it is causally connected to the fact that makes it true, via a reliable causal process — and this causal connection need not be known or accessible to the believer.

Extracting a notion of justification from Goldman’s externalism yields reliabilism: a belief is justified if it is produced by a cognitive process that is reliable (that is, tends to produce true beliefs). Perception is a paradigm example. I see the red book, and visual perception is a reliable belief-forming process. My belief that there is a red book on the desk is justified because it was produced by vision — regardless of whether I can articulate an argument to that effect.

This externalist picture also, interestingly, addresses Bonjour’s demand that there be a good reason for thinking a basic belief is likely true. On a reliabilist account, there is such a reason: the belief was produced by a reliable process. The difference is that on the externalist view, this reason does not need to be possessed by the believer. The reason exists objectively, in the causal structure of the world, even if the believer is unaware of it.

The Deeper Divide: Internalism versus Externalism

Alston’s defense of minimal foundationalism ultimately rests on the claim that the regress argument forces the adoption of some stopping point in the chain of justification, and that externalism provides a principled account of what that stopping point can be. Coherentism also answers the regress, but by dissolving the linear structure of justification into a holistic web. Whether one finds foundationalism or coherentism more compelling may, in the end, turn on a prior commitment regarding internalism and externalism. If you think knowledge always requires that the believer have cognitive access to justifying grounds, coherentism’s picture of mutually supporting beliefs within a system is probably more appealing. If you are willing to allow that justification can accrue to beliefs independently of what the believer can access or articulate, then foundationalism — at least in its minimal form, supported by an externalist account of basic justification — becomes the more plausible alternative.

Socializing Epistemology

The Social Scene of Epistemology

The Abstract Character of Traditional Epistemology

Looking back over the epistemological ground covered in the course — skepticism, the problem of induction, the relationship between experience and theory, the analysis of knowledge, and the structure of knowledge — a common thread becomes visible: the discussion has been strikingly abstract. The paradigm epistemic agent has been a featureless “S” in the schema “S knows that P if and only if…” This S has no gender, no race, no social position, no historical location. Similarly, the Cartesian knower of Descartes’ Meditations is a solitary individual, stripped of social context, meditating in isolation on what can be known with certainty.

Nothing in this tradition encourages us to ask: whose knowledge? From what social position? Under what conditions of inclusion or exclusion? Social epistemology is the branch of epistemology that takes precisely these questions seriously. It studies knowledge in the context of social relationships among knowers, acknowledging that knowledge acquisition is never purely individual — it is socially embedded, collectively produced, and shaped by institutional structures.

Within social epistemology, some of the natural topics include the epistemology of testimony (how knowledge is transmitted from one person to another, which is inherently social), and the epistemology of collective epistemic agents (how communities — scientific communities, juries, regulatory bodies — form justified beliefs, and how individual justification relates to collective justification). Both areas ask not just about what individuals know, but about the social structures that facilitate or hinder the production of justified belief.

Socially Situated Knowers

A further dimension of social epistemology concerns a more radical idea: that there is no such thing as a generic, abstract knower. All knowing takes place from a social location, and that location differentially positions knowers as more or less advantaged in the epistemic enterprise. Some knowers occupy positions of relative privilege in communities where knowledge is produced and validated; others are marginalized, their voices systematically discounted or ignored.

This observation generates two very different responses. The traditionalist response holds that because knowledge is supposed to be objective and socially neutral, social particularities are obstacles to be overcome. The epistemically correct stance is to aspire to a “view from nowhere” — a perspective that transcends particular social locations. A more recent and arguably more fruitful response, however, argues the opposite: the diversity of social perspectives is an asset to epistemic practices, not a liability. Suppressing diversity — insisting on a single dominant perspective — can itself be a source of distortion and error. And crucially, if diverse perspectives have something to teach, then making those perspectives visible and audible becomes itself an epistemic imperative.

Feminist Epistemology: Overview and Motivation

Feminist epistemology has been one of the most powerful engines driving this second response. It is an approach to epistemology that takes seriously the specifics of women’s social location, the history of women’s exclusion from knowledge-producing institutions, and the ways in which that exclusion has impoverished our collective understanding.

The political motivation is straightforward: women have been marginalized in science, philosophy, and other knowledge-producing disciplines for centuries. Aristotle viewed women as biologically inferior. Scientific institutions long excluded women from participation, and even where women were present, their contributions were devalued. Helen Longino, one of the leading figures in feminist epistemology, puts the stakes vividly: the sciences have “stacked the cards against women for centuries,” and since science is the epitome of knowledge and rationality in modern Western society, women need either to find different ways of knowing or to show how feminist perspectives can contribute to and improve existing ways of knowing.

One might object that the problem is simply one of sexism among practitioners, not a problem with epistemology itself. If we could eliminate sexist biases, wouldn’t traditional epistemology do just fine? Longino takes this objection seriously but finds it overstated. Her more compelling motivation for feminist epistemology is internal to philosophical developments: the decline of strong foundationalism (the recognition that no knowledge is beyond doubt), and the rise of naturalism in epistemology (the recognition that psychological and sociological facts about knowers are relevant to how knowledge works). These developments, Longino argues, already create space for feminist perspectives within epistemology. If knowledge is fallible and socially conditioned, then the specifics of women’s social conditioning and women’s perspectives become philosophically relevant, not merely politically motivated.

Suggestions for a Feminist Epistemology

What Feminist Epistemology Is — and Is Not

Longino’s essay In Search of Feminist Epistemology is as much a searching examination of what feminist epistemology cannot be as a positive account of what it can be. Some initial proposals fall short, and understanding why they fall short is crucial to appreciating Longino’s positive contribution.

One initial proposal is that feminist epistemology describes uniquely feminine ways of knowing — special access to truths rooted in women’s biological experience (childbirth, menstruation) or distinctively female modes of learning and cognition. Longino finds value in work that opens a “door on the diversity of ways in which humans apprehend their world,” but insists that cataloguing women’s experiential domains is “not yet epistemology.” Epistemology requires generality: it must say something about the nature of knowledge and the conditions of its acquisition, not merely enumerate particular knowledge-types.

Another proposal draws on standpoint theory. A standpoint, for Longino, is “a perspective afforded by social location” — one’s beliefs about the world develop from, and reflect, the standpoint one occupies. Dorothy Smith applies this idea to sociology, arguing that dominant sociological practice embodies a “ruler standpoint,” projecting the categories of a bureaucratic elite onto the populations it studies, while remaining blind to the power relations involved in that projection. A woman sociologist, by contrast, encounters power relations both as subject and object, putting her in a position to illuminate what the male sociologist cannot see.

Longino acknowledges the value of this analysis but finds it still insufficient for a full feminist epistemology. Smith’s insight is specific to sociology — to a particular discipline’s construal of the social world — not to knowledge in general. And more fundamentally, Longino argues that there is not enough uniformity in women’s experiences to ground a new general epistemology. Women’s lives are too various; “women” is not a homogeneous epistemic category from which stable general conclusions about knowledge can be extracted.

Doing Science as a Feminist: Underdetermination and Theory Choice

Longino’s positive proposal shifts the frame entirely. Rather than asking what uniquely feminist ways of knowing exist, she asks how feminist values can shape participation in existing knowledge-seeking practices. Feminist epistemology is not a replacement for mainstream epistemology; it is mainstream epistemology done as a feminist.

To illustrate this, Longino focuses on the problem of underdetermination in science. Underdetermination names the gap between observational evidence and theory choice: the evidence available at any time does not uniquely determine which theory should be accepted. Multiple competing theories can be compatible with the same body of observations. This is not a peculiarity of difficult cases — it is a pervasive feature of scientific reasoning. The Copernican revolution illustrates it nicely: in the sixteenth and early seventeenth centuries, both the heliocentric hypothesis and various geocentric theories were compatible with the available astronomical observations. Something beyond observation — what philosophers of science call background assumptions — was needed to settle the debate.

Background assumptions can include quite varied things: beliefs drawn from adjacent scientific domains, methodological principles like Ockham’s Razor (prefer simpler theories, all else equal), or even broader cultural beliefs (such as the assumption that, because humans are central to God’s creation, Earth must be central to the universe — a background assumption that kept geocentrism alive long after it should have been abandoned).

Philosophers of science, most notably Thomas Kuhn, have articulated lists of the factors that influence theory choice in underdetermination contexts: accuracy (empirical adequacy), consistency, breadth of scope, simplicity, and fruitfulness. Some of these — accuracy, consistency — are tied to evidence. Others — scope, fruitfulness, simplicity — are genuinely non-evidential: they reflect values about what makes a theory good, values that go beyond any immediate evidential comparison.

Feminist Standards of Theory Choice

Longino’s insight is that if non-evidential standards inevitably shape theory choice, then feminist values can legitimately enter the picture as alternative or supplementary standards. She identifies several such feminist-oriented standards:

Empirical adequacy is not strictly a non-evidential standard, but it can be applied with a feminist slant — for instance, by demanding that theories be empirically adequate with respect to women’s experience, not just the experience of dominant groups. This can expose unsupported biases in existing research, such as causal biological accounts of sex differences that are poorly evidenced.

Ontological heterogeneity is a preference, among otherwise comparable theories, for those that acknowledge diversity and individual difference over those that assume homogeneity. This standard has a feminist motivation because it blocks the tendency to treat one group’s behavior (characteristically male behavior, in many research contexts) as normative, dismissing deviations as deficiencies. As Longino puts it, “Difference is a resource, not a failure.”

Applicability to current human needs — particularly the human and social needs traditionally served by women — is another potential feminist standard. This is not merely a call for applied science, but for science oriented toward needs that dominant research programs have systematically neglected.

What makes these standards feminist? Longino’s answer is careful. The standards are not intrinsically feminist — a non-feminist scientist could endorse ontological heterogeneity for reasons having nothing to do with gender. But feminists have a specific reason to endorse them: they tend to make gender and women visible, preventing them from being washed out by assumptions of homogeneity or by research frameworks designed around male experience. This is what Longino calls the bottom-line requirement for feminist inquiry: standards should “reveal or prevent the disappearing of the experience and activities of women and/or reveal or prevent the disappearing of gender.”

Finally, the mere identification of such standards is not itself epistemology. Genuine epistemological work consists in reflecting critically on the standards: examining their interrelations, evaluating their utility, asking whether they serve feminist goals and whether alternatives might serve those goals better. This reflective activity, conducted from a feminist orientation, is what constitutes feminist epistemology in Longino’s sense.

Broadening the Social Perspective — Mills on Alternative Epistemologies

From Feminist Epistemology to Alternative Epistemologies

While Longino carves out a place for feminist values within the existing practice of epistemology, Charles W. Mills advances a more philosophically radical claim. His essay “Alternative Epistemologies” argues for the legitimacy of epistemological frameworks associated with socially subordinated groups — not just women, but also racial minorities and the working class. Mills’s proposal is more ambitious: rather than finding a way for feminists to intervene in existing epistemology, he argues that subordinated groups actually possess epistemic advantages over dominant groups, and that acknowledging these advantages requires something genuinely new.

An alternative epistemology, in Mills’s sense, is an epistemology associated with a socially subordinated group and grounded in the differential experience of that group. The idea has Marxist roots: Marx argued that the proletariat, living under the constraints of capitalism, could see through the illusions that capitalism generates — illusions that the ruling class itself is blind to or actively invested in perpetuating. Mills extends this structure to racial minorities (who can perceive the pervasive reality of racism that white dominant groups deny or minimize) and to women (who can see through the structures of patriarchy in ways that men, typically, cannot).

The Basis of Epistemic Advantage: Social Causation and Differential Experience

What accounts for this epistemic advantage? Mills considers and rejects biological accounts: the claim that biological differences between groups — such as the distinctive physical experiences of pregnancy and nursing available to women — produce distinctive cognitive capacities. The science is inconclusive, and in any case, arguing from biology would require appeal to the very mainstream epistemological standards that alternative epistemologies are supposed to transcend.

Mills’s preferred account is social causation. The social positioning of a group — not its biology — is what determines its epistemic resources. Drawing on Marx, Mills observes that social causation has both positive and negative epistemic effects. Under capitalism, all classes are subject to certain shared illusions — most notably, the illusion that capitalists and workers freely and voluntarily engage in economic exchange as equals. This liberal ideology mystifies the actual relations of production. The capitalist class is especially invested in this illusion; it naturalizes their dominance. But the working class is not automatically immune either — people can and do internalize the very systems that oppress them.

The key is differential experience. Workers experience not only the ideological surface of capitalism (the marketplace, the formal freedom of contract) but also the brutal material reality of their position: the coercive dependency on wages, the absence of real alternatives, the disillusioning conditions of factory work. This experience “dramatically undercuts” the voluntarist illusion. It gives workers cognitive resources to see through what the capitalist class takes as natural law. Similarly, women’s experience of harassment, the ever-present threat of sexual violence, and the structural constraints of patriarchy puts them in a position to perceive dynamics that men who have never confronted such constraints are systematically likely to miss. Black minorities, whose experiences of housing discrimination, police violence, and employment prejudice are invisible in official narratives, can “come to realize how pervasive, despite official denials, white racism continues to be.”

Mills distinguishes differential experience from two other proposed bases of epistemic advantage. Simple oppression is not enough: suffering is not necessarily cognitively illuminating, and oppressed people can be as captivated by the illusions of a system as anyone. Nor is it sufficient to argue, as Marx sometimes did, that the working class standpoint is somehow universal — this claim seems arbitrary and has the troubling implication that the standpoints of women and racial minorities are merely secondary. Only differential experience, understood as a structural feature of one’s social position that shapes one’s categories of perception and interpretation, satisfies Mills’s requirements.

The Relationship between Alternative and Mainstream Epistemology

Mills is aware that alternative epistemologies can seem like a frontal assault on mainstream epistemology — the tradition centered on “that familiar Cartesian figure, the abstract disembodied individual knower beset by skeptical and solipsistic hazards.” Some proponents of alternative epistemologies do present them as transcending or replacing mainstream epistemology, diagnosing its classic problems as arising precisely from “the privileged universalization of the experience and outlook of a very limited (particularistic) sector of humanity — largely white, male, and propertied.”

But Mills’s own view is more nuanced. He sees alternative epistemologies as complementary to, rather than hostile toward, mainstream epistemology. In virtue of a shared humanity, there is a “universal zone” of common experience that makes traditional epistemological inquiry meaningful: the abstract problems of perception, memory, inference, and the Gettier puzzle do not disappear. Mainstream epistemology addresses what is common to human epistemic life. Alternative epistemologies take up what falls outside the trajectory of hegemonic groups — areas of experience, and hence areas of knowledge, that dominant groups systematically fail to encounter and therefore fail to theorize.

The New and the Old in Epistemology

Objections: Relativism and the Problem of Internal Diversity

Alternative epistemologies face two serious challenges. The first is the diversity objection, raised in a form we already encountered in Longino: there is not enough uniformity of experience within any subordinated group to ground a coherent standpoint. Women’s experiences are immensely varied; so are the experiences of black people, or workers. Moreover, individuals belong to multiple groups simultaneously: a white working-class woman is oppressed in some dimensions while privileged in others, which threatens to dissolve any stable group standpoint into an individualistic morass.

The second challenge is relativism: if all standpoints are shaped by social positioning, why should any particular standpoint — even a subordinated one — be considered more accurate than any other? What prevents the claim that “subordinated groups see more clearly” from being just another perspective, with no claim to privileged accuracy?

Mills’s response to both challenges draws partly on naturalized epistemology. Empirical psychology can investigate the mechanisms that sustain cognitive distortions — including those underlying racism, sexism, and class ideology. Such investigations make the claim that dominant groups are susceptible to systematic cognitive distortions more than a mere assertion. Meanwhile, subordinated groups, whose differential experience brings them into confrontation with those distortions, may develop schemas and interpretive frameworks better equipped to detect them. What matters for establishing a standpoint is not complete uniformity of experience, but a degree of commonality in the experience of oppression relative to a particular hegemonic group — enough to generate shared interpretive resources that differentiate the subordinated group’s epistemic position from that of the dominant group.

On relativism, Mills argues that the progressive articulation of different standpoints should be seen not as a descent into relativism, but as a cumulative correction toward greater accuracy about social reality. The theoretical articulation of women’s standpoint corrected the limitations of a working-class standpoint that ignored gender; the standpoint of black women corrected the limitations of white feminist analysis. Each new articulation introduces resources for identifying what the previous articulation missed. The overall trajectory is toward a more complete and accurate picture of social dynamics. As Mills puts it, what is needed is “a synthesis of the alternative epistemologies that recognizes both the multiplicity and the unity, the experiential subjectivity and the causal objectivity of hierarchical class-, gender-, and race-divided society.”

Socializing Metaphysics

Can Reality Be Socialized?

From Social Epistemology to Social Metaphysics

The final module of the course moves from social epistemology back to metaphysics — specifically, to the question of whether metaphysics itself, and in particular the concept of reality, can be understood as socially constructed. Sally Haslanger’s essay Ontology and Social Construction addresses this question with considerable philosophical care. Ontology, the branch of metaphysics concerned with what exists and what kinds of things are real, here becomes the site of a confrontation between social constructionist philosophy and a defensible notion of independent reality.

The context for this confrontation is familiar from Week 11: Mills’s alternative epistemologies are explicitly epistemologies of social reality, which implicitly presupposes a distinction between social realities and other kinds. But what if that distinction collapses? What if all reality, in any sense we can coherently talk about it, is in some measure socially constructed? And if so, does that threaten the notion of an independent reality — a reality that exists independently of our thought, practices, and language?

Haslanger’s article is, in part, motivated by the observation that certain strands of socially engaged philosophy tend toward strong constructionist conclusions that she finds philosophically untenable. Her aim is not to dismiss social construction — she considers it a genuine and important philosophical phenomenon — but to show how much we can acknowledge it while still defending a moderate realism about independent reality.

Types of Social Construction

Haslanger distinguishes several importantly different kinds of social construction, and the distinctions are crucial for her argument.

Causal social construction: Something is causally socially constructed when social factors play a role in creating it or determining its characteristics. The traits we associate with masculinity and femininity, for instance, are arguably products of socialization rather than biological necessity — they are causally brought about by social processes.

Constitutive social construction: Something is constitutively socially constructed when its very definition makes reference to social factors. The concept of gender may be constitutively constructed in this sense: no adequate definition of gender can avoid reference to the broad network of social relations that structures the distinction between masculine and feminine.

Pragmatic construction: Classifications and distinctions are pragmatically constructed when their use is determined, at least in part, by social factors. Haslanger discriminates between a weak and a strong version of this notion. The weak version holds that social factors only partly determine the distinction’s use. The strong version holds that social factors wholly determine its use, and the distinction fails to reflect any independent “fact of the matter.”

Nearly all of our classificatory schemes are weakly pragmatically constructed, for the simple reason that language — the medium through which all categories are articulated — is itself a social phenomenon. The categories of “cat” and “dog” may well be weak pragmatic constructions in this sense. But this does not entail that there are no real differences in the world between cats and dogs — the weak pragmatic construction of the distinction is compatible with there being genuine, mind-independent differences among the animals to which the categories apply.

Strong pragmatic construction is another matter entirely. Here, the category reflects no independent fact about what it is applied to; it is a projection of the interests and perspectives of some social group, presented under the illusion that it picks out an intrinsic characteristic of its objects.

Examples of Strong Pragmatic Construction

Haslanger offers several illuminating examples, drawing substantially on the work of legal theorist Catharine MacKinnon.

Consider the category of “cool,” applied to people. There is no objective property that cool people possess — no intrinsic characteristic that “cool” picks out. What the label actually does is project the preferences and standards of an in-group, while creating the illusion that it is identifying something real in the people to whom it applies. The concept is strongly pragmatically constructed: it is wholly determined by social factors, and there is no independent fact of the matter answering to it.

A more consequential example is ideal womanhood — the culturally operative conception of women’s nature. On MacKinnon’s analysis, this ideal is an externalization of male desire, projected onto women and presented as intrinsic to them. “Women’s nature” is not an independently existing property that the concept tracks; it is a social construction that serves the interests of a particular social group while creating the illusion of describing something objective.

The most striking example is rape as legally defined. Standard legal definitions center on the forced penetration of the vagina by the penis — a definition that, on MacKinnon’s analysis, reflects a conception of women’s sexual integrity built around male, not women’s, understanding of sexual violation. Women’s actual experiences of sexual violation extend far beyond this narrow legal category. The legal concept of rape thus constitutively constructs women’s sexual integrity in terms of male interests, projecting onto women a conception of integrity that does not answer to women’s own experience. MacKinnon and Haslanger do not conclude that rape is “unreal” — rather, the reality of rape (legally defined) is a social reality, not an independent one. “One cannot violate something that does not exist”: if the conception of women’s sexual integrity embedded in the legal definition is itself a social construct answering to male interests, then legal rape, understood on those terms, is real only as a social construct, not as a violation of some independently existing integrity.

An important clarification: strongly pragmatically constructed categories are typically also cases of constitutive construction (their definitions make reference to social factors) and can produce further causal social constructions. Applications of the cool/uncool distinction cause people to change their behavior to conform to in-group standards — “cool dudes” are, in this sense, causally socially constructed, in a process Haslanger calls discursive construction.

The central question that emerges from all of this is: is the category of reality itself a strong pragmatic construction? Does acknowledging the pervasiveness of social construction force us to conclude that there is no independent reality at all — that “reality” is just another illusion projected from a particular social position?

Reality, Construction, and Course Conclusion

Two Arguments for Radical Constructionism

Haslanger considers two arguments for the conclusion that the notion of independent reality is itself a social construction. She finds both unpersuasive, but her responses illuminate what can and cannot legitimately be claimed on behalf of social construction.

The indirect argument attempts to generalize from familiar cases of strong pragmatic construction. We have seen that categories like “cool,” “women’s nature,” and “rape” (legally defined) are strongly pragmatically constructed. Can we not simply generalize: since all our categories are, in some measure, shaped by social perspective, anything we say about “reality” is actually only a description of how things seem to us from our social location? As Haslanger describes MacKinnon’s line of thought: “because what we think or know about things is always conditioned by our particular social position, all we can ever meaningfully say about them is how they are related to that position… we’re actually only speaking of how things seem to us — that is, social facts.”

Haslanger’s response is to draw a distinction between the criteria we use for applying a category and the content of the category. The criteria by which we judge something to be water — its look, taste, and feel — do reflect our perspective, our sensory apparatus, and our interests. But the content of the category “water” is another matter: water is H₂O, a substance with a chemical constitution that has nothing to do with any human perspective and would characterize water even if no humans existed to taste it. Similarly, social forces may be responsible for my turning my attention to the street outside my window and noticing it is wet — perhaps the sound of rain drew me there, a social interaction prompted me to look — but the content of my belief that the street is wet (the pavement’s actual physical properties of moisture and reflectance) is independent of those social motivations. Social factors govern why and how we form beliefs; they do not thereby determine the truth-conditions of those beliefs.

The direct argument — what Haslanger calls the anti-objectivist argument — proceeds differently. It begins by defining reality in terms of objective knowledge: an object or fact is real just in case it is or can be objectively known. Then it argues that the notion of “objective knowledge” — knowledge purified of all social influence — is illusory. Human inquiry is always conditioned by social interests; no perfectly neutral vantage point exists. Therefore, the notion of “objective knowledge” collapses, and with it, any notion of reality defined in terms of such knowledge.

Haslanger grants the first step: the idea of objective knowledge as completely insulated from social factors is indeed bankrupt. But she denies the conclusion, because she denies the premise that reality should be defined in terms of objective knowledge in the first place. Why should we say that a thing is real only if it is or can be objectively known? Realists typically reject precisely this connection. The simplest and most defensible conception of reality is just: to be real is to exist. And there is no good reason to think that existence depends on any relationship to our knowledge — social, objective, or otherwise. If we sever the tie between “being real” and “being known,” then showing that knowledge is socially conditioned says nothing about whether reality is independently constituted. As Haslanger puts it: “When I say that something is real, my assertion is true just in case the thing in question exists; this is so even if the criteria I employ in making the judgement are socially loaded.”

Haslanger’s Moderate Realism

Haslanger’s conclusion is a form of moderate realism. Social construction is genuine and pervasive: nearly everything we think and say about the world is shaped by social conditioning and social interests, and some of our most consequential categories — “women’s nature,” legal rape — are strongly pragmatically constructed projections with no independent reality behind them. These are serious and important philosophical observations with real political stakes.

But none of this establishes that independent reality itself must be surrendered. The distinction between epistemology and ontology — between what we can know and what exists — survives the challenge from social constructionism. Social forces condition the former; they do not thereby constitute the latter. We need not choose between acknowledging the depth of social construction and retaining a defensible notion of a mind-independent world.

Course Conclusion: Metaphysics and Epistemology Together

Because this is the final lecture of the course, Callaghan uses Haslanger’s argument as a vantage point from which to reflect on the course as a whole. Haslanger’s defense of the epistemology-ontology distinction is, in a sense, a defense of the very division of labor that organized this course: the first part dealt with metaphysical questions (the nature of causation, the structure of reality, the problem of induction in its metaphysical dimensions), and the second part dealt with epistemological questions (the analysis of knowledge, justification, the structure of belief, and now the social dimensions of knowledge).

Social constructionism might suggest that this division is unstable — that one cannot cleanly separate questions about what exists from questions about how we know what exists, because “what exists” is always already shaped by how we know. Haslanger, as we have seen, resists this conclusion. But even granting her resistance, the course has illustrated how deeply these two areas of philosophy implicate each other.

Epistemology needs metaphysics: without a defensible account of the realities we hope to know, we have no framework for asking how knowledge of them is possible. We cannot ask how we know about causation unless we have some account of what causation is. We cannot ask how scientific theories are justified unless we have some understanding of what the theories are about.

Metaphysics needs epistemology: without careful reflection on the conditions and limits of knowledge, metaphysical claims float free, unmoored from any basis in what can actually be known. The history of philosophy is littered with metaphysical systems that collapsed when epistemological scrutiny revealed their foundations to be unsupportable.

The course has traced this interdependence through an extended philosophical journey: from the empiricist and rationalist accounts of the origins of knowledge, through Hume’s skepticism about induction and causation, through the analysis-of-knowledge debates and the Gettier problem, through foundationalism and coherentism as competing accounts of epistemic structure, and finally to feminist epistemology, alternative epistemologies, and social constructionism. Throughout, the questions of what we can know and what there is to know have been inseparable companions. That inseparability is, perhaps, the deepest lesson of the course.