Insight Editions · No. 294
Think
A Compelling Introduction to Philosophy
Simon Blackburn
Blackburn presents philosophy as conceptual engineering: the study of how the concepts we live in are built, how they fail, and how they might be improved.
A study guide to Think by Simon Blackburn. Shared free for personal reading; not for sale and not endorsed by the authors or publishers. Please read and buy the original.
The Book on One Page
Think (Oxford University Press, 1999) is Simon Blackburn's introduction to philosophy as conceptual engineering: the study of the structure of thought, worked through nine classic problems from Descartes's doubt to what to do. Each chapter takes a famous argument, reconstructs it, finds the hidden flaw, and asks what picture of ourselves produced it. The book's recurring discovery is that most philosophical traps are set by one picture, a ghostly inner self and a God's-eye reason standing outside nature, and that we think better once we see ourselves as natural creatures whose concepts are tools for organising experience and living together.
Philosophy is conceptual engineering. Cartesian pictures (a Real Me behind the brain, a soul substance, a perfect being provable from its concept, a normative order perceived by reason) generate puzzles they cannot solve; naturalism, Humean mitigated scepticism and Kantian organising principles do better, though the deepest problems stay open.
Nine problem areas: the point of philosophy, knowledge, mind, free will, the self, God, reasoning, the external world, ethics. The method is close reading of Descartes, Locke, Berkeley, Hume and Kant, with Wittgenstein as a constant interlocutor, plus thought experiments and a little logic and probability.
Descartes's Meditations (1641); Hume's Dialogues (1779) and "Of Miracles"; Kant's Copernican revolution (1781); Bayes's theorem and the 1 in 11 base-rate case; the Golden Harp lottery with its 6^9 tickets.
Anyone who wants to read the classics "with a mind prepared", and practitioners (architects, data scientists, risk managers) who need sharper tools for evidence, inference, identity, accountability and value in systems that now include AI.
Why bother. Philosophy studies the mental housing we live in. Its questions arise from self-reflection and cannot be settled by measurement, so doing it well is a skill. It pays at three levels: understanding is good in itself; reflection is continuous with practice (beliefs about selfishness, fate or dualism change how people live and do science); and ideas are what people kill each other for, so the sleep of reason produces monsters.
Knowledge. Descartes's Evil Demon strips away everything but the Cogito; his rebuilding via clear and distinct ideas and a non-deceiving God fails through a weak trademark argument and the Cartesian circle. Hume shows that using reason to certify reason is incurable, leaving natural foundationalism, Neurath's coherentism or scepticism, which "permanently beckons" because knowledge implies a sense of our own reliability.
Mind, freedom and self. Dualism leaves Zombies and Mutants wide open, which is reason to reject it; the real choice is Locke's brute annexation of mind to matter versus Leibniz's demand that the mental be intelligible from the physical, and colour science shows how far the latter can go. Free will needs no ghost: the dilemma of determinism or chance is met by inside control, flexible decision-making responsive to true and available considerations. The self is not an object Hume could find but, as for Kant, the structural point of view that organises experience; Locke's memory criterion, Reid's brave officer and scrambling cases show that the crisp "me or not me" of our hopes has no metaphysical match.
God. The ontological argument confuses real with imagined greatness; the cosmological ends in regress or an arbitrary stop that might as well be the world; Hume's Philo demolishes design, and from a mixed world infers at best a mixed god, so the problem of evil stands. Hume on miracles, in Bayesian form, shows testimony can almost never beat a miracle's improbability; Pascal's wager rests on false options; and fideism must be judged ethically, by its fruits.
Reasoning and the world. Logic asks only whether premises could be true and the conclusion false; Frege's quantifiers expose scope fallacies. Hume's problem of induction, dramatised by Straightie and Kinkie, survives Bayes, which nonetheless guards against the base-rate fallacy. Science cannot see why nature must behave as it does, but it corrects itself. In the world, the argument from perceptual relativity drives colours into the mind, Berkeley and Hume turn it on solidity, Faraday dissolves things into powers, and the search for a Straightjacket behind laws repeats the cosmological argument. Kant recasts space, time and causation as organising principles, but no one has found a stable place between Philonous and Moore.
What to do. Practical reason rests on concerns, not mere desires, and psychological egoism is "entirely wrong". Ethics is the set of concerns we expect of each other, internalised as a voice within. Blackburn reads value-talk as expressing concerns (non-cognitivism): there is no mysterious normative order, but concerns are better or worse by the lives they produce. Argument should respect the other's point of view; values should be coherent, objective and imaginative; and relativism's "just us" fails where practices trespass on concern and respect. We sail Neurath's boat, and nobody else is in dry dock either.
| Ch | Title | One-line payload |
|---|---|---|
| 0 | Introduction | Philosophy is conceptual engineering of the mental housing we live in, worth doing on high, middle and low grounds. |
| 1 | Knowledge | Descartes's Demon leaves only the Cogito; his rebuilding is circular, and Hume shows reason cannot certify reason, so scepticism beckons. |
| 2 | Mind | Dualism opens Zombies and Mutants; Leibniz's demand for intelligibility beats Locke's brute annexation, as colour science shows. |
| 3 | Free Will | Neither determinism nor chance gives control; inside control responsive to true, available considerations does, without any ghost. |
| 4 | The Self | Hume finds no self; Kant shows it is a structural point of view; Locke, Reid and scrambling show crisp survival is an illusion. |
| 5 | God | Ontological, cosmological and design arguments fail; evil stands; Hume on miracles wins in Bayesian form; Pascal's wager rests on false options. |
| 6 | Reasoning | Validity, truth-tables, quantifiers; Hume's induction problem (the Golden Harp) survives Bayes, which still exposes the base-rate fallacy. |
| 7 | The World | Relativity drives secondary qualities into the mind; Berkeley and Hume threaten matter itself; Kant's organising principles promise a middle way. |
| 8 | What to Do | Ethics rests on concerns we expect of each other; non-cognitivism without a normative order; respect, coherence, and no "just us". |
Introduction
Blackburn recasts philosophy as conceptual engineering: the study of the structure of thought, the way an engineer studies the structure of material things. Its questions arise from self-reflection, they resist the agreed procedures that settle empirical questions, and doing them well is a skill, a knowing how, more than a body of facts. He then defends the point of it on three grounds (high, middle, low) and declares the book firmly on the side of reason and the critical tradition.
Philosophy as Conceptual Engineering
The book is for anyone who wants to think about the big themes: knowledge, reason, truth, mind, freedom, destiny, identity, God, goodness, justice. These are not a specialist preserve; they structure how everyone thinks about the world and their place in it. Blackburn admits the word "philosophy" suggests the impractical and weird, and says he would rather introduce himself as doing conceptual engineering. An engineer studies how the parts of a material structure function and interconnect, and what would happen, for better or worse, if they were changed. The philosopher does the same for the structure of thought.
Our concepts form the mental housing in which we live, and before we can decide whether to be proud of it or to tear it down, we have to know what it is.
The book is self-standing, but it is built to be read alongside the classics it quotes most: Descartes's Meditations, Berkeley's Three Dialogues, Hume's Enquiry Concerning Human Understanding and his Dialogues Concerning Natural Religion. A reader who finishes it should be able to open those texts, and logic or ethics books, "with a mind prepared".
What Are We to Think About?
Blackburn sorts the big questions into three families. They look ordinary, but every one of them stalls the moment you ask how you would actually settle it.
| Family | Sample questions |
|---|---|
| About ourselves | What am I? What is consciousness? Could I survive bodily death? Are others' sensations like mine? If I can't share their experience, can I communicate with them? Do we always act from self-interest? Am I a programmed puppet? |
| About the world | Why is there something rather than nothing? What separates past from future? Why does causation run past to future, and could the future affect the past? Why is nature regular? Does the world presuppose a Creator, and why would he (she, they) create it? |
| About ourselves and the world | Is the world as we take it to be? What is knowledge and how much have we? What makes inquiry a science (psychoanalysis? economics?)? How do we know about numbers, values and duties? Are our opinions objective or subjective? |
Why These Questions Are Strange
The contrast case is the time of high tide. That is an empirical question, a matter of experience: there are authoritative tide tables, I roughly know how they are made, and failing all else I can go and measure the sea. Agreed procedures of looking, measuring and applying tested rules settle it. The philosophical questions are different: we don't know where to look, and we may not even be sure what we mean or what would count as an answer.
The puppet headline. Imagine the newspaper headline: "Neuroscientists discover human beings not puppets." Try to picture the experiment behind it. What reading on which scanner would show it? The difficulty is not a shortage of data. The neuroscientists would first need to know what being a puppet, or not being one, consists in, and that is a conceptual question the scanner cannot answer. Blackburn's point: some questions are hard because we lack the concepts to say what evidence would look like, and sorting out those concepts is the philosopher's job.
Where They Come From: Self-Reflection
The source of the bafflement is self-reflection. Humans are relentlessly able to step back: we act from habit, then reflect on the habit; we think things, then reflect on what we think; we ask whether we know what we are talking about, whether what we say is "objectively" true or just our own "take". At that point we are examining concepts and procedures we normally just use, "looking at the scaffolding of our thought".
This can happen in the middle of normal work. A historian will at some point ask what "objectivity", "evidence" or "truth" means in history. A cosmologist pauses over equations full of the letter t to ask what is meant by the flow, direction or beginning of time. At that moment they become philosophers, whether they recognise it or not, and the activity can be done well or badly.
Philosophy as a Skill
Learning philosophy is better described as acquiring thinking skills. Doing it well is "more like playing the piano well": a knowing how as much as a knowing that. The model is Socrates of Plato's dialogues, who prided himself not on knowledge but on being the only one who knew how little he knew, and whose talent (estimates of his success differ) was exposing weak claims to know.
- Avoid confusion. Keep distinct things distinct.
- Detect ambiguities. Notice when a word or claim can be read two ways.
- Hold one thing at a time. Keep ideas separate in mind while working on them.
- Make reliable arguments. Move from premises to conclusions that actually follow.
- Become aware of alternatives. See the other positions the evidence leaves open.
Blackburn's summary image: concepts are the lenses through which we see the world, and in philosophy the lens itself is the topic. Success is measured not by what you know at the end but by what you can do when "the seas of argument rise" and confusion breaks out.
What Is the Point?
Reflection bakes no bread and flies no aeroplanes, so why bother? Blackburn gives three answers: high ground, middle ground and low ground.
| Answer | Claim | Blackburn's assessment |
|---|---|---|
| High ground | Questions the question: understanding ourselves is good for its own sake, like pure mathematics, music or reading Jane Austen; it is mental health, and brings the pleasure of mental exercise | Not wrong, but it only persuades people "half-convinced already" |
| Middle ground | Reflection is continuous with practice: how you think about what you do changes how you do it, or whether you do it | The main argument; supported by four examples |
| Low ground | Down "in the basement": ideas are what people kill each other for; critical reflection is the antidote to the sleep of reason | Explains why reflection is feared and resisted |
The High Ground
The high-ground reply is the classic philosophical move of going up one level: what do we even mean by "the point"? Architecture, music, art, history and literature bake no bread either. Much of life is a loop of raising more hogs to buy more land to raise more hogs; the time we take out of it is time to be cherished, in which we "cosset our mental health". The trouble is only that it appeals to people who did not ask the question "in a very aggressive tone of voice".
The Middle Ground: Four Examples
The middle-ground claim is that reflection is continuous with practice: it can direct research, attitudes to people who live differently, or a whole life. Blackburn's examples:
- Life after death. Belief in an afterlife may make someone willing to face persecution they would not face if convinced, as many philosophers are, that the notion makes no sense.
- Fatalism. The belief that the future is fixed whatever we do is purely philosophical, yet it can paralyse action; politically it can express acquiescence in low social status, a convenient pay-off for those of higher status who encourage it.
- Homo economicus. The view that we are all at bottom selfish shapes how people treat each other.
- Copernicus, Einstein and mind-body dualism. Two cases where better reflection advanced science, and one where bad reflection distorts medicine.
The cost of believing everyone is selfish. The leading paradigm of the social sciences, homo economicus, pictures each person looking after himself in competitive struggle. People who come to believe this become less trusting, less cooperative, more suspicious. Cooperative ventures get harder or impossible, and they risk Thomas Hobbes's (1588 to 1679) "war of all against all". In markets they pay heavy transaction costs: if "a verbal contract is not worth the paper it is written on", I need lawyers to draft penalty clauses, and if I trust the lawyers only to pocket their fees, more lawyers to check the first lawyers. All this may rest on a philosophical mistake, viewing motivation through the wrong categories. If people can care for others, or at least for doing their bit and keeping promises, a more optimistic self-image may be one they can come to live up to.
Two trading towns. Picture two towns with identical resources. In one, everyone is taught from school that all apparent kindness is disguised self-interest; in the other, that most people keep their word unless given strong reason not to. Nothing about the inhabitants' brains differs. Yet the first town needs notaries, escrow, insurance on every deal and auditors for the auditors, while the second runs on handshakes and spends the difference on schools. The thought experiment isolates Blackburn's claim: a theory of human nature can be partly self-fulfilling, so getting the categories right is a practical matter with a price tag.
The scientific examples share one structure: reflection on how we know (epistemology) produced a leap in physics. Epistemology comes from the Greek episteme, knowledge.
- Copernicus (1473 to 1543). He saw that perception of motion is perspectival: whether things look as if they are moving depends on whether we are moving. We know the effect from trains and airports, when the neighbouring train seems to pull away and then, with a jolt, we realise it is we who are moving. So the apparent motions of stars and planets might be due to the motion of the observers, as turned out to be the case: "the first spectacular leap of modern science".
- Einstein. His reflection on how we know whether two events are simultaneous had the same shape: measurements depend on how we are travelling relative to the events we clock. This led to the Special Theory of Relativity, and Einstein himself acknowledged earlier philosophers for sensitising him to these epistemological complexities.
Dualism in the clinic. If mind and body are strictly separate things, it becomes hard to see how they interact. Doctors may then conclude from the armchair that treatments addressing mental causes of physical conditions must fail, since good science says physical effects need physical and chemical causes. Drugs and electric shocks look "right"; humane treatment, counselling and analysis look unscientific. Blackburn's reply: that certainty rests not on science but on a false philosophy. Mind-body interaction is commonplace: thinking of something makes people blush; thinking of danger makes hearts pound, fists clench, guts constrict. So there is nothing incomprehensible in cheerful optimism affecting skin conditions or even cancer remission. Whether it does is a purely empirical question; the armchair certainty that it cannot is itself unscientific.
The cancer example is a principle, not a finding. Blackburn's logical point stands: whether mental states affect bodily disease is an empirical question, not something to be ruled out from the armchair. But readers should not take the example as evidence. Large prospective studies have generally not found that emotional state or optimism predicts cancer survival, and the "positive thinking cures cancer" idea has been criticised for blaming patients who do not recover. The mind-body link is well established for stress responses and placebo effects on symptoms such as pain; it is weak to absent for tumour remission.
The middle-ground conclusion: a system of thought is something we live in, just as much as a house, and if our intellectual house is cramped, we need to know what better structures are possible.
The Low Ground: The Sleep of Reason
The low-ground answer takes the point "down in the basement where human life is a little less polite". The Spanish painter Goya etched a satire titled "The Sleep of Reason Produces Monsters", believing many human follies come from reason's sleep. There are always people telling us what we want, how they will provide it and what we should believe, and convictions are infectious. We readily believe our religion, politics or God-given rights trump theirs, or that our interests require pre-emptive strikes against them.
In the end, it is ideas for which people kill each other. Wars, oppression carried out with a good conscience, and even acquiescence in one's own oppression all run on beliefs about who the others are and what our rights require. Critical awakening is the antidote: stepping back to see our perspective as possibly distorted, and asking whether there is real argument for preferring our ways or merely subjective preference.
Why Reflection Is Resisted
Because nobody can tell in advance where reflection will lead, it can look dangerous, and many people are discomfited or outraged by philosophical questions. Some fear their ideas will not stand up. Others take refuge in the "politics of identity": identifying with a tradition, group, nation or ethnicity in a way that licenses ignoring outsiders, declaring their values "incommensurable" with outsiders' and intelligible only to "brothers and sisters within the circle". People like to stay inside thick, comfortable, traditional folkways. Reflection opens them to criticism, and so ideologies become closed circles, primed to feel outraged by the questioning mind.
For two thousand years the philosophical tradition has opposed this complacency. It has insisted that the unexamined life is not worth living (Socrates), that rational reflection can winnow out bad elements in our practices, and that critical self-reflection is a form of freedom: only when we see ourselves properly can we steer.
Marx (as Blackburn reads him). Previous philosophers sought to understand the world; the point is to change it. Blackburn calls this "one of the silliest famous remarks of all time", and says Marx's own intellectual practice belied it.
Blackburn. Without understanding the world you will know little about how to change it for the better. Rosencrantz and Guildenstern cannot play a pipe yet try to play Hamlet, who retorts: "'Sblood, do you think I am easier to be played on than a pipe?"
A harsh reading of Marx. The original (the eleventh Thesis on Feuerbach, 1845) says philosophers have "only interpreted" the world in various ways. Most commentators read it as a complaint that interpretation stopped short, not a dismissal of understanding; Marx spent decades on economic analysis precisely because he thought change needs understanding. Blackburn's parenthesis half-concedes this. Likewise, his brisk verdict on contemporary critics of truth and reason (below) is a declared stance, argued for later in the book, not a neutral survey.
Against the Sceptics of Reason
Some academic currents question truth, reason or the possibility of disinterested reflection. Blackburn's verdict: mostly they do bad philosophy without knowing it, "conceptual engineers who cannot draw a plan, let alone design a structure". The book stands unashamedly with the tradition and against modern or postmodern scepticism about reflection.
He closes with Goya's full motto, which reconciles imagination and reason rather than opposing them:
Imagination abandoned by reason produces impossible monsters: united with her, she is the mother of the arts.
"Conceptual engineering" is now a research programme. When Blackburn used the phrase in 1999 it was a metaphor. Since about 2018 it has become a named field: Herman Cappelen's Fixing Language (2018)* and the volume Conceptual Engineering and Conceptual Ethics (2020, eds. Burgess, Cappelen and Plunkett) ask not only what our concepts are but what they should be, and whether we can deliberately change them. A forerunner is Sally Haslanger's "ameliorative" analysis (2000)* of gender and race, which asks what concepts would best serve our legitimate purposes. Live debates include whether changing a concept changes the topic, and how much control anyone has over meaning.
Concept design is architecture work. In AI governance for banking, many disputes are conceptual before they are technical: what counts as an "explanation" of a credit decision, what "fairness" means when demographic parity and equal error rates conflict, when a model's output is a "decision" rather than "decision support". Blackburn's engineer's question applies directly: how do the parts of the concept function, and what breaks if we change them? Regulators and model-risk teams that skip this step end up building controls around a word nobody has defined.
Knowledge
The childhood thought that the whole world might be a dream is the doorway to epistemology. Blackburn reads Descartes's Meditations as the founding attempt to answer it: doubt everything, find the Cogito as one certain rock, then rebuild via clear and distinct ideas and a non-deceiving God. The rebuilding fails (the trademark argument is weak and the procedure looks circular), which leaves four live options: rational foundationalism, natural foundationalism, coherentism and scepticism, with scepticism permanently beckoning because knowledge implies a sense of our own reliability that outlandish scenarios undermine.
Losing the World
The problem of appearance and reality (things as we take them to be versus things as they are) is as old as thought. It appears in the Indian Vedas (about 1500 BC), and the generation before Descartes had Montaigne, whose motto and essay title was "Que sais-je?", what do I know? But the modern form has a date: 10 November 1619, when René Descartes (1596 to 1650) shut himself in a stove-heated room in Ulm, southern Germany, and had a vision and dreams that he took to reveal his life's work: the one true way to knowledge, which required sweeping away everything taken for granted and building up from the foundations.
Why Descartes Mattered
Descartes was Jesuit-educated, but wrote as Copernicus's heliocentric model and Galileo's "mechanical" science were reshaping the world picture: space containing only material atoms moved by mechanical forces. Both Copernicus and Galileo fell foul of the Inquisition, since if science tells us all there is, it is unclear what becomes of the soul, freedom and God.
- The scientist. Descartes invented standard algebraic notation; Cartesian coordinates, which give algebraic equations for geometric figures, carry his name; he made fundamental advances in physics, especially optics.
- The believer. He was also a pious Catholic, so it mattered to him to show the "vast, cold, inhuman, and mechanical" new world still had room for God, freedom and the human spirit.
- The project. The Meditations (1641) are subtitled as demonstrating the existence of God and the distinction between soul and body. The subtext is to rescue the scientific world view from the charge of atheism and materialism by reflecting on the foundations of knowledge.
Blackburn starts with Descartes because he was the first great philosopher to wrestle with the modern scientific world view; the Greeks and medievals are often too far away for us to think in their shoes. He presents only "edited highlights", like a sportscast, but enough to light up most of later philosophy.
The Six Meditations: the First Two
The Method of Doubt and the Evil Demon
In the First Meditation Descartes resolves that to establish anything "stable and likely to last" in the sciences he must demolish all his ordinary opinions. The case builds in stages:
- The senses deceive. It is "prudent never to trust completely those who have deceived us even once".
- The madman objection. Only madmen, who think they are dressed in purple when naked, have heads of earthenware, or are pumpkins or made of glass, deny the obvious evidence of the senses.
- The dream reply. In dreams we represent things just as convincingly as the senses now do, with no relation to reality.
- The painter objection. Dreams are like paintings: a painter rearranges scenes, but uses real colours. So even if familiar things are imaginary, they depend on simpler, more universal things that are real.
- The Evil Demon. Descartes then supposes that not God but "some malicious demon of the utmost power and cunning" devotes all his energies to deceiving him, so that sky, air, earth, colours, shapes, sounds and all external things are delusions devised to ensnare his judgement.
His only defence is to guard resolutely against believing any falsehood. "A kind of laziness" pulls him back to normal life, but intellectually he must labour in the "inextricable darkness" he has raised.
Cogito, Ergo Sum
The Second Meditation opens with Descartes overwhelmed, supposing he has no senses and no body. Yet if the deceiver is deceiving him, he must exist to be deceived: "I am, I exist, is necessarily true whenever it is put forward by me or conceived in my mind." This is the Cogito: "I think, therefore I am."
Asking what this self is, he finds he cannot imagine it, since imagination pictures corporeal (extended) things, and he knows nothing of those yet. The tall or tubby figure in the mirror is beside the point. What is inseparable from him is thought: he exists for as long as he is thinking, and is "in the strict sense only a thing that thinks".
The ball of wax. Take what people think they understand best: bodies. A piece of wax has taste, scent, colour, shape and size, and makes a sound when rapped. Put it by the fire: the taste goes, the smell goes, the colour changes, the shape is lost, the size increases; it turns liquid and hot and no longer rings. Yet no one denies it is the same wax. So what was understood in the wax was none of the features delivered by the senses. Descartes concludes there is a "pure mental scrutiny" of the wax, and that bodies are perceived "by the intellect alone", so that he can perceive his own mind more easily and evidently than anything else.
Motivating the Doubt: Arguments and Fallacies
Descartes calls his doubt "hyperbolic" (excessive, exaggerated); it is also called Cartesian doubt. How is it motivated? The crude version is a fallacy:
- The senses sometimes deceive us.
- Therefore, for all we know, they always deceive us.
The same form takes you from truth to falsehood: newspapers sometimes make mistakes, so for all we know they always do. Worse, there are instances where the conclusion cannot be true: some banknotes are forgeries, so for all we know all are forgeries. Forgeries are parasitic on the real; forgers need genuine notes to copy.
- Valid. An argument is valid when there is no possible way for the premises to be true and the conclusion false.
- Sound. An argument is sound when it is valid and its premises are true, so its conclusion is true too.
- Suppressed premise. A premise the author presupposed but did not state, which charitable interpretation may supply.
- Weaker conclusion. Charity can also reinterpret an author as aiming at less than the text seems to claim.
The charitable repair adds a premise and weakens the conclusion:
- The senses sometimes deceive us.
- We cannot distinguish the occasions when they do from the occasions when they don't.
- Therefore, for all we know, any particular sense experience may be deceiving us.
This looks valid, but the conclusion has shrunk. It concerns any particular experience, not all experience en bloc: the difference between "any particular note may be a forgery" and "all notes may be forgeries". The first can be true when the second is not.
The Second Premise and Self-Correction
Premise 2 is also suspect. We often cannot detect an illusion at a glance (that is what makes it an illusion), but given time we can: we learn to distrust shimmering water in the desert as a mirage. More damaging, Descartes's own argument presupposes that we can detect deception, since premise 1 is offered as a known truth, and we only know the senses sometimes deceive because further use of the same senses shows it: going closer, looking harder, touching, listening. We know a quick guess at the size of the Sun is wrong only because laborious observation shows it is many times the size of the Earth.
The senses are self-corrective. Premise 2 is true only as "we cannot tell at a glance"; Descartes needs "we cannot tell even over time and with care", which is false. The very evidence that the senses sometimes mislead us comes from the senses themselves.
Dreams and the Pinch Test
Dreams look like a fix: perhaps all experience is a dream. The same distinction applies: we may not tell at once, but memory distinguishes past dreams from past encounters with reality. Still, how would one test it from inside? Pinching yourself fails, since you might dream that the pinch hurts; any cunning experiment might itself be dreamt, along with its reassuring result. Appeals to the scale and coherence of waking life (dreams are jerky and spasmodic; experience is "large and spacious and majestic") invite the reply that the coherence might itself be deceptive. That is what the Evil Demon is for: to show that all experience might be like a dream, totally disconnected from the world.
Two points of interpretation matter:
- The doubt is for the study. Descartes knows that as living agents we cannot keep such doubt alive outside the study; it is psychologically impossible. It matters only for his task of finding foundations.
- Consider, not believe. He does not ask you to believe in the Demon, only to consider it on the way to dismissing it. Unless it can be dismissed, the challenge of scepticism (we have no knowledge at all) remains.
The brain in a vat. A mad scientist removes your brain and keeps it alive in a vat of nutrients, feeding the optic nerve and the nerves of hearing, touch and taste with inputs as if you were living a normal life: eating, playing golf, watching TV. Feedback is built in: send the output "raise hand" and you get experience as of your hand rising. You would apparently have no way of knowing. Blackburn adds a twist: Descartes's version doesn't need vats. Our beliefs about brains and nerves are themselves beliefs about the world, so they too could be inputs; brains and nerves belong to the virtual reality. And coherence won't help, since the Demon can make experience as coherent and large as he likes.
Why the Cogito Survives
A better translation is "I am thinking, therefore I am": the premise is like "I am skiing", not the habitual "I ski". For Descartes thinking includes experiencing, so even in a virtual reality it is I who experience it. From the Demon's point of view, his project hits a logical wall. He cannot make both of these true at once:
- I think that I exist.
- I am wrong about whether I exist.
If (1) is true, I exist to do the thinking, so I am right. Similarly, I can think I am skiing when I am not, but I cannot think that I am thinking when I am not, because thinking that I am thinking is itself an instance of thinking. This is the only case where the thought guarantees its own truth.
The Elusive "I"
Outside the doubt, the thinker is a describable person: for Blackburn, a middle-aged philosophy professor with a personality, history, family and social network. Inside the doubt, all that is swept into the virtual reality. What remains is a pure subject of thought, which might not even have a body. Peering inward, you find experiences, never the "I" that has them; trying to imagine it fails because imagination frames extended things, and we are certain of the self while still uncertain of all extended things.
The Masked Man Fallacy
One reconstruction has Descartes arguing:
- I cannot doubt that I exist.
- I can doubt whether bodies exist.
- Therefore, I am not a body.
Souls are certain, bodies are doubtful, so souls are not bodies. This is superficially plausible but invalid. Parallel: I cannot doubt that I am in the room; I can doubt whether a person who will get bad news tomorrow is in the room; therefore I am not a person who will get bad news tomorrow. "A nice proof with a welcome result!" The pattern is the masked man fallacy: I know who my father is; I don't know who the masked man is; so my father is not the masked man.
What the fallacy shows. Psychological attitudes like knowing and doubting do not track identity: the same thing can be known under one description and doubted under another. So differences in what we can doubt prove no difference in what things are.
Blackburn's charity. He doubts Descartes commits the fallacy in this Meditation. The point there is epistemological: knowledge of the self does not depend on knowledge of bodies, because one can be certain while the other is not. That is not yet a proof that they are distinct.
Lichtenberg: "It Thinks"
Georg Christoph Lichtenberg (1742 to 1799), an aphorist who influenced Nietzsche (1844 to 1900), said we should say "it thinks" just as we say "it thunders"; even "cogito" is too much if translated "I think". There is no "it" that thunders; thunder is going on. Within the doubt, Descartes may be entitled only to "there is a thought going on", not to an "I".
That sounds bizarre: surely there can't be thoughts without a thinker, thoughts floating round a room waiting to be caught, any more than dents floating around waiting for a surface. But if Descartes cannot experience, imagine or confront the self, what entitles him to certainty that it exists, or even tells us what that claim means? (The question returns in Chapter 4.)
The Wax Reconsidered
Descartes sidesteps this by showing the same puzzle about ordinary bodies. Reconstructed:
- At one time my senses report a shape, colour, hardness and taste belonging to the wax.
- At another time they report different ones.
- The senses show me nothing but these diverse sensory qualities.
- Yet I judge that it is the same wax on both occasions.
- So it is the nature of the wax to be able to bear different sensory qualities at different times.
- Therefore, to understand what the wax is I must use my understanding, not my senses.
He is not denying that the senses tell me the wax is there. They are messengers delivering information that needs interpreting, and identifying one object across many appearances is the work of the understanding, using categories whose credentials can themselves be investigated. So the wax is some elusive "thing" that takes on different bodily properties, and the self some equally elusive "thing" that thinks different thoughts. Perhaps selves are no harder to understand than bodies, and we think otherwise only from prejudice. (The wax returns in Chapter 7.)
Climbing Off the Rock
The first two Meditations combine depth, imagination and rigour to a degree seldom matched. But they leave Descartes on bedrock with no building materials. His own standard, demon-proof knowledge, seems to forbid even "self-evident" reasoning, since the Demon could deceive us with bad reasoning as easily as with false perceptions; our reasonings are more fallible than our senses.
Clear and Distinct Ideas
Descartes instead asks what makes the Cogito so certain, and decides it is his "clear and distinct" perception of it. The model is mathematical. Draw a circle, a diameter, and chords from both ends to any point on the circumference: they always meet at a right angle. At first you only suspect a reason. Then you go through the proof (join the centre to the apex and solve the two triangles) and just see that it must hold, perhaps in a "flash" of certainty.
- Rationalism. The ideal that pure unaided reason can see from the armchair that things must be one way and cannot be otherwise, and that mind, body, God, freedom and human life might be grasped with the same rush of clarity.
- A priori knowledge. Knowledge achieved by rational insight, seen to be true without any experience of the way of the world.
The Trademark Argument
Looking into himself, Descartes finds an idea of perfection, argues that only a perfect cause could produce it, and concludes that God exists and stamped the idea in us like a craftsman's trademark. Since God is perfect, he is no deceiver, so if we trust only clear and distinct ideas, the world will be as we understand it. Reconstructed:
- I have the idea of a perfect being.
- This idea must have a cause.
- A cause must be at least as perfect as its effect.
- So something at least as perfect as my idea caused it.
- Therefore such a thing exists, and it must be perfect: God.
- Premise 1 is not free. Some theological traditions deny we have any idea of God's perfection, which defies understanding.
- Premise 2 is not demon-proof. On his lonely rock, Descartes cannot exclude events that "just happen" without cause; and if he believes otherwise, the Demon might be making him believe it.
- The perfectly punctual person. I can define perfect punctuality (never late, perhaps never early) without meeting anyone like that, and even if nobody ever is. Ideas of perfection need no perfect cause.
- The perfect mathematician reply. Descartes might say a perfect mathematician would be inventive, and I cannot grasp inventions before they arrive, so only a perfect mathematician could give me a clear and distinct idea of one. But then I plainly lack a clear and distinct idea of perfection; my idea is a punctuality-style list (never errs, never unkind, never unable) plus perhaps an imagined glow, which adds nothing. Claiming full comprehension of God would be presumptuous, even blasphemous.
- Descartes's own mountain. He wrote that we can touch a mountain but not put our arms around it: to know something it is enough to touch it in thought. If we only touch God's qualities by definition, we cannot argue back to an archetype that let us comprehend them.
- The relay-baton theory of causation. Premise 3 rests on an inherited principle that a cause passes something on to its effect (heat makes hot, motion makes motion). It is not even true: a piece of iron moving in a magnetic field causes an electric current, which resembles nothing in the cause.
- Idea of X shares X. Descartes may have slipped into thinking an idea of infinity is an infinite idea, so an idea of perfection is a perfect idea. (Is an idea of something solid a solid idea?)
Most readers find the argument easy to resist even outside extreme doubt, let alone demon-proof. Descartes's optimism owed much to inherited assumptions, the baton principle above all.
The Cartesian Circle
With God guaranteeing clear and distinct ideas, the Demon seems to have been quietly forgotten. Antoine Arnauld (1612 to 1694) cried foul: Descartes argues in a circle. Write CDp → Tp for "if I clearly and distinctly perceive p, then p is true", and G for "God exists and does not deceive".
- At some points Descartes holds: I can know (CDp → Tp) only if I first know G.
- At other points he holds: I can know G only if I first know (CDp → Tp).
- So neither can come first, and the rebuilding never starts.
It is the morning impasse: you need coffee to get out of bed, and you need to get out of bed to make the coffee.
Escape 1: God validates memory only. While you are actually perceiving something clearly and distinctly, you need no guarantee at all. G is needed only later, when you have forgotten the proof and must trust your memory that you once proved it.
Escape 2: no general rule needed. Descartes sees G clearly and distinctly without first holding the general rule (CDp → Tp). He is certain of an instance without being certain of the rule.
The second escape carries a very important truth: we are often more certain of particular verdicts than of the principles we would cite to defend them. I can know a sentence is grammatical without knowing any rule that licenses it. I recognise a Pomeranian, a member of the Rolling Stones or my wife without knowing the principles (perhaps "algorithms" in my perceptual system) that produce the verdict. Socrates habitually got people to assert something, showed they could not defend it with general principles, and concluded they had no right to their claim. The grammar case shows that inference is bad, and the habit "is only apt to give philosophers a bad name".
Still, why should Descartes's "seeing" that God exists be a clear and distinct case of seeing the truth? Blackburn's "dark suspicion" is that mention of God clouds the mind rather than clarifying it. Either way there is a whiff of double standards: the Demon is forgotten as soon as Descartes wants off his rock, which suggests he has stranded himself on a desert island with no escape.
Foundations and Webs
David Hume (1711 to 1776) gave the decisive criticism. Cartesian doubt recommends doubting even our faculties, then reassuring ourselves by reasoning from some infallible original principle. But no principle has that prerogative, and even if one did, we could not advance beyond it except by using the very faculties we are doubting.
If the project is to use reason to defend reason against universal doubt, it must fail: the Cartesian doubt, were it achievable (and it plainly is not), would be entirely incurable.
So what follows? Blackburn sets out the alternatives.
Natural Foundationalism (Hume)
Hume keeps foundations but denies them Descartes's rational status. The veracity of our senses and reasoning is itself part of the foundation; it cannot be demonstrated from anything deeper. Outside the study, trusting common experience comes naturally; growing up, we learn the danger areas (illusions, mirages) against a background of natural belief, and self-correction is all we need. Blackburn calls this non-rational or natural foundationalism: not irrational, only lacking a demon-proof way of "standing to reason".
- Empiricists and rationalists. The British philosophers of the seventeenth and eighteenth centuries distrusted unaided reason; their paradigm of mind meeting world was seeing and touching a familiar object, not a proof crystallising. Hence empiricists, against the card-carrying rationalist Descartes.
- Labels leak. Under pressure Descartes says the good thing about clear and distinct ideas is that you cannot doubt them while you have them, a natural potency much like Hume's basic beliefs; and John Locke (1632 to 1704), champion of empiricism, is elsewhere "as rationalist as the best of them". Great philosophers resist labelling.
- God as a label. On this view, whether experience and reasoning en bloc match the world en bloc can only be settled by an act of faith; "God" just names whatever ensures the harmony. Normal life never raises the question, so the hyperbolic doubt is "unreal".
This can sound complacent, but calling the doubt unreal does not mean ignoring the gap between appearance and reality; we can study it from within our ordinary beliefs. When Hume did exactly that, he found difficulties strong enough to reintroduce scepticism (Chapter 7).
The Evolutionary Reassurance
One optimism is available two centuries after Hume: evolution would not have selected our senses and reasoning, in their present shape, if they did not work. Eyesight that failed to report predators, food and mates when they were present would be useless. The harmony between our minds and the world exists because the world is responsible for our minds. This is not an answer to the Demon, since it relies on things we take ourselves to know about the world, and those, Hume will show, can themselves be made doubtful.
"Built to get things right" is contested. Evolution selects for fitness, which tracks truth only where truth pays. Alvin Plantinga's evolutionary argument against naturalism (1993) claims that, given unguided evolution, the probability that our faculties are reliable is low or inscrutable. Donald Hoffman's "interface theory" (The Case Against Reality, 2019) argues from evolutionary game simulations that perception is tuned to fitness payoffs, not to the structure of reality. Most philosophers think the reliability argument works well for middle-sized, survival-relevant matters (is there a predator?) and much less well for theoretical beliefs about quantum fields or the age of the cosmos, which evolution never tested.
Coherentism (Neurath)
A different response drops foundations altogether. Waking beliefs hang together, dreams are fragmentary, and coherent structures need no foundations: a ship or a web gets its strength from interconnections, each part supported by others, with no part supporting everything else while unsupported itself. A challenged belief is defended by the rest; if nothing supports it, drop it. Otto Neurath (1882 to 1945):
We are like sailors who on the open sea must reconstruct their ship but are never able to start afresh from the bottom.
Any plank can be replaced provided there is enough of the rest to stand on, but the whole cannot be challenged en bloc. The coherentist motto: every argument needs premises, but nothing is the premise of every argument. The trouble is that we seem to understand the Demon scenario: a belief system extensive, coherent, interlocking and completely wrong. If the choice is coherentism or scepticism, Descartes might say scepticism is the more honest.
| Option | Champion | Structure of knowledge | Answer to the Demon | Main weakness |
|---|---|---|---|---|
| Rational foundationalism | Descartes | Built up from demon-proof certainties (Cogito, clear and distinct ideas) | Prove God exists and is no deceiver | Circular; cannot reason past the Cogito without trusting reason |
| Natural foundationalism | Hume | Rests on natural trust in senses and reason, corrected over time | The doubt is unreal; settling it en bloc takes faith | Can look complacent; Hume's own later doubts reopen the gap |
| Coherentism | Neurath | A web or ship: mutual support, no base | Doubt must be piecemeal, never en bloc | A coherent system could still be wholly false |
| Scepticism | Broods over all | No knowledge | The Demon cannot be ruled out | Unliveable outside the study |
Whichever the reader prefers, "he or she will find good philosophical company". One might think Descartes got almost everything right or almost everything wrong; the baffling thing is to defend either answer.
Local Scepticisms
Scepticism can be local: confidence in science combined with doubt about ethics, politics or literary criticism, where mere caution, not hyperbolic doubt, unsettles us. But some general local scepticisms are just as baffling as the Demon.
Russell's young world. Bertrand Russell (1872 to 1970) asked how I know the world did not spring into existence a few moments ago, complete with delusive traces of great age, including the brain modifications we take to be memories. Victorian thinkers reconciling Genesis with the fossil record said much the same of geology: God laid down misleading evidence of an ancient Earth a few thousand years ago (and, we may add, misleading signs of an ancient universe). The move was never popular, perhaps because scepticism about time quickly becomes scepticism about everything, or because it makes God a large-scale practical joker.
The twist is that Russell's scenario can be argued to be scientifically more probable than common sense. Low-entropy (highly ordered) systems are improbable, and as systems evolve entropy increases: smoke never returns to the cigarette, toothpaste never goes back into the tube. So it is "easier" for a moderately disordered world like ours to arise from nothing than for its lower-entropy ancestor to have done so. Compare Scrabble: there are far more ways to draw a seven-tile hand containing a four- or five-letter word than one containing a seven-letter word. "In a straight competition for probability between Russell's outlandish hypothesis and common sense, Russell wins." Blackburn leaves it for the reader to ponder.
The figures need updating, and one is misstated. Victorian young-Earth views (the apparent-age version is Philip Gosse's Omphalos, 1857) typically followed Archbishop Ussher's chronology, which dates creation to 4004 BC, about 6,000 years ago, not the 4,000 the text implies. Current estimates put the Earth at about 4.54 billion years (Blackburn: about 4,000 million) and the universe at about 13.8 billion years (Blackburn: about 13,000 million). None of this affects the argument.
Russell's puzzle now has a name: Boltzmann brains. The entropy argument is a version of a problem physicists take seriously. If the universe is very long-lived and fluctuates, random fluctuations producing a single momentary brain with false memories should vastly outnumber whole low-entropy universes, so a typical observer "should" be such a Boltzmann brain. Physicists such as Sean Carroll treat a cosmological model that predicts this as "cognitively unstable": if it were true, the evidence for it would be unreliable, so we reject it. The standard escape, defended by David Albert (Time and Chance, 2000), is the Past Hypothesis: we simply posit that the universe began in a very low-entropy state, and condition our probabilities on that. Note the shape of the move: it is a natural-foundationalist posit, not a demonstration.
The Moral: Knowledge and Reliability
Knowledge implies authority: those who know are those we should listen to. It implies reliability: those who know register the truth like good instruments. Claiming knowledge implies claiming a sense of one's own reliability, and granting authority to a person or method means seeing it as reliable.
The scenarios of Descartes and Russell unseat that sense. We could regain it by showing them impossible, or at least without real chance, but they are hard to prove impossible, and in such abstract matters we have no good sense of probabilities. Any argument that they have no chance relies on the very opinions they query. So scepticism permanently beckons, or threatens. In engineering terms, the structure of our thought spans a gap between how things appear and how they might be, and we hand ourselves the right to cross it. Without a good sense of our own reliability, any confidence in harmony between belief and world looks like a pure act of faith.
Descartes left three problems: knowledge; the place of the mind in nature (Chapter 2); and, through the scientific revolution he helped found, understanding the world we are placed in.
How the debate has moved since 1999.
- Putnam's semantic reply (1981, already available). Hilary Putnam argued in Reason, Truth and History that a brain that had always been in a vat could not even think "I am a brain in a vat" truly, because its word "vat" would refer to virtual vats, not real ones. The argument relies on semantic externalism and is still debated; it does nothing against a brain recently envatted.
- The simulation argument (2003). Nick Bostrom argued that at least one of three is true: civilisations almost never reach the ability to run ancestor simulations; those that can almost never do; or we are almost certainly living in a simulation. It turns the Demon into a probabilistic hypothesis with empirical premises.
- Virtual realism (2003, 2022). David Chalmers ("The Matrix as Metaphysics", 2003; Reality+, 2022) argues that if we are in a simulation, our tables and chairs are still real, made of digital processes rather than whatever we thought; so the simulation hypothesis is not a sceptical hypothesis about most of our beliefs.
- Knowledge first (2000). Timothy Williamson's Knowledge and Its Limits argues that knowledge is basic and cannot be analysed into belief plus extra conditions, and that our evidence is just what we know, which blocks the sceptic's claim that the envatted and the embodied share the same evidence.
- Contextualism. David Lewis ("Elusive Knowledge", 1996) and Keith DeRose hold that "knows" sets higher standards when sceptical possibilities are raised: in the study you don't "know" you have hands, in the street you do. This is one formal version of Hume's point that the doubt is real only in the study.
- Predictive processing. Neuroscientists and philosophers such as Andy Clark and Anil Seth model perception as the brain's best prediction of the causes of its inputs, corrected by prediction error; Seth calls it a "controlled hallucination". On this picture dreaming is prediction with the error signal switched off, which gives a scientific gloss on Descartes's point that waking and dreaming share their machinery and differ in how they are checked against the world.
What the chapter leaves out. Blackburn treats knowledge through reliability and scepticism and never pauses on what knowledge is. The standard starting point, knowledge as justified true belief, was undermined by Edmund Gettier (1963), whose cases show a person can have a justified true belief by luck and still not know. Much later epistemology (Goldman's reliabilism, virtue epistemology, Williamson's knowledge-first view) is a response to Gettier. Blackburn's emphasis on reliability fits reliabilism closely, but the reader is not told this is one theory among several.
The flawless counterfeit. Suppose a counterfeiter produces notes physically identical to genuine ones, down to the molecule, and slips them into circulation alongside real ones. Now "for all we know, any particular note is a forgery" is true and no amount of careful inspection will help. Notice what this does to Blackburn's self-correction reply: it works only where errors leave detectable traces. The Demon and the vat are precisely traceless errors. The thought experiment shows why the en bloc doubt survives the refutation of the crude argument: it is not about the frequency of error but about whether error would be detectable even in principle.
Reliability is the right frame for AI outputs. Blackburn's moral, that knowledge implies authority and authority requires a sense of reliability, is the logic of model governance. A model that is usually right but cannot tell you when it is wrong is like the senses at a glance; what earns trust is the self-corrective loop: calibration measured against ground truth, monitoring for drift, and evaluation sets the model never saw. Hallucinated citations are traceless errors in the counterfeit sense above, which is why retrieval with source verification, not more fluent generation, is what restores warrant. Synthetic media add a practical twist: as deepfakes improve, the "check over time" route needs provenance infrastructure (signed capture, content credentials), because inspection of the item alone stops working.
Mind
Descartes's dualism, mind as a ghost in the bodily machine, leaves wide open the possibility that others are Zombies (no experience) or Mutants (different experience), and makes the analogy from one's own case hopelessly weak. Blackburn uses these possibilities as a test that dualism fails, then frames the whole mind-body problem as a debate between Locke (mental life is annexed to the physical by God's good pleasure, a brute fact) and Leibniz (the mental must be a rational expression of the physical). He sides with Leibniz in spirit, tries behaviourism, functionalism and identity theory, and shows how colour science can close the gap the inverted spectrum seemed to open.
The Problem Left Over
Set aside the general harmony between belief and world and assume we know what we take ourselves to know. Descartes still leaves each of us with our own mind as a special, intimate object of immediate knowledge. Even if I climb onto the Cogito, I cannot climb onto the nature of your mind. How do I know you see blue as I do? Might some feel pain more but make less fuss, and others feel less but make more?
The Ghost in the Machine
Descartes was also a working scientist: he made foundational discoveries in optics, practised dissection, and knew that nerves carry impulses to the brain by physical transmission (a "pull" or "violent motion", for us an electrochemical impulse). Senses activate nerves, nerves signal the brain, and patterns of activation spread within it and back to the body. All of this is neurophysiology, and in principle it can be shown to a classroom with the right instruments.
Then comes "the magic moment": the res cogitans (thinking thing) is affected and the world of experience opens up: colours, sounds, textures, temperatures, tastes, smells. These mental events are private: "the whole classroom may see some neurones firing, but only the one person feels the pain". Descartes, for quite sensible neurophysiological reasons, located the crossing point in the pineal gland, a central structure in the brain.
- Substance dualism (Descartes). There are two kinds of bearer: mental events are modifications of an immaterial substance, "ghost-stuff or ectoplasm"; movement and position belong to matter. In "I thought of the Queen and I saluted", strictly two different "I"s are involved. It is theologically convenient, since soul-stuff need not share the body's life span, opening the way to immortality.
- Property dualism. There are two kinds of property, mental and physical, but one kind of bearer, whatever large animals are made of. Compare mass and velocity: very different properties, both had by one projectile.
- The ghost in the machine. Gilbert Ryle's (1900 to 1976) label for the Cartesian picture: machine events are like other physical events (molecules, atoms, fields, forces), ghost events are altogether different, yet the two correlate and interact (a pin causes pain; remembering a blunder causes groaning and blushing).
The view is not peculiar to Descartes; it is presupposed by many world religions and by any doctrine that we survive bodily death or that soul and body can go separate ways. Yet it faces "enormous, and arguably insurmountable, problems".
Zombies and Mutants
The first problems are epistemological. Dualism says mental and physical events correlate closely in normal humans, but what entitles us to believe that?
The Zombie possibility. Zombies look and behave exactly like us; open their brains and they function exactly as ours do; prick one and it says "ouch". But there is no ghost within: no consciousness at all. Nothing could tell us which people are Zombies. Having raised the possibility, I see I cannot be sure about you; perhaps consciousness is an extremely rare correlate of complex brains and I am the only example.
The Mutant possibility. Mutants are physically and behaviourally identical to us and are conscious, but their experiences bear no relation to ours. A pricked Mutant may experience something like hearing middle C on a clarinet, yet still say "ouch", because her brain works like ours. When she does hear middle C, perhaps she feels awful pain but smiles beatifically. She may see British post-boxes as yellow and daffodils as blue. Perhaps, compared with me, the rest of you are all Mutants.
On the Cartesian account these possibilities are wide open. They are unnerving, and Blackburn suspects they cross our minds more often than the Demon of Chapter 1.
The Argument from Analogy
One response is to bite the bullet: I can never know others' minds, but I can reasonably use myself as a model for the rest, a conjecture rather than knowledge. This is the argument from analogy for other minds:
- In my own case, a body and brain of this kind go with a mental life of this kind.
- Others have bodies and brains of the same kind.
- Therefore, probably, others have mental lives of the same kind.
It seems incredibly weak. Ludwig Wittgenstein (1889 to 1951) asked: "how can I generalize the one case so irresponsibly?" One case, perhaps arising by luck, is flimsy ground for a universal association.
The beetle in the box. Blackburn's use: if I have a box with a beetle in it, that gives me very poor grounds for thinking everyone else's box contains a beetle. Worse, it gives poor grounds for denying beetles exist outside boxes: maybe things physically very unlike us, rocks or flowers, are conscious just as I am. A sample of one supports no generalisation, in either direction.
Wittgenstein's beetle makes a stronger point. In the Philosophical Investigations (§293), the beetle in the box is not mainly about weak induction. Everyone has a box only they can look into, and calls what is inside "beetle". Wittgenstein's conclusion is that the thing in the box drops out of consideration as irrelevant: it could be different for everyone, or constantly changing, or absent altogether, and the word "beetle" would still work in the shared language. The moral is that if sensation words got their meaning from private objects, the objects would play no role in that meaning. Blackburn reaches this deeper point later, under the private language argument and Wittgenstein's objection that one could not even form the concept of another's pain.
Why Not Shrug Them Off?
You might dismiss Zombies and Mutants as unverifiable philosophical fantasy. Blackburn calls that "not an intelligent reaction". It is true that neurophysiologists cannot display conscious experience on the lecture-theatre screen. But on Cartesian dualism the possibilities we all believe in, that others are neither Zombies nor Mutants, are equally unverifiable: "blind articles of faith". The Zombie believer is no worse off than the rest of us. Indeed, if it is not a priori false that others are Zombies, why should it be a priori less probable than that they are conscious like me?
Bizarre possibilities are instruments for testing a conception. If mind and matter are conceived the Cartesian way, intolerable possibilities about which we could know nothing open up. The aim is not to wallow in scepticism but to draw back from any philosophy that opens up the sceptical possibilities: if Cartesian dualism leaves Zombies and Mutants wide open, that shows something is wrong with Cartesian dualism.
The positive claim: it is really not possible that someone howling over a stubbed toe is doing so because they are in a state like my hearing middle C on a clarinet. That state cannot be expressed by howling. The tie between what a mental state feels like and its expression is closer than dualism allows; we know the toe-stubber feels something very like what I feel when I stub mine.
Wittgenstein's Deeper Objection
Wittgenstein's main objection to the analogy argument is not its weakness. He argues that if you learned about mental events entirely from your own case, you could not even think of other people's consciousness. If I drop a brick on your toe, there would simply be no pain about (I feel none), and that would be that. Since we do think in terms of others' experience, we must conceptualise minds some other way.
The Metaphysical Problem: What Is Consciousness Doing?
There is also metaphysical pressure. The Zombie's physical functioning is identical to ours; his projects succeed and fail the same way, his health depends on the same variables, he laughs in the right places, weeps at tragedies, and is good fun to be with. So what is the lack of consciousness doing, or what is consciousness doing for us?
- Epiphenomenalism. Perhaps in us mental events exist but do nothing, like the whistle on a steam engine, no part of the machinery that makes things happen.
- The evolution problem. If minds do nothing, why did they evolve? Why did nature go in for them?
- The memory problem. If mental states do nothing, how do they enter memory?
This is the problem of brain-mind interaction as it confronts Cartesian dualism.
Locke, Leibniz and God's Good Pleasure
The issue is beautifully captured in a debate between John Locke and Gottfried Wilhelm Leibniz (1646 to 1716). Locke worried about the moment when motions of particles in the brain give rise to ideas such as colour in the mind. A violet, by the impulse of insensible particles of particular shapes and motions, produces the ideas of blue colour and sweet scent; it is no harder to conceive God "annexing" such ideas to motions they do not resemble than to conceive him annexing pain to the motion of steel dividing our flesh.
- Intelligible causation. Like Descartes and Newton, Locke found some causal processes intelligible, chiefly motion passed from particle to particle by impact.
- Opaque causation. Body-to-mind causation, motions producing something entirely different (smell, colour, pain), is wholly obscure. It is due to the "arbitrary will and good pleasure" of God, "the wise architect", who annexes conscious states to physical events.
- Brute fact. In Descartes's terms, we have no clear and distinct idea of which systems God might choose to "superadd" consciousness to. That some systems are conscious, and change consciousness when they change physically, is a brute fact.
Occasionalism
Locke is close to the occasionalism of Nicolas Malebranche (1638 to 1715): physical events do not cause mental events at all; they are occasions on which God himself inserts appropriate mental events into our biographies. There is "precious little difference" between God arranging that steel in flesh brings about pain and God directly injecting pain whenever steel divides flesh.
Leibniz's Reply
In the New Essays, a blow-by-blow commentary on Locke, Leibniz's spokesman Theophilus answers Locke's spokesman Philalethes, quoting the violet passage. Ideas of colour and pain are not arbitrary; "it is not God's way to act in such an unruly and unreasoned fashion". There is a resemblance of a kind, the way an ellipse, parabola or hyperbola resembles the circle of which it is a projection on a plane, each point corresponding to a point by a precise relation. Pain does not resemble the movement of a pin, but it may thoroughly resemble the motions the pin causes in our body, and represent them in the soul.
Locke: two jobs for God. First fix all the physics and its laws; second, decide how to annex mental events to physical ones. The world has two biographies, physical and mental, and their relation is a separate decision. Independent variation is possible: God could keep the physics fixed and not annex pain to pinpricks. So your physical twin could be a Zombie or a Mutant; on occasionalism, God might insert pain when I stub my toe but not when you do.
Leibniz: one job for God. Fix the physical stuff and laws, and everything else follows. To make pinpricks painless, God would have to change the nerves and pathways. No independent variation: you and your twin stubbing toes with the same force must have the same mental "expression", just as identical shapes projected at the same angle cast identical figures. (On another reading, Leibniz allows variation but God has chosen the best pairing.)
The mathematical analogy is telling. Leibniz was an even better mathematician than Descartes and invented the calculus, but more to the point he held that the whole order of nature must eventually be transparent to reason. His principle of sufficient reason: when things fall out one way, there must be a reason why they do, which we should be able to see. "God does nothing arbitrary" is less theological optimism than a demand for intelligibility. That confidence makes Leibniz, like Descartes, a rationalist.
The Leibnizian must therefore deny the possibility of Zombies and Mutants: fix the physical biography and the mental biography is fixed with it. The philosophical task is to understand how the entire physical story makes the mental story true.
Locke's Rationalism
Locke left open whether an immaterial thing or the physical system itself does our thinking, since God can superadd thought to anything he likes. But he was clear that "it takes a mind to make a mind": unthinking particles, however arranged, gain only "a new relation of position", which cannot yield thought. This a priori certainty about what can cause what marks Locke too as fundamentally a rationalist, if a more nervous one than Descartes and Leibniz.
Leibniz's mill. Leibniz has his own famous thought experiment (Monadology, §17). Imagine a machine that thinks and perceives, then enlarge it to the size of a mill so you can walk inside. You would find only parts pushing one another, never anything that explains a perception. Leibniz concluded that perception cannot be explained mechanically and must be sought in a simple, unextended substance. The image gives the opposing intuition its sharpest form: inspection of mechanism reveals only mechanism. It also shows that the "Leibnizian" position in Blackburn's debate is a reconstruction: the historical Leibniz was no materialist, and his "expression" of body in soul sits inside a metaphysics of monads and pre-established harmony, in which mind and body do not causally interact at all.
Where the Debate Stands
"Thinkers about mind and matter have not got much beyond Locke and Leibniz." Two contemporary camps stay with Locke:
- The new mysterians. We shall never understand the mind-matter relation; it stays rationally inexplicable, God's good pleasure in secular dress.
- Modern epiphenomenalist dualists. Since the physical is a closed system, a complete physical chain runs from pin to wince and explains the wince; so it must be false that you wince because you are in pain. This "bit of common sense has to be given up", and these thinkers inherit Locke's interaction problem (Chapter 3).
On the Leibnizian side Blackburn introduces two approaches: analysis (showing the mental as an a priori expression of the physical) and scientific reduction or identity.
Analysis
Analysis tries to say what makes true some mysterious kind of statement, using terms from a less mysterious class. The homely example is the average man, with 2.4 children and 1.8 automobiles. Statements about him are made true by total children divided by number of progenitors and automobiles divided by owners; he is what Russell called a "logical construction" out of aggregates of facts. (Not every statement about averages is useful: the average person has one testicle and one breast.) Analyses provide reductions of one kind of statement to another.
The method's standing has varied. Russell and G. E. Moore (1873 to 1958) thought analysis the essential goal of philosophy; W. V. Quine (1908 to 2000) and others queried its prospects, helped by "the depressing fact that very few philosophical analyses seemed successful"; in 1999 it was enjoying a cautious revival. For present purposes the point is simple: analysing the mental in physical terms would vindicate the Leibnizian dream of an a priori route from physical to mental.
Logical Behaviourism
Take pain. We identify it chiefly by what it makes us do, which is also what it is for, in evolutionary terms: it demands attention, makes us immobilise the injured part, distracts us, and is unpleasant. Sum these up as dispositions to behaviour and propose that being in pain just is being so disposed. That would be an a priori result, the mystery of consciousness would vanish, and you and your twin, sharing dispositions, would share sensations, because that is what sensations are. This is logical behaviourism, and Blackburn believes "there is something right about it".
- Same pain, different behaviour. You can stub your toe and make a fearful fuss one day, and smile bravely with the same pain another. Behaviour is not a transparent guide to feeling. (Hence the joke about two behaviourists in bed: "That was great for you, how was it for me?")
- Reply: suppressed dispositions. The brave smiler is still disposed to more expressive pain behaviour that she suppresses. Total suppression is almost impossible, and observers are good at telling a child who isn't hurt from one who is hurt but being brave. Disposing in this way seems essential to pain.
- Brain damage cases. Some brain-damaged patients sincerely say the pain is still there but they no longer mind it.
- Reply: hard to make sense of. Imagine a sturdy pain (touching a hotplate, swinging your toe into the wall): it is very hard to imagine that very state without imagining it as incredibly unpleasant and as tending to cause typical behaviour.
Pain asymbolia is now well documented. The brain-damage cases Blackburn mentions are known as pain asymbolia, typically linked to lesions of the insula; patients detect and can describe painful stimuli but show little aversion, sometimes even laughing. Related dissociations followed frontal lobotomy and cingulotomy. Neuroscience now routinely distinguishes sensory-discriminative components of pain from affective-motivational ones, carried by partly different pathways. This supports Blackburn's opponents in one way (sensation and aversion can come apart) and his reply in another: what remains after the split is something patients themselves are unsure whether to call pain.
Functionalism
Contemporary thinkers prefer functionalism, which also identifies a mental state by its function but more relaxedly: a network of relations including typical causes, behavioural dispositions and effects on other mental states, provided those are eventually expressed in physical dispositions. The idea is essentially the same as behaviourism's.
Qualia: The Hard Residue
Pain suits analysis because it has a distinctive natural expression; so do sadness, fear, anger and joy. But some states relate to behaviour only very indirectly: the taste of coffee. There is something it is like to taste coffee (not for Zombies), but it does not make us do much. Contemporary philosophers speak of qualia or raw feels, and friends of qualia are often glum about reducing them to dispositions. They are thereby back with Locke: qualia are superadded, could have been otherwise, and Zombie and Mutant scepticism returns.
A Scientific Model
A popular middle route rejects Locke's happenstance without demanding Leibniz's rational transparency: perhaps there is a metaphysical identity between mental and physical facts that cannot be known a priori.
Temperature and kinetic energy. Classical physics identifies the temperature of a gas with the mean kinetic energy of its molecules. God has one thing to fix: fix the molecules' energy and the temperature is fixed. There are no Zombie or Mutant gases. Yet the identification was not an armchair result: it took experiment, observation and general theoretical considerations, so it is mostly a posteriori, unlike the purely mathematical fact that a circle on a tilted plane casts an ellipse.
Such identifications run through bridge principles within the sciences. Genes are defined in classical biology by their function of making characteristics heritable; molecular biology found that bits of DNA have that function. Analysis is not entirely absent (we must know what genes are meant to do), but the big discovery is contingent and scientific: what does the job.
Applied to mind: find a brain state that all and only people in pain share (vaguely, "C-fibres firing") and propose that this is pain, as DNA segments are genes. This is the psycho-physical identity theory, a complete reduction of the mental to the physical.
- "Feigning anaesthesia." Opponents say you can hold it only by pretending to feel nothing: everything distinctively mental has been left out. Blackburn's rebuttal: ask the challenger what exactly has been left out, "and watch him squirm on the difficulties of dualism".
- First-person authority. From the subject's view, anything that feels like pain is pain, whether C-fibres or something else. After a mini-transplant replacing organic C-fibres with silicon that produces the same results, it is still pain. Knowledge of one's pain is not hostage to one's biological engineering.
- Marginal creatures. We might know whether shrimp have C-fibres yet be uncomfortable declaring them to suffer, or not, purely on that basis. So the identity is less straightforward than temperature (though "this could be challenged").
The stock example is neurologically naive. C-fibres are unmyelinated peripheral nerve fibres that carry slow, burning pain signals to the spinal cord; their firing is neither necessary nor sufficient for felt pain (phantom-limb pain occurs without them; anaesthetised patients' C-fibres fire without pain). Philosophers use "C-fibres firing" as a placeholder for "whatever the brain state turns out to be", and Blackburn's "vaguely" signals this. The silicon point is the classic multiple realisability objection, pressed by Hilary Putnam (1967), which pushed most philosophers from identity theory towards functionalism.
Blackburn would be content to see mental and brain events related as clearly as temperature and kinetic energy, whether by pure thought or by experiment, so we can share Leibniz's objection to Locke without his full rationalism. But when we think hard about mind and brain, "it usually seems to be our thinking rather than mere scientific ignorance that is letting us down."
- Correlation is not intelligibility. Scientists have identified brain states implicated in normal conscious functioning (Blackburn mentions brain waves of a particular frequency). From Locke's side, this shows only what consciousness is annexed to, by happenstance; it does not make the combination intelligible.
- Correlation presupposes the answer. Establishing any correlation already assumes the right to shove Zombies and Mutants aside; otherwise the scientist could establish it at best in her own case.
- The mysterian verdict. Neither science nor philosophy will ever let us side wholeheartedly with Leibniz.
The science of consciousness since 1999. The brain-wave reference is presumably to the ~40 Hz (gamma) oscillations that Francis Crick and Christof Koch proposed in 1990 as a correlate of consciousness; that specific proposal has since lost favour. The field has since organised around rival theories: Global Neuronal Workspace (Dehaene, Changeux, Baars), Integrated Information Theory (Tononi, Koch), higher-order theories and recurrent processing theory. A 2023 adversarial collaboration (the Cogitate consortium) pitted IIT against GNW with pre-registered predictions; both theories had key predictions fail. Later in 2023 an open letter signed by over a hundred researchers called IIT "pseudoscience", provoking a sharp dispute. Blackburn's point survives all of it: these programmes find correlates and mechanisms of access, and whether that closes Locke's gap is exactly what is argued about. David Chalmers's "hard problem" (1995) and his zombie conceivability argument (The Conscious Mind, 1996) gave the Lockean side its modern form.
Two newer moves on qualia. Illusionism (Keith Frankish, 2016, building on Daniel Dennett) says phenomenal consciousness as the qualia-friend conceives it does not exist; what exists is a powerful introspective representation of our states as having ineffable qualities, and the real problem is the "illusion problem": why we represent ourselves that way. This is a radical Leibnizian move: close the gap by denying one side of it. At the other extreme, panpsychism has revived (Philip Goff, Galen Strawson), taking seriously Blackburn's passing thought about beetles in rocks and flowers: perhaps experience is a basic feature of matter.
Inverted Spectra and Private Languages
Colour seems to open the Mutant possibility widest: physically identical people whose colour spectra are completely inverted, so that what one experiences from red-end light the other experiences from blue-end light, with nothing to tell them.
Turning the Doubt on Yourself
Dualism seems at least to leave us secure about our own experience. But apply the Zombie and Mutant possibilities to your own past: do things look the same colour today as yesterday? Did they look any colour at all, or are you merely remembering experiences you never had? We would surely notice if the sky now looked like grass did yesterday. But is that security guaranteed by dualism?
- Physical memory is checkable. I remember putting the car in the garage, and there it is; I remember the way to the kitchen and get there.
- Mental memory is not. On dualism, what checks my memory of the mental world? Why should it not be God's good pleasure to annex certain experiences to me today together with a delusive memory that similar ones were annexed yesterday?
Always get rid of the idea of the private object in this way: assume that it constantly changes, but that you do not notice the change because your memory constantly deceives you.
This is the heart of the private language argument in Wittgenstein's Philosophical Investigations (published posthumously, 1953), "one of the most celebrated arguments of twentieth-century philosophy". If mental life were divorced from the physical as dualism says, it would be "too slippery or ghostly" to be an object even of our own memories or intentions, and no significant thought about one's past or future mental life would be possible. So the possibilities should unnerve us only about the dualist picture.
The private diary. Wittgenstein's own illustration (Investigations §258): I decide to keep a diary of a recurring sensation, writing "S" each time it occurs, with no outward sign to link it to. I concentrate on the sensation and "impress on myself" the connection. But next time, what makes my writing "S" correct? Whatever seems right to me will be right, and "that only means that here we can't talk about 'right'". A rule I can never be checked against is no rule. The experiment shows that meaning needs public criteria of correctness, which is why Blackburn ties the argument to closing the gap between the mental and the physically expressed.
Colour Science as Leibnizian Engineering
Can we flesh out the Leibnizian alternative, on which my twin (or myself yesterday) functioning the same way physically cannot have a different mental life? Blackburn sketches an answer from colour science.
- Three cone types. The central retina is packed with cones of three kinds, L, M and S (long, medium, short), each "spiking" more readily for light of its wavelength band.
- Opponent coding. Perceived colour depends on comparisons between cone excitations. S much more excited than L codes blue; L much more than S codes yellow; L more than M codes red; M more than L codes green. The channels are opponents: the outcome depends on which side wins.
- No yellowy blue, no reddish green. Yellow and blue are mathematical opposites (L > S versus S > L); likewise red and green. But bluish green and yellowish red (orange) are fine.
- No bright brown. Brown is darkened yellow: a surface coded yellow but with low overall energy relative to other light sources in view. (Hence no brown flame.)
- No bright grey. Grey is darkened white (hence no grey flame).
- Yellow is lighter than violet. Yellow light lies near the frequency of maximal visual sensitivity; both ends of the spectrum (red and blue) lead towards the dark, where we cannot respond at all.
- No transparent white. A surface is seen as white only when it scatters light, so the nearest to a white gem is milky opal, while transparent red, blue and green gems exist.
- White light, but no black light. Black is the absence of the response.
What looked like brute facts about the Cartesian theatre of mind begin to look like expressions of physical facts. This "only scratches the surface" but gives "a glimmering" of how things make sense.
A simplified schema. In current models the yellow-blue channel compares S cone signals against the sum of L and M, not L against S alone, and the red-green channel compares L and M; the cortical story is more complex still. Blackburn's version is a fair teaching simplification. The argument from asymmetries of colour space against inverted spectra is associated with C. L. Hardin (Color for Philosophers, 1988). Two caveats: a few laboratory reports (Crane and Piantanida, 1983) claimed observers saw "reddish green" and "yellowish blue" under special image-stabilised conditions, a finding still disputed; and critics reply that asymmetries rule out only simple inversions, leaving open more complex mappings that preserve the whole structure.
Inverting Light and Dark
Start with monochrome vision. Could a physical duplicate see as dark what I see as light, like a photographic negative? The snap answer is yes, but it fails:
- If I make a piece of grey glass lighter, I see better through it; darker, worse.
- My twin is a physical duplicate, so the same holds for him.
- For him, though, clearing the glass would seem like adding soot (it gets subjectively darker), and sooting it would seem like clearing it.
- So he would see better and better as the glass looked darker, and worse as it looked lighter.
- That does not make sense; so the light-dark inversion is not a coherent possibility.
Inverting Yellow and Blue
Now a twin who sees yellow as I see blue and vice versa. He must respond as I do, so he cannot say yellow is a dark colour (that would be a physical difference). So how does he see blue as bright and yellow as dark? And how does he see brown, which is darkened yellow, when yellow is already dark for him? Or orange? It becomes very hard to see how his physical discriminations could match mine given so complete a mental disparity.
The inverted spectrum becomes much less clear, and we may feel our way to denying it is a possibility at all. That would be conceptual engineering that closes the gap between a functioning visual system and supposedly superadded qualia, vindicating Leibniz: subjective colour becomes the inevitable, rationally explicable expression of how creatures like us function. If the same could be done for every element of consciousness, the problem is solved.
The colour-blind engineer. Imagine a vision scientist who has never seen colour, a version of Frank Jackson's "Mary" (1982), who knows everything physical about colour vision from inside a black-and-white room. Blackburn's strategy predicts that she could derive a great deal about colour experience from the structure of the opponent channels: that there is no reddish green, that brown is a dark yellow, that yellow is the lightest hue. Jackson's intuition is that, released, she still learns something new: what red looks like. The two thought experiments pull apart the structure of experience (which looks derivable from function) from its intrinsic feel (which looks left over). How much of Mary's surprise survives once all the structure is known is one way to measure how far Leibniz's programme can go.
Thought and Intentionality
Consciousness is not only sensations and qualia; it is largely thoughts, and thoughts are strange. They have representational power: a thought represents the world as being some way. A sensation seems to just sit there, pointing at nothing beyond itself. (Some deny this, holding that pain is a perception of bodily injury, representing the body as injured; Blackburn leaves the reader to ponder it.) This intentionality or directedness is deeply puzzling if thoughts are "things" in consciousness: how can a thing in and of itself point at another thing?
- The signpost. A signpost points to a village only because of how it is taken; we learn to read it. Another culture could treat the same object as a display board, a totem or abstract art.
- The dog. When you point at something, dogs typically look only at the finger, "to their owners' irritation".
- The thought. By contrast, a creature with our thoughts but without our "take" on them seems incoherent: "It is the 'take' that makes the thought."
The dog example is factually wrong. Domestic dogs are unusually good at following human pointing, better than chimpanzees in comparative studies (Hare, Brown, Williamson and Tomasello, Science, 2002), and even young puppies show the skill. Blackburn's point about signs needing interpretation survives with a different example (a cat, or a pointing gesture shown to a culture that lacks it), but the dog is a poor illustration.
Against Reification
The right response is probably to deny that thoughts are things. Supposing that every noun names a thing is the mistake of reification, a charge philosophers often trade. It is people who think, and their thinking is not the presence of a blob in brain or mind, even a blob conceived as a little sentence written in the brain. Thinking is taking the world to be one way or another, and such taking is a matter of dispositions, not of what hangs about inside us.
Perhaps thinking about absent, distant, past and future states of affairs should be no more puzzling than attending to the world at all. Yet it seems to set us apart from other animals, which presumably perceive but which we are nervous about crediting with representing absent states.
Animals and absent states. The nervousness has been eroded by experiments. Nicola Clayton and Anthony Dickinson (Nature, 1998) showed that scrub jays remember what food they cached, where and when, retrieving perishable worms before durable nuts: "episodic-like" memory. Later work showed jays caching for tomorrow's needs (Raby and colleagues, 2007), and great apes saving tools for future use (Mulcahy and Call, 2006). The gap between human and animal thought about absent states is now generally seen as one of degree and format, not presence or absence.
Interpreting the Thinker
The most popular approach attributes thoughts to the well-functioning person on the basis of behaviour, linguistic and non-linguistic. The hoped-for reduction:
- The schema. X thinks that p if and only if X's plans, desires or behaviour are "somehow in line" with the world being such that p.
- The guided missile. A missile "thinks" there is a plane a mile away and two hundred feet up if its systems point it as appropriate to a plane being there, given its aim or function of bringing down planes.
- The person. She thinks the weekend weather will be fine if her behaviour is appropriate, given her aims, to that being the weather.
- The obstacle. The difficulty is filling out "somehow in line" without relying on other mental states of the subject, and "nobody knows how to do" that.
| Position | What a mental state is | Zombies/Mutants possible? | Main problem |
|---|---|---|---|
| Substance dualism (Descartes) | A modification of immaterial substance | Yes, wide open | Other minds unknowable; interaction; why it evolved |
| Locke's annexation / occasionalism (Locke, Malebranche) | A state God attaches to physical events | Yes | Mind-body link a brute fact |
| Epiphenomenalism | A real but causally idle by-product | Yes | Can't explain evolution, memory, or why pain makes us wince |
| New mysterianism | Unknown; the link is beyond our grasp | Unknowable | A counsel of despair |
| Logical behaviourism | Dispositions to behaviour | No | Same pain, different behaviour; qualia like coffee taste |
| Functionalism | A role: typical causes, effects, links to other states | No | Qualia seem left out; analyses never completed |
| Identity theory | A brain state, found a posteriori (like temperature) | No | Multiple realisability (silicon, shrimp); "what's left out?" |
| Illusionism (post-1999) | A representation of states as having qualia | No | Seems to deny the obvious |
The chapter ends by setting thought aside. Two further elements of our world view nourish Cartesian dualism: thoughts about our freedom (Chapter 3) and about our identity (Chapter 4).
Other minds meet large language models. Blackburn's question "how do I know about your mind?" has acquired a new test case. LLMs produce fluent, apparently reflective talk about experience with no shared biology to ground an analogy, and their outputs are trained on human testimony about experience, so behavioural evidence is unusually uninformative: the system would say "ouch" whether or not anything is felt. In 2022 a Google engineer's claim that the LaMDA chatbot was sentient was widely rejected. A 2023 report by Patrick Butlin, Robert Long and colleagues assessed AI systems against "indicator properties" drawn from scientific theories (global workspace, recurrent processing, higher-order, attention schema) and concluded that no current system was a strong candidate, while finding no obvious technical barrier to building ones that satisfy the indicators. Chalmers (2023) judged current LLMs unlikely to be conscious while taking seriously that successors within a decade might be. On intentionality, Emily Bender and Alexander Koller (2020) argued that a system trained only on text cannot acquire meaning, since it lacks connection to what the words are about; others reply that grounding can be indirect. The functionalist in Blackburn's table must say the right functional organisation would suffice; the identity theorist says it would not; the Wittgensteinian asks what public criteria we would use, and whether those criteria were built for creatures whose behaviour is not itself learned from descriptions of behaviour.
Functional evidence, not testimony. For anyone deploying conversational AI in a bank, the lesson of the Zombie argument is practical: what a system says about its own states is not evidence of those states. A copilot that "explains its reasoning" produces text shaped by training, which may or may not track the computation that produced its answer. Treat self-reports like the Zombie's "ouch": verify through functional probes (counterfactual tests, consistency checks, attribution and interpretability tools) rather than taking the narration at face value. Blackburn's anti-reification point also helps: a model's "belief" is best treated as a disposition across many inputs, measured by behaviour under variation, not as a sentence stored somewhere inside.
Free Will
If every event follows from earlier causes, it seems we control nothing; if some events are uncaused, they are mere chance, and chance is as relentless as necessity. Blackburn rejects the dualist escape (a Real Me intervening in the brain) and builds instead a compatibilist account of "inside control": we are free when action flows through flexible, reason-responsive decision processes, the way a thermostat controls temperature by being part of the causal order. He then tests the account against mini-Martian puppeteers, the Twinkie defence, Strawson's worry about objectifying people, and the old fatalist arguments, and concludes that the real enemy of human dignity is a false choice between ghosts and trams.
The Bonds of Fate
The chapter opens with Lucretius, the Roman poet of Epicurean atomism, asking how free will can exist if atoms never "swerve" to "break the bonds of fate", and with a limerick (usually credited to Maurice Evan Hare) about a young man who finds he is "not even a bus, but a tram": a creature moving in predestinate grooves. Chapter 2 asked what the brain produces; this one asks about what we do. We do not only register the world; we deliberate, choose and take control.
We usually regard ourselves as free agents living in an open space of possibilities. Blackburn went to the mountains this year but could have gone to the seaside; he could not have gone to the Moon, because that was not feasible. Consciousness of freedom seems allied to consciousness itself: when we imagined Zombies we pictured jerky, robotic, inflexible creatures, and we do not take ourselves to be like that.
- Freedom grounds praise and blame. We take pride in controlling obsessions and addictions, deserve praise when we succeed, and may deserve punishment when we abuse our freedom.
- Inability excuses. Nobody deserves punishment for failing to do what they could not do: it would be unjust to punish Blackburn for not going to the Moon, or a prisoner for missing an appointment outside the prison.
- So our practices presuppose alternatives. Ordinary thought and moral reaction both assume that sometimes, even when we acted badly, we could have done otherwise.
The Argument Against Control
Lucretius and the young man can be given an argument, a thumbnail of what is now called the Consequence Argument:
- The past controls the present and future.
- You can't control the past.
- You can't control the way the past controls the present and future (the laws of nature).
- So you can't control the present and future.
- In fact, you can't control anything at all, past, present or future.
Premise 1 is determinism: every event is the upshot of antecedent causes; the state of the world at any moment evolves from its state immediately before according to unchanging laws of nature. Premise 2 looks certain. Premise 3 reminds us we cannot control the laws. The conclusion looks to follow. People who accept it are hard determinists, or incompatibilists, since they think freedom and determinism are incompatible.
Two labels run together. Blackburn treats "hard determinist" and "incompatibilist" as one camp. In the literature they come apart. Incompatibilism is only the thesis that freedom and determinism cannot both be true. Add "determinism is true" and you get hard determinism; add "we are free" and you get libertarianism (in the metaphysical sense). Peter van Inwagen gave the argument its classic modern form in An Essay on Free Will (1983), and he is an incompatibilist who believes in free will. A third option, hard incompatibilism (Derk Pereboom), says free will is incompatible with determinism and with indeterminism, which is Blackburn's dilemma below turned into a position.
The Dilemma of Determinism
Perhaps we restore freedom by denying determinism. The best current science, quantum physics, is standardly read as positing uncaused events: two systems can be in exactly the same state (no "hidden variables") and a quantum event occurs in one but not the other. Physics assigns a probability but cannot say, from the state at one moment, whether the event will happen.
But this introduces randomness, not control. If little "fits and starts" occur at a micro level in your brain, you can hardly be responsible for what follows: you can't control electron jumps, and if they are genuinely indeterministic nothing can. A random jump the wrong way is as much bad luck as an outside accident frustrating your intentions. Putting the accident into your brain does not restore your responsibility.
- Either determinism is true or it is not.
- If it is true, every action is fixed by causes beyond my control, so I am not free or responsible.
- If it is false, the undetermined events "just happen", so they are not under my control either, and I am not free or responsible.
- So either way we lose freedom and responsibility.
Quantum indeterminism is an interpretation, not a datum. Blackburn's "standardly interpreted" is fair for the textbook (Copenhagen-style) reading, but several live interpretations are deterministic at the fundamental level: Bohmian mechanics (hidden variables, made non-local, which Bell's theorem allows; it rules out only local hidden variables) and Everettian many-worlds (the wave function evolves deterministically; apparent chance reflects branching). Either way the philosophical moral survives: neither determinism nor chance by itself supplies control. Neuroscientists also doubt that single quantum events are amplified into choices; neural noise at the level of ion channels and synaptic release is real, but it is the organisation of the system, not the noise, that would have to carry agency.
Fig Trees and Waterfalls
In Mark 11:12 to 14 and 20 to 21, a hungry Jesus finds a fig tree with leaves but no fruit, "for the time of figs was not yet", and curses it; next morning it has withered from the roots. Setting aside the story's social and ecological problems, Blackburn focuses on the injustice to the fig tree.
Counsel for the fig tree. Jesus seems to argue: "You sometimes bear figs, so you could be bearing figs now." The tree has a complete defence: fig trees bear in summer, and it is winter. Bearing figs takes a set of circumstances; even the best tree does not bear out of season, any more than it bears plums. A thoughtful fig tree, remembering only that it sometimes bears figs and ignorant of the precise causes, might feel guilty for not bearing now. That guilt would be irrational ignorance. The question the story leaves hanging, and which returns later in the chapter, is whether every human failure is also "winter".
You might reason like the fig tree: I just know I am free; I stand here able to raise my arm or not; I raise it, and feel myself controlling events. The German philosopher Arthur Schopenhauer (1788 to 1860) replies with a parable.
Schopenhauer's water. A man stands in the street at six in the evening: he could go for a walk, to the club, climb the tower to see the sunset, go to the theatre, visit a friend, run out of the gate into the wide world and never return; all strictly up to him; but "with just as free a will" he goes home to his wife. This is exactly as if water said to itself: I can make high waves (yes, in a storm), rush downhill (in a river bed), plunge foaming (in a waterfall), rise into the air (in a fountain), boil away (at the right temperature); but I am doing none of these, and voluntarily remain quiet in the reflecting pond. The water is unaware of the causal set-ups it needs; it only remembers sometimes doing these things, so it attributes its calm to its own decision. If it "tries" to boil at the wrong temperature it will soon learn otherwise. Wittgenstein made the same point with an autumn leaf in the wind saying, "Now I'll go this way, now I'll go that."
Schopenhauer denies that self-consciousness displays real freedom. Blackburn reads him as attacking this argument:
- I am not conscious of the causal background needed for me to do Y.
- I know I sometimes do Y.
- So I am conscious that there is no causal background needed for me to do Y.
The argument is invalid. Being unconscious of something cannot be parlayed into being conscious of its absence. When Blackburn speaks he is unaware of the musculature, breath control, tongue, palate and jaw that make speech possible, but all are necessary, as he would quickly discover if one went wrong.
The pilot who never sees the autopilot. Imagine an aircraft whose autopilot makes thousands of small corrections, each silently routed through the pilot's control column so that it moves under his hands. The pilot feels he is flying the plane. His feeling is perfectly genuine and tells him nothing about whether the corrections originate in him. Introspection reports the felt result, never the wiring behind it. The case shows why the phenomenology of choosing cannot settle the metaphysics of choosing, in either direction.
The Real Me Behind the Machine
Next comes a tempting thought. Within the physical world we face determinism or random indeterminacy, so suppose there is another level: behind brain and body stands the Real Me, receiving information and occasionally directing operations, so that brain and body go a way they would not have gone left to themselves. This is mind-body dualism again: messages come in "perhaps through the pineal gland", and "a breath of soul then fans neurones and synapses into action", initiating new causal chains.
Beyond the interaction problem of Chapter 2, Blackburn raises a different objection: how do we understand the freedom of the soul?
- The dilemma of determinism applies to any states that change over time.
- If ghost-stuff obeys laws, so that one ghostly state fixes the next, the ghost is determined.
- If it does not, ghost-stuff is subject to random fits and starts.
- Neither yields control or responsibility, as we saw for the brain.
- So adding a ghost does not solve the problem; it relocates it.
- Mental is not the same as free. There is no correlation between an event being "mental" and being under my control: I cannot wish away pains, desires, obsessions, unwelcome thoughts and confusions.
- The homunculus problem. We secretly picture the soul's freedom on the model of human freedom: the ghost is an "ethereal little human being" that takes in information, deliberates, wants things and acts. Far from explaining human freedom, the idea depends on it.
- Causal closure. The physical system is a closed system; it takes a physical cause to produce a physical effect, so an immaterial nudge has nowhere to enter.
The fundamental mistake of throwing in another thing. The dualist sees a problem and adds a new kind of entity, forgetting to ask how the new entity escapes the problems that beset ordinary things. Blackburn flags the same pattern for the philosophy of religion in Chapter 5.
Epicurus and the Swerve
Epicurus (341 to 270 BC) had already tried to reconcile freedom with a world of small, hard, indivisible atoms by letting the spirit make atoms "swerve"; Lucretius speaks of a minute swerving "neither place nor time determinate". The laws of motion are not hospitable: linear momentum is conserved, so it would shatter physics as badly for the Real Me to make the Moon change direction by thinking as to make it speed up or slow down.
- The atomists were better off than Descartes in one respect. They held that spirit is itself mechanical, made of especially fine, small, mobile particles ("of seeds extremely small, through veins, flesh, sinews, woven"; thin enough that "dreams of smoke and mist can move it"), so there is no in-principle barrier to its pushing the body's larger particles.
- But the swerve was never explained. Lucretius does not return to how tiny particles could break the bonds of fate. The ancients compared soul acting on body to wind acting on a ship, but the wind is itself part of the infinite chain of cause and effect, and on this model so is the soul.
Pulling Yourself Together
The hard determinist argument says nothing about which kinds of causal route lead to an action. Some routes are totally independent of what we think: once a swimmer is irreversibly trapped under water, drowning is inevitable for Einstein and for a donkey alike. Other routes run through high-level neural processes, which is just to say we often move as we do because our brains are working properly.
- Scanner. Takes in information about the situation.
- Tree producer. Generates the options for behaviour in light of what the scanner reports.
- Evaluator. Ranks the options by the concerns programmed into it, perhaps by attaching emotional indicators such as fear or joy to different paths.
- Producer. Fixes on the top-ranked option and outputs the neural signals that move muscles and limbs.
This is a "software" description of parts of the brain. When a decision is the upshot of these modules functioning normally, you chose it; it was not forced on you as drowning is forced on the trapped swimmer.
The Dog-Chucker
You come into Blackburn's room and throw his peaceable old dog out of the window. Accused, you invoke incompatibilism: childhood events beyond your control "set" your modules so that a dog-free environment has top priority; your scanner reported a dog and a window, your tree producer listed the option, your evaluator selected it, your producer acted. Why blame me?
Blackburn is not impressed. He does not care how you came to be set as you are; what bothers him is that this is your set. You are "a nasty piece of work" and he will thump you. Perhaps it was bad luck that you became like this, and now doubly bad luck, since you will be thumped for it. And by your own argument you cannot blame him for thumping you: that is just how he is set.
- Resetting the evaluator. Thumping may lead your evaluator to rank dog-throwing below tolerating the dog, perhaps by attaching a "risk-of-being-thumped" flag to the option, or by shocking you into general re-evaluation.
- Deterrence. Even if you do not change, it signals to other would-be dog-chuckers.
- Expression. It also relieves Blackburn's feelings.
This differs from blaming a swimmer for drowning. There the route runs through basic animal physiology that education and the attitudes of others cannot alter; praise and blame cannot "reset" it. Dog-throwers can be deterred, changed and warned away.
The stupid pupil and the lazy one. Teachers say: "I don't mind a stupid pupil, but I do dislike a lazy one." Under the grip of hard determinism this looks like prejudice, since both traits are tough luck. But the asymmetry is justified if laziness responds to incentives and stupidity does not: respect for the teacher can make you work harder but cannot make you smarter. The teacher is in the business of resetting your evaluating module, and how far modules are reset by others' reactions is an empirical fact learned from experience.
Compatibilism and Its Labels
This is the beginning of compatibilism: the attempt to show that, properly understood, determinism is consistent with holding people responsible. It is sometimes called "soft" determinism, a poor label for two reasons.
- It is not a different determinism. It accepts determinism in exactly the same sense as anyone; no ghostly power interferes with the causal order.
- It is not soft in practice. Morally the compatibilist may be pretty harsh: offered the heartrending excuse that biology or environment made you this way, "she turns deaf"; no facile equation of crime with illness; people can pull their socks up, and she will use punishment to help them.
The compatibilist still accepts some excuses: constraint, where no upshot could be good however well your modules work (the drowning swimmer); and actions "out of character", for example under medication that disorients or depresses you, forgivable once you are yourself again.
A subject acted freely if she could have done otherwise in the right sense. She could have done otherwise in this sense provided she would have done otherwise if she had chosen differently. The compatibilist says this conditional "could" is all that is needed to justify holding people responsible.
Interventionist Control Versus Inside Control
Interventionist control. The ghost is distinct from the causal order of nature yet mysteriously able to alter it: "contra-causal" freedom. In the literature this is libertarian freedom, a confusing name, since it has nothing to do with the political libertarianism of free markets and minimal government. Blackburn sticks to "interventionist control".
Inside control. You are entirely situated inside the causal order; your freedom lies in how action flows out of your cognitive processes. You control things by being part of the way the past controls the future. When you exercise inside control badly, you can justly be held responsible, and blamed if blame is the fitting reaction.
The compatibilist answers the opening argument with a parody:
- The past controls the present and future.
- A thermostat cannot control the past.
- A thermostat cannot control the way the past controls the present and future.
- So a thermostat cannot control the future.
Something must be wrong, because a thermostat controls the future temperature; that is what thermostats do. It does so by being part of the way the past controls the future. The same form of argument therefore cannot show that we control nothing.
But we attribute no freedom to the thermostat, and compatibilism can look like a dismissal of the problem rather than a solution. So it seemed to Immanuel Kant (1724 to 1804), who called it the "freedom of clockwork" (more precisely, the freedom of a turnspit) and "a wretched subterfuge".
Kant was not simply a hard determinist. Kant accepted that the empirical world, the world of appearances, is thoroughly deterministic, and he scorned compatibilism; but he held that as rational agents we must act "under the idea of freedom" and that our freedom belongs to us as things in themselves, outside time (transcendental freedom). Blackburn uses Kant only as the voice of dissatisfaction. Kant's own solution is a version of the two-standpoint view discussed later in this chapter, but with a metaphysical self behind it that Blackburn would reject.
Puppets and Martians
Kant's worry can be sharpened. The modules complicate the causal picture, but do they alter it fundamentally? Recall counsel for the fig tree. If I acted badly, perhaps it was winter for me too: my modules had been badly set by causal chains stretching back before my birth. Your anger may alter my system for the future, but that does not show I could have acted differently in the past.
As we learn the causes behind actions, we switch into less moralistic modes:
- Depression. We might blame someone for being depressed all the time until we learn a chemical story explaining it.
- Illness. We might be angry with someone who cannot stir himself until we learn he has mononucleosis.
- Social pressure. Parents might blame a teenage daughter for squandering time and income on valueless cosmetics, but a better reaction is to understand the social and commercial pressures that paralyse her better judgement.
The determinist says there are always things like this to learn, quite apart from growing neurophysiological evidence of "brainwashing" or "conditioning".
The invasion of the mini-Martians. Tiny, organised, mischievous beings invade our brains and walk around in them, setting our modules at will; we become puppets. They might set us to do what we would have done anyway, or throw chemical switches so that we do terrible things. Science then invents a scan that detects Martian invasion. Surely we would sympathise with anyone found infested and recognise at once that he was not responsible. The incompatibilist then asks: why does it matter whether the setter was a mini-Martian or causal agencies of a more natural kind?
The ant and the fluke. In case the Martians sound far-fetched, Blackburn points to a real parasite that colonises ants' brains: under its influence the ant climbs a blade of grass, where it is more likely to be eaten by a passing sheep, which the parasite then infects. The individual parasite in the ant's brain perishes, but others hitch-hike in the ant's body. "For all one knows, the ant feels free as air." The parasite is the lancet liver fluke, Dicrocoelium dendriticum; infected ants clamp onto grass tips in the cool of evening and release in the heat of the day.
The objector's target is the compatibilist "could have done otherwise". It is all very well that she would have done otherwise had she chosen differently; but suppose she was set so that she could not have chosen differently, her choosing modules locked by Martians or chemicals. The compatibilist so far shrugs, interested only in whether the outcome is good or bad. The objector finds the history important, and some of our own reactions agree with the objector.
The manipulation argument. The mini-Martians anticipate the most discussed incompatibilist argument of the last 25 years. Derk Pereboom's four-case argument (Living Without Free Will, 2001) describes Professor Plum killing White in four scenarios: manipulated moment by moment by neuroscientists; programmed at the start of life; conditioned by his community in childhood; and produced by ordinary deterministic causes. In each, Plum meets the compatibilist conditions (he reasons, reflects, acts on desires he endorses), and there seems no principled difference between adjacent cases, so if he is not responsible in case 1 he is not in case 4. Compatibilists reply either with a "hard-line" response (Plum is responsible even in case 1, if all the compatibilist conditions really hold; Michael McKenna) or with historical conditions on how one's values were acquired (Alfred Mele).
Obsessions and Twinkies
Blackburn thinks the best compatibilist reply is to query the word "set". The first move distinguished decision-making from drowning by introducing flexibility into the causal process: modules capable of being tuned. The reply now introduces another level of flexibility. In the brainwashed teenager or the Martian case the modules really are set: subjects become inflexible, immune to argument, unresponsive to changes in the decision scenario. Normally agents are not so set in their ways: their freedom consists in responsiveness to new information and new differences in the situation. They are not driven to chuck dogs out of windows or stand all day at the cosmetics counter.
The idea sharpens into three successive definitions.
| Definition | Free if... | Handles | Still leaves open |
|---|---|---|---|
| First compatibilist | She would have done otherwise had she chosen differently. | Drowning: no choice could have saved her. | Agents whose choice itself was locked (Martians, brainwashing). |
| Revised | Also, under the impact of other thoughts or considerations she would have chosen differently. | Martians and brainwashing: locked modules ignore new considerations. | Considerations she could never have had (the arsenic case). |
| Revised revised | Those thoughts are true and available: they represent her situation accurately and are ones she could reasonably be expected to take into account. | The poisoned coffee; the Twinkie defence. | Negligence; and the incompatibilist claim that "available" must mean freely selected. |
- Spinoza's contribution. Baruch Spinoza (1632 to 1677) associated freedom with increased knowledge and understanding: we are free in so far as we understand things. This ties freedom of the will to political freedoms of information and speech, since lack of information denies us opportunities. The "true and available" clause builds Spinoza's thought into the definition.
- The arsenic coffee. Blackburn cunningly puts arsenic in your coffee; you drink it. You would have avoided it had you chosen differently, and the thought that it might be poisoned would have made you choose differently, but since there was no reason for that thought to enter your mind, you were a victim rather than a free agent.
- Bad luck, and its consolation. It may be bad luck that the right thoughts did not arise. The compatibilist agrees, and adds once more that anger and thumping may stop it recurring.
Negligence and the Golfer
What of the person to whom the thoughts just didn't occur? So far free choice, interventionist or compatibilist, was treated as necessary for responsibility. But sometimes we do not treat a missing thought as "mere" bad luck: the thought should have arisen. Someone setting fire to buildings for fun cannot seriously plead that "it never occurred to him" someone might be hurt, unless he is a child or mentally deficient. Recklessness and negligence are faults, and we are responsible for them even though there was no free choice to put people at risk. Aristotle "rather desperately" held that negligent people chose to make themselves negligent, perhaps in early childhood, and that only this makes them responsible.
The golfer, the seagull and the breeze. On day one a golfer hits a fine ball, but a passing seagull gets in its way. On day two he hits an equally fine ball, but a little breeze blows it off course. Both are bad luck, but of different kinds. The seagull is pure bad luck, an act of God. The breeze is bad luck a really good golfer is expected to foresee and play around: it lies within the player's purview. Enough of the second kind and we think less of the golfer, and the same holds for agents generally. Hence the pianist's reply to an admirer who gushed about his luck in having so much talent: "Yes, and the more I practice the luckier I get."
Moral luck became a named problem in 1976. The golfer case belongs to a debate opened by Bernard Williams and Thomas Nagel in twin papers both titled "Moral Luck" (1976). Nagel distinguished resultant luck (how things turn out: two equally careless drivers, one of whom hits a child), circumstantial luck (the situations one faces), constitutive luck (the kind of person one is) and causal luck (how one is determined by antecedent circumstances, which is the free will problem itself). Blackburn's distinction between incidental bad luck and bad luck that "casts a reflection on the agent" is a version of the line between luck that should and should not affect our assessment. Later work (George Sher, Angela Smith) argues that responsibility for unwitting omissions like negligence does not require prior choice, contra Aristotle.
Conceptual Engineering and Grey Areas
Blackburn describes this work as conceptual engineering: making explicit real elements in our thinking. We do distinguish changing the past (cannot do) from acting otherwise than we do (sometimes can); we have discriminating practices of blame; we distinguish being ill from being bad; we allow some excuses and disallow others. Philosophy aims to show all this as the application of a reasonable, defensible set of concepts, not "an irrational jumble of disconnected habits", and that is why it is hard.
- Blackburn's verdict. He believes the revised revised definition does the job "pretty well", answering to how we think when we are best in control of the problems we face.
- The other side. Others take Kant's objection more seriously: our "interpersonal reactions" depend on a lingering affection for interventionist freedom, so if that is metaphysically bankrupt our attitudes ought to change. The problem would then be that interventionist control is untenable and inside control inadequate.
- Grey areas are not failures. Return to the cosmetics-buying teenager. Whether she could do otherwise now hinges on whether the thought that cosmetics do little for her popularity (a decent mind would do more, "perhaps by reading a book like this") is "available" to her. Perhaps people under her influences just cannot get themselves to believe it; the culture is awfully good at blinding teenagers. Blackburn inclines to see her as a victim rather than an agent, but the point is that the analysis pinpoints the issue even when it does not settle it.
- The road still to travel. An incompatibilist can insist that thoughts are "available" only if they are themselves objects of free interventionist selection, which would put us back to square one.
The Twinkie defence (San Francisco, 1978). On 27 November 1978 a former city employee, Dan White, entered San Francisco City Hall with a gun, evading the metal detectors through a basement window, and shot dead Mayor George Moscone and Supervisor Harvey Milk. Defence psychiatrist Martin Blinder testified that White had been depressed, leading him to binge on high-sugar junk food such as Twinkies, which as an ex-athlete he knew was bad for him, deepening the depression; so he could not have acted with the premeditation or real intent required for first-degree murder. The jury convicted him of voluntary manslaughter instead. California later closed the space for such defences. Blackburn's verdict: White plainly acted with intention and premeditation (that is why he brought a gun and used the basement window), and the revised revised analysis is "not at all hospitable": sugar does not make us incapable of the thought that murder is a bad idea.
The Twinkie story is partly a myth, as Blackburn half-repeats it. Blinder's testimony presented the junk food mainly as a symptom of depression, not a cause of the killings, and the defence's legal argument was "diminished capacity" grounded in depression; the "Twinkie defence" label was coined by the press. White served about five years and took his own life in 1985. California abolished the diminished capacity defence in the early 1980s, by statute and by the 1982 ballot measure Proposition 8. Blackburn adds, as a jab at juries, an anecdote about a Michigan man who won damages claiming a rear-end car collision made him homosexual; he gives no details, and it should be read as an illustrative aside rather than a documented case.
Fail-Safes and Overdetermination
One difficulty faces all the definitions: they generate control from what the agent would have done in other circumstances, and causal overdetermination cases pull these apart. A thermostat might control the temperature even though a fail-safe would have kept it the same had the thermostat failed. Likewise an agent might do something bad, in control, with intent and responsibility, even though, had he chosen otherwise, unknown mechanisms would have clicked in to ensure the bad act anyway. Picture the mini-Martians sitting idle, ready to intervene only if the outcome looks set to be one they don't want. The cases are "surprisingly tricky", but they make it no harder to define control for people than for thermostats, and since the mechanical problem must have a solution, so must the human one.
These are "Frankfurt cases". The idle Martians are a version of Harry Frankfurt's (1929 to 2023) counterexample in "Alternate Possibilities and Moral Responsibility" (1969). Black, a neurosurgeon, will intervene if Jones shows signs of deciding not to kill Smith; Jones decides to kill on his own, so Black never acts. Jones could not have done otherwise, yet seems fully responsible. Frankfurt concluded that the Principle of Alternate Possibilities (responsibility requires the ability to do otherwise) is false, which turns Blackburn's worry around: if Frankfurt is right, compatibilists need not analyse "could have done otherwise" at all, only the actual sequence that produced the act. Libertarians reply with the "flicker of freedom" (Jones could at least have shown the sign) and the dilemma defence (Robert Kane, David Widerker: if the sign determines the choice, the case begs the question against indeterminism). John Martin Fischer and Mark Ravizza (Responsibility and Control, 1998) built on Frankfurt a theory of "guidance control": the action must issue from the agent's own, moderately reasons-responsive mechanism, which is close to Blackburn's revised revised definition.
Objectifying People
A further worry: the compatibilist story describes organic beings with brains in terms of what happens, not in terms of agency, of my doing things. So it seems to leave out something essential to my humanity and my regard for others: that we are active agents, not passive patients and victims. That is how we regard ourselves and others, and how we want to be regarded.
Peter Strawson (1919 to 2006), in a famous paper, contrasts an "objective" or impersonal attitude with a "personal" or participant attitude. On the objectifying track, other people are blocks to our progress, to be "managed or handled or cured or trained", looked at as if mad rather than as intelligent agents who can be understood.
Strawson's gestalt switch. At first the attitudes of blame look hard and harsh, and replacing them with "understanding" attitudes, treating people as patients rather than criminals, looks humane. Strawson asks what this loses: much of what makes human relationships distinctively human. If I explain my behaviour and see in my listeners' eyes that my talk is just another symptom, I have been dehumanised. I want my decision understood, not patronised; I want people to "hear my voice", to see how things appear to me, rather than wonder what causes this organism to behave so. (Blackburn returns to this against the therapy industry in Chapter 8.)
The compatibilist reply: the account does not deny agency; it gives an account of it. Data are taken in, alternatives generated and ranked, until an output comes on line. True, these events "just happen", but they are what happens, and all that happens, when you, the person, do something. Describing you as acting for a reason is a personal-level description of the upshot of multiple micro-level happenings.
Two Stances, No Conflict
Some thinkers speak of two perspectives: the deliberative, first-person stance you adopt in making a choice, and the "objective" third-person stance of a scientist who sees a complex, determined neurophysiological system. The problem is to reconcile them.
- Reconciliation is a problem only if what the deliberative stance discloses is incompatible with what the third-person stance discloses.
- The deliberative stance discloses nothing about causation. To think otherwise is Schopenhauer's water's error: mistaking absence of awareness of the brain's functioning for awareness of its absence. The first is universal; the second is impossible, since without the functioning there could be no awareness.
- So nothing seen from the deliberative stance conflicts with the scientific world-view.
- So there is no deep problem of reconciliation, only a moral problem: approaching one another with full human understanding, enriched rather than undermined by knowledge of what brings about others' decisions.
Strawson's paper and its legacy. The paper is "Freedom and Resentment" (1962). Strawson's central argument is that the reactive attitudes (resentment, indignation, gratitude, love, hurt feelings) are too deeply woven into human life to be given up wholesale on the strength of a general theoretical thesis like determinism; excuses work inside the practice (he didn't mean it, he was pushed), not by overturning it. This "Strawsonian" approach is now a major family of theories (R. Jay Wallace, Susan Wolf, Michael McKenna's conversational model). Blackburn's own projectivist ethics, which grounds moral claims in attitudes, makes him a natural ally. Note that Strawson died in 2006; Blackburn's 1999 text gives his dates as "1919 to" because he was then alive.
Fate, Oracles and Death
An old man Blackburn knew had been an officer in the First World War. His problem was getting men to wear their helmets under fire. Their argument: if a bullet "has your number on it", precautions are pointless; if none does, you are safe for another day and need not wear the cumbersome helmet. This is the "lazy sophism". The smoker says: if I'm going to get cancer, I'll get it. If determinism is true, isn't the future already fixed in the womb of the past? Shouldn't we resign ourselves, withdraw, and sit in an orange shawl saying "Om" all day?
Death in Samarkand. A Sufi's disciple in Baghdad overhears the Angel of Death say he has calls to make in the city. Terrified, he hires the fastest horse and rides day and night to Samarkand. Death meets the teacher and asks where the disciple is; at home studying, says the Sufi. "That is surprising," says Death, "for here he is on my list. And I have to collect him tomorrow, in Samarkand, of all places."
Oedipus. In Sophocles' Oedipus Rex, King Laius of Thebes is told his son will kill his father and marry his mother. He has the baby Oedipus crippled and left to die on a hillside; a shepherd saves him and he grows up in Corinth, believing himself the king's son. Hearing rumours of his destiny, he consults the oracle at Delphi, which confirms it, so he flees away from Corinth, and at a place where three roads meet he encounters Laius. The twofold attempts to thwart destiny are exactly what make the doom unfold.
Two Replies to the Soldiers
- The epistemic difference. Oedipus knows his fate and tries to avoid it anyway; the soldiers do not know whether they will die today. So the proper reply is open: whether a bullet has your number on it may depend on whether you wear a helmet. A bullet that would have had your number may be kept unwritten-on by this simple precaution, and since you don't know and would like none to have it, take the precaution.
- Doing nothing is a choice. Failing to put on a helmet, or donning the orange shawl, represents a choice; to have your choosing modules set by the lazy sophism is to be disposed to that kind of choice.
The lazy sophism, as an argument for action, has a twin:
Lazy version. 1. The future will be what it will be; its events are already in time's womb. 2. So, do nothing.
Energetic version. 1. The future will be what it will be; its events are already in time's womb. 2. So, get cracking.
The first would be better only if human actions made no difference to what unfolds, as if we watched a game behind one-way glass, players deaf and blind to us. Normally it is not like that. The event of someone eating an omelette is always preceded by someone breaking an egg; reaching the summit is always preceded by starting out; doing nothing is invariably followed by no omelette and no summit. Which events unfold depends on what we decide: that is what inside control means, for a person or a thermostat.
- As advice, fatalism is refuted. There is no conceivable reason to prefer "do nothing" to "get cracking". In the practical sphere, accepting one of these arguments is equivalent to admiring or desiring to be someone whose modules have that shape, and those who genuinely follow "do nothing" are simply good-for-nothings.
- As a debunking of choice, fatalism is empty. Fatalism usually aims to show choice is an illusion. One conception of choice is an illusion: interventionist choice, the uncaused intervention of the Real Me. But how could thoughts about the passage of time show that the operations of flexible choosing modules are unreal? That is no more plausible than saying computers, thermostats or chainsaws do not really operate.
The Stoics answered the lazy argument the same way. The ancient "lazy argument" (argos logos), reported by Cicero* in On Fate, ran: if it is fated that you recover from illness, you will recover whether or not you call a doctor, so don't call one. The Stoic *Chrysippus (c. 279 to c. 206 BC) replied that some events are "co-fated" (confatalia): recovery and calling the doctor are fated together, just as Blackburn's omelette and broken egg arrive as a package. Blackburn's reply is thus over two thousand years old, which is a mark in its favour.
The View From Nowhen
Suppose the future is not known to you but is known, perhaps to God, or simply knowable. As we deliberate we picture an open future and a fixed past: arrows of possibility spreading out from now. Perhaps that is illusory. From a "God's eye view", or "the view from nowhen", time is laid out like a celluloid film, one frame per moment. We can be aware only of past frames (prophets are sometimes thought to see future ones), but there is no metaphysical asymmetry between past and future. Then trying to influence the future seems as useless as trying to influence the past, and God watches our efforts laughing. This is the implication of the Sufi story, and why the soldiers said a bullet already has your number.
But why is God laughing? Even with the timeless view:
- He sees no upshot without the doing. God does not see omelettes at one date without eggs being broken a little earlier. Knowing whether we have an omelette in one frame, he also knows whether we set about preparing it in the previous one.
- He is not like a doctor with a fatal diagnosis. A doctor may know a cancer will kill us whatever we do: there would be frames in which people behave in many ways and die of the cancer anyhow. The view from nowhen sees our doings and their upshots, but not upshots without doings.
- Samarkand reread. The story implies Death had the disciple on his list before he fled, so would have come wherever he was. But perhaps Death had him on the list only because of his flight (he ran under a bus in Samarkand, say). If Death was having a field day in Baghdad, because of plague, the flight might have been perfectly rational, though unlucky in the event; and it might have been the flight that kept him off the list.
What then of the past-future asymmetry? Even God does not see us preparing an omelette after eating one (unless greedily making seconds), so it really is useless to try to influence the past. But this leaves "a huge and intractable philosophical problem": is it a mere contingency, a pattern that might differ in other regions of space and time, that we cannot influence the past? If it is only a matter of patterns seen from the timeless view, it seems it should be. Blackburn leaves this "as an exercise (an extremely difficult one)".
Two threads since 1999. The phrase "view from nowhen" is Huw Price's (Time's Arrow and Archimedes' Point, 1996), and Price argues that the direction of causation is perspectival: agents' deliberative standpoint builds in a past-future asymmetry that the physics alone does not contain, which leaves room in principle for "retrocausal" interpretations of quantum mechanics. Physicists and philosophers of physics (David Albert, Time and Chance, 2000, and Barry Loewer) instead trace the asymmetry of influence and records to the "Past Hypothesis": the universe began in a state of very low entropy, and the thermodynamic arrow then explains why we have records of the past and can affect only the future. Neither view treats the block universe as a threat to deliberation, which agrees with Blackburn.
Fatalism as a Mood
There is no general philosophical or rational justification for fatalism. It corresponds to a mood, in which we feel out of control, spectators of our own lives. That is not always unjustified: people are sometimes largely powerless, politically or psychologically (inflexible, brainwashed, or in the grip of obsessions). When best efforts come to nothing often enough, thoughts of unfolding destiny or karma can console. But they are not appropriate when we are acting: we cannot safely think while driving that it makes no difference whether we turn the wheel or hit the brake.
The sealed envelope. A perfect predictor watches you deliberate about a job offer, writes "accepts" on a card and seals it in an envelope before you decide. You agonise for a week and accept. Did the envelope make you accept? No: the card says "accepts" because of the deliberation it foresaw; had your reasons pointed the other way, it would say "declines". Opening the envelope early would simply give you one more input for your scanner, possibly a self-defeating one, which is why reliable predictions about deliberators are so hard to publish to the deliberator. The case shows Blackburn's point in miniature: foreknowledge tracks doings; it does not replace them.
Flexibility and Dignity
Mind-body dualism runs deep as an ideology: not a specific argument but a framework of thought, a guiding idea that is supposed to make freedom, dignity and experience itself possible, underwriting the "big words that get on banners". The last two chapters tried to disconnect these things from dualism. People fear the alternative: are we reducing people to "drab monochrome machines", conditioned this way or that, or passive vehicles for our selfish genes? "Absolutely not."
- The false dichotomy. The alternatives are posed as exhaustive: either a free spirit floating apart from the natural order, or a determined machine like a bus, or even a tram. Blackburn calls this misrepresenting the alternatives and says we will meet it again.
- Who denigrates nature. It is not compatibilism that denigrates human nature but this way of posing the choice, which assumes nature is so awful it takes a divine spark from the ghost to make it sing: either clockwork (Zombies) or Ghosts. Blackburn invokes Wittgenstein: it is humiliating to appear "like an empty tube, which is simply inflated by a mind".
- Flexibility is the key word. You cannot tell a priori how flexible human behaviour is. Biology gives us the modules; how they are programmed in different environments is another matter. Compare language: biology gives us the structures for learning a language, which no other animal has to anything like the same degree, but which language we learn is fixed by environment, as infants imitate their mothers and kin.
- The evidence favours flexibility. Our awareness, our capacity to think of alternatives, our evaluations and behavioural routines might have been highly inflexible, but evidence suggests the reverse: people grow up caring about a huge variety of things, and universal patterns are hard to find. Human beings can grow to make killing fields, and they can grow to make gardens.
- Beware the gurus. Theorists like to impose a pattern: people are all selfish; driven only by class interests; hate their parents; can be conditioned; men are aggressive; women are gentle; people cannot help themselves. These are interpretations imposed on evidence, and like all stereotypes they are dangerous, because people can be caused to conform to them and become worse as a result. Conceptual engineering supplies clearer outlines of alternative structures of thought.
| Position | Determinism true? | Free will compatible with it? | Source of control | Main objection |
|---|---|---|---|---|
| Hard determinism | Yes | No | None; freedom is an illusion | Cannot explain our discriminating practice of excuses; the thermostat parody. |
| Interventionist (agent-causal) libertarianism | No | No | The self or soul as uncaused cause (Chisholm, O'Connor) | Homunculus regress; causal closure; how the ghost escapes the dilemma. |
| Event-causal libertarianism | No | No | Indeterministic deliberation in the brain (Kane's "self-forming actions") | Luck objection: chance is as relentless as necessity. |
| Classic compatibilism (Hobbes, Hume) | Either | Yes | Acting as one wills, without constraint | Locked choices: mini-Martians, brainwashing. |
| Blackburn's revised revised compatibilism | Either | Yes | Inside control responsive to true, available considerations | Manipulation arguments; "available" may smuggle in interventionism. |
| Hierarchical / reasons-responsive (Frankfurt, Fischer and Ravizza) | Either | Yes | Endorsed desires; reasons-responsive mechanism that is one's own | Manipulated agents can meet the conditions (Pereboom). |
| Strawsonian (P. F. Strawson) | Irrelevant | Yes, practice needs no metaphysics | The reactive attitudes and their internal excuses | May describe our practice without justifying it. |
| Hard incompatibilism (Pereboom, G. Strawson) | Either | No, nor with indeterminism | None for basic desert; forward-looking practices remain | Seems to make resentment and gratitude unfitting. |
What neuroscience did to the debate. Blackburn wrote before the brain-imaging era of free will. Benjamin Libet's 1983 experiment had subjects flex a wrist whenever they chose and report, from a fast clock, when they first felt the urge (W ≈ 200 ms before movement); scalp EEG showed a readiness potential beginning ≈ 550 ms before movement, roughly 350 ms before awareness. Chun Siong Soon, John-Dylan Haynes and colleagues (2008) used fMRI to predict a left/right choice up to about 7 to 10 seconds before reported awareness, though only at about 60% accuracy. Then Aaron Schurger, Jacobo Sitt and Stanislas Dehaene (2012) showed the readiness potential can be modelled as spontaneous neural noise accumulating to a threshold; on this reading the early ramp is not a hidden decision but an artefact of averaging trials aligned to movement onset. Uri Maoz and colleagues (2019) found the readiness potential absent for deliberate, meaningful choices (donations to charities) as opposed to arbitrary ones. Blackburn's framework predicts the moral: these findings threaten the interventionist picture, in which a conscious ghost must start the causal chain, but not inside control, on which decision processes were always partly unconscious. Popular hard-determinist books continue (Robert Sapolsky, Determined, 2023), as do naturalistic defences of agency (Kevin Mitchell, Free Agents, 2023; Daniel Dennett, Freedom Evolves, 2003).
Counterfactual explanations are Blackburn's definition, applied to models. In credit decisioning, regulators and model-risk teams ask what a model would have done under different inputs: "you would have been approved had your income been £4,000 higher". These counterfactual explanations (formalised by Wachter, Mittelstadt and Russell, 2017) test exactly the compatibilist "would have done otherwise if...", and the revised revised clause adds the right constraint: the alternative inputs must be true and available to the applicant, not impossible changes such as age. The same logic governs human accountability around AI. The UK Senior Managers and Certification Regime holds named individuals responsible if they failed to take "reasonable steps": the golfer's breeze, not the seagull. A model failure that should have been foreseen by monitoring is negligence; a genuinely novel shock is bad luck. Architects can make that line defensible only by recording what was known and available at decision time.
- Lucretius (c. 99 to c. 55 BC): Roman Epicurean poet; the atomic swerve as the origin of free will.
- Epicurus (341 to 270 BC): Greek atomist; the spirit makes atoms swerve.
- Chrysippus (c. 279 to c. 206 BC): Stoic; answered the lazy argument with co-fated events.
- Aristotle (384 to 322 BC): held the negligent chose to become negligent.
- Baruch Spinoza (1632 to 1677): freedom as understanding.
- Immanuel Kant (1724 to 1804): dismissed compatibilism as a "wretched subterfuge", "freedom of clockwork".
- Arthur Schopenhauer (1788 to 1860): the water parable; consciousness does not reveal freedom.
- Ludwig Wittgenstein (1889 to 1951): the falling leaf; the "empty tube" inflated by a mind.
- Peter Strawson (1919 to 2006): objective vs participant attitudes; reactive attitudes.
- Harry Frankfurt (1929 to 2023): counterexamples to alternate possibilities; second-order desires.
Frankfurt's other contribution: wanting what you want. In "Freedom of the Will and the Concept of a Person" (1971), Frankfurt distinguished first-order desires (to smoke) from second-order desires about which first-order desire should move you (to want not to want to smoke). A "wanton" has no second-order volitions; the unwilling addict has them but is moved against them; the willing addict endorses his addiction. Freedom of the will is having the will you want to have. This gives Blackburn's "pulling yourself together" a precise structure: the self that pulls is the second-order evaluator, still entirely inside nature. Susan Wolf (1987) objected with JoJo, son of a sadistic dictator, who wholeheartedly endorses his cruelty; she argued responsibility also requires the ability to recognise the good, which parallels Blackburn's "true and available considerations" and his verdict on the cosmetics-buying teenager.
The Self
Descartes salvaged the "I" from universal doubt, but when Hume looks inward he finds only perceptions, never a self. Blackburn follows Locke and Kant to show that an immaterial soul substance would explain nothing about personal identity, works through Locke's memory theory and Reid's brave officer, rejects Hume's bundle of unowned experiences, and adopts Kant's idea that the self is a structural requirement, the point of view from which experience is organised, not an object found within it. That diagnosis dissolves the imaginings that make reincarnation and disembodied survival seem possible, and leaves one stubborn residue: the feeling that our future holds three crisp options, which "scrambling" thought experiments show to be an illusion that is very hard to shake.
An Immortal Soul?
Chapter 1 ended with Descartes keeping the self alone from the wreckage of doubt, and Georg Christoph Lichtenberg (1742 to 1799) querying his right to do even that, since all he was entitled to was "it thinks". Blackburn starts from two lists of things we think about ourselves.
| Thought about myself | List | Fits "I am a large human animal"? |
|---|---|---|
| I was once very small. | 1 (actual) | Yes: animals grow. |
| Barring accident, I will become old. | 1 | Yes: natural change. |
| When old I will lose memories, want different things, and my body will change. | 1 | Yes: survival through change. |
| The organic material of my body (except my brain) changes roughly every seven years. | 1 | Yes, as with a tree's cells. |
| If I lost body parts in an accident, I would have to cope. | 1 | Yes: same animal, damaged. |
| I might have been born at another time and place. | 2 (possible) | No: the self floats free of origin. |
| I might survive bodily death and live on as a spirit. | 2 | No: outlives the animal. |
| I might have had a different body. | 2 | No. |
| I might have had a different mind. | 2 | No. |
| I might have had both a different body and a different mind. | 2 | No: nothing of the animal remains. |
| I might be the reincarnation of some historical personage. | 2 | No. |
| I might have to live again, e.g. as a dog, unless I behave well. | 2 | No. |
"Possible" here means only that we seem to understand these thoughts, not that we believe them. Some people, and whole religions, do: on Blackburn's account Christianity holds the second-list item about survival as a spirit, and Hinduism the last.
- List 1 fits a straightforward view. I am a large human animal, with a biography like other animals': natural birth, natural changes, natural death, firmly bounded in space and time.
- List 2 suggests something mysterious. Something only contingently "fastened to a dying animal" (the phrase is W. B. Yeats's, from "Sailing to Byzantium"), able to change body and mind and exist without a body at all, with a biography spanning centuries and endless changes of character, "rather like an actor".
Two claims in the lists need correction.
- The "seven years" figure is folklore. Radiocarbon dating of DNA using the Cold War "bomb pulse", pioneered by Kirsty Spalding, Jonas Frisén and colleagues (Cell, 2005) and extended in later studies, showed that turnover varies enormously by tissue: gut lining renews in days, fat cells at roughly 10% a year, while most neurons of the cerebral cortex, the lens of the eye and a large share of heart muscle cells last essentially a lifetime. At the level of molecules and atoms, by contrast, even long-lived cells exchange material constantly, so there is no single cycle at all. Blackburn's philosophical point, that we survive replacement of our material, is untouched.
- Christian orthodoxy is resurrection, not spirit survival. Mainstream Christian doctrine (the creeds' "resurrection of the body") expects bodily resurrection; disembodied survival between death and resurrection is held by many traditions but is not the whole doctrine. Hindu traditions do teach rebirth (saṃsāra), including in animal form, governed by karma.
Hume's Elusive Self
Descartes thought we had a "clear and distinct" perception that the self is distinct from the body, and list 2 seems to support him: something, my soul or essence, endures through list 1's changes and could endure through list 2's. But what is it? David Hume (1711 to 1776) reports that when he enters most intimately into what he calls himself, he always stumbles on some particular perception: heat or cold, light or shade, love or hatred, pain or pleasure. He never catches himself without a perception, and never observes anything but the perception. In sound sleep he is insensible of himself and "may truly be said not to exist"; were all perceptions removed by death, he would be "entirely annihilated".
- The self is unobservable. Look inside to catch it and you miss: you meet experiences and emotions, never the "I" that is their subject.
- Yet we know ourselves with peculiar intimacy. Descartes thought this self-knowledge survived even hyperbolic doubt.
Reid's Simple Self
Many philosophers thought this nugget of self had another remarkable property: it is simple, not composite. Thomas Reid (1710 to 1796), Hume's contemporary and the leader of Scottish "common sense" philosophy, declared "a part of a person is a manifest absurdity".
- Loss without diminution. A man who loses estate, health or strength is still the same person and has lost nothing of his personality; cut off a leg or arm and he is the person he was.
- The amputated limb test. If the limb were part of his person, it would have a right to part of his estate and be liable for part of his engagements, entitled to a share of his merit and demerit: "manifestly absurd".
- Successive thoughts, permanent self. Thoughts, actions and feelings change every moment; they have "a successive existence", but the self to which they belong is permanent and stands in the same relation to all of them.
This simple, enduring "I" is exactly what Hume could never stumble on; Reid bangs the table and announces its existence.
The Simplicity Argument for Immortality
Simplicity conveniently opens the door to a traditional argument for immortality:
- All change and decay is the coming together or falling apart of composite things.
- So anything that is not composite cannot change or decay.
- The soul is not composite.
- So the soul cannot change or decay.
- Defending premise 1. In any natural change something is conserved: break a biscuit and its matter is conserved. Chemistry was once the rearrangement of conserved atoms; now we dig deeper to energy or sub-atomic particles. Either way only compositions change; the fundamental stuff "just keeps on".
- If both premises held, the argument would be good. Premise 1 would need to be an a priori truth, premise 3 grounded in Reid.
- The deflating corollary. It equally proves that my Self existed before my birth, "which might be a bit deflating".
Before asking whether list 2's possibilities obtain, Blackburn asks whether they are even coherent.
Kant answered the simplicity argument directly. The argument goes back to Plato's Phaedo (the soul's affinity with the simple and unchanging) and was restated by Moses Mendelssohn in his own Phaedon (1767). In the second edition of the Critique of Pure Reason (1787), Kant's "Refutation of Mendelssohn's Proof of the Permanence of the Soul" grants simplicity for the sake of argument and replies that a simple being need not perish by decomposition; it could fade away by diminishing degree of intensity, its consciousness dwindling to nothing ("elanguescence"). So premise 1 is false even for simples. More broadly, Kant's Paralogisms argue that the formal unity of "I think" tells us nothing about any substance underlying it, which is the line Blackburn follows below.
Oak Trees and Ships
List 2 prises the self away from everything that seems to give it identity: body, history, memory, even mind. To test whether this makes sense, turn to the identity of other things. John Locke (1632 to 1704) observed that a plant is one plant so long as it has an organisation of parts "partaking of one common life", even though that life is communicated to new particles of matter.
- Same tree, new matter. We have the same oak through time though its atoms, cells or molecules change. What matters is organisational or functional unity: sapling and mature tree, after branches drop off, are the same oak.
- Same human being. "Same man or woman" works like "same tree" or "same monkey": it accommodates growth and change so long as there is continuity of organised life. This is how we reidentify the large mammal you see in the mirror.
Confine attention to plants and animals and list 2 gets no foothold. We do not think of an oak, "that tree might have been a maple", except to mean we could have planted a different tree; but then it would have been a different tree, not that very oak "dressed up as a maple". Nor do we imagine a tree surviving organic death and coming back as a daffodil. So if being the same self is nothing over and above being the same human being, settled as we settle animal identity, none of list 2 makes sense.
Tolerance of Change
- The Everest cloud. The cloud streaming off the summit may be the same cloud to the mountaineer for hours or days, though the wind tears water molecules through it at a hundred miles an hour.
- Teams and regiments. We support the "same team" year after year as players, management and even ground change; the glorious history of the regiment would be far less glorious if the regiment could be traced back only as far as its present members.
- Functional artefacts. It is the same computer after more memory, a new screen and a system update.
- Carelessness. The joke of the Irish axe, in the family for generations though it has had three new heads and five new handles.
The ship of Theseus, with Hobbes's twist. On a long voyage Theseus replaces, plank by plank, every sail, spar, rope and plank of his ship and tosses the old ones overboard. Does he come home in the same ship? We would probably say yes. But an entrepreneur follows behind, collects the discarded parts and reassembles them. Can't he claim the original ship? Yet two different ships cannot each be identical with the original. The puzzle is ancient (Plutarch reports it); the reassembly version is Thomas Hobbes's (De Corpore, 1655). What it shows is that continuity of form and function and continuity of matter are two criteria that normally coincide and here come apart, so "same ship" has no single answer until we decide what the concept is for.
Souls and Elastic Balls
To make list 2 intelligible, one might invoke an "immaterial substance", the simple soul of Me; list 2 might even seem to argue for Cartesian dualism, the only framework in which it makes sense. Here Locke makes "an extremely interesting move": since plants and animals survive change of material substance, why shouldn't persons survive change of soul substance?
Locke asks whether, if the substance that thinks were changed, it could be the same person, or the same substance remaining, different persons. Those who place thought in a purely material constitution plainly locate personal identity in something other than identity of substance, as animal identity lies in identity of life. So those who place thinking in an immaterial substance must show why personal identity cannot be preserved through a change of immaterial substances, as animal identity is preserved through a change of material ones.
Locke's wonderful move. Even if we are deeply worried about survival through time and change, invoking immaterial soul substances won't help. We count plants through time regardless of their material elements, and we count persons through time without any reference to immaterial substances. The soul, if there is one, does no work in our judgements of who is who.
Kant's elastic balls. In the Critique of Pure Reason, Immanuel Kant (1724 to 1804) notes that one elastic ball striking another in a straight line passes on its whole motion, and so its whole state (counting only position in space). Postulate, by analogy, substances that pass on their representations together with consciousness of them: the first transmits its state and consciousness to the second, the second its own state plus the first's to a third, and so on. The last substance would be conscious of all the earlier substances' states as its own, because they arrived together with consciousness of them. Blackburn's modern gloss: perhaps our immaterial substance is replaced every evening, like swapping the disk drive in a computer while preserving all the software and files. Nothing in experience would reveal it.
Kant continues: from our own consciousness we cannot determine whether, as souls, we are permanent. Since we count as belonging to our identical self only what we are conscious of, we must judge that we are one and the same throughout the time we are conscious of, but that judgement holds only from the inside, not from the standpoint of an outside observer. The only permanent item in the soul is the representation "I" that accompanies and connects our thoughts, and we cannot prove that this "I", a mere thought, is not in flux like the rest.
- The negative point. Nothing in our inner musings about "myself" licenses a permanent inner substance, able to survive the most remarkable changes. That puts grave doubt over the simplicity argument for immortality.
- The positive points. Locke and Kant each have something constructive to offer, taken up in the next two sections.
The Brave Officer
Locke says it is "the same consciousness that makes a man be himself to himself", and neither the subject nor onlookers care whether that consciousness is carried by one enduring substance or a succession of them. He sharpens this into a memory criterion:
Person A at a later time is the same person as person B at an earlier time only in so far as A is conscious of B's experiences: A remembers thinking what B thought, and remembers sensing, feeling and acting as B sensed, felt and acted.
Consequences we might like. - I am not Cleopatra reincarnated, since I am not conscious of anything Cleopatra did or felt: memory wipe-out destroys personal identity. - I will not live again as a dog: no dog could remember my doings; if one did (think of the neural complexity required), it would be not a dog but at best "a doggiform human being", and dogs are not doggiform human beings.
Consequences we might not like. - I cannot survive complete amnesia: whoever remains cannot be me. - Partial amnesia splits persons: if I commit a crime and, through an untoward rush of blood or adrenalin, retain no memory of it, then I am not the person who committed it, though the same human being. One human being is inhabited by multiple successive persons as memories come and go.
Locke's own prince and cobbler. Locke imagined the soul of a prince, carrying all his memories, entering the body of a cobbler whose own soul has departed. The cobbler-bodied man would be the same person as the prince, accountable for the prince's deeds, but the same man as the cobbler to everyone who saw him. The case shows why Locke needs "person" and "man" (human being) as separate sortals: they can in principle come apart, and when they do, our sense of who deserves the prince's reward or punishment follows memory, not flesh.
Reid's Objection
Reid presented the problem as the brave officer objection. A boy is flogged at school for robbing an orchard; as a young officer he takes a standard from the enemy in his first campaign; in old age he is made a general. When he took the standard he remembered the flogging; when made a general he remembered taking the standard but had absolutely lost the memory of the flogging.
- The officer remembers the flogging, so on Locke's criterion the officer = the schoolboy.
- The general remembers taking the standard, so the general = the officer.
- Identity is transitive, so the general = the schoolboy.
- The general does not remember the flogging, so on Locke's criterion the general ≠ the schoolboy.
- So the general both is and is not the same person as the schoolboy: Locke's theory yields a contradiction.
Locke's Reply: Man Versus Person
Locke saw the problem. Suppose I wholly lose the memory of parts of my life beyond retrieval: am I not still the person who did those things? He answers that the word "I" here is applied to the man only; because the same man is presumed to be the same person, "I" is easily taken to stand for the same person too. But if the same man could have "distinct incommunicable consciousness" at different times, the same man would at different times make different persons.
- Either "same person" goes along with "same human being", or it does not. If it does, we have one human being from infancy to death regardless of mental capacities, and none of list 2 makes sense.
- Locke's reason for denying it. If one man has distinct, incommunicable consciousnesses, we have different persons successively inhabiting one body (compare multiple personality disorder). Then it is "surprising, but correct" that the senile general is not the schoolboy.
Personal Identity as a Forensic Notion
Locke's reason for his view is the very thing Reid disliked: we need "same person" to justify claims of responsibility. Personal identity is "forensic", a notion whose home is in courts of law.
The amnesiac war criminal. A doddering eighty-year-old is charged with crimes committed sixty-five years ago as a naive conscripted teenager. If he genuinely retains no memory of them, it is to him as if he is condemned for a completely different person's deeds: he cannot repent of them, since they are no part of his self-consciousness and cannot weigh on his conscience. That seems unjust.
- Courts don't work that way. Locke knew amnesia is not a valid legal excuse, but thought this reflected suspicion, since amnesia is too easy to claim. In God's eyes, real amnesia really excuses: God would treat the genuine amnesiac as a different person from the war criminal.
- Degrees of forgetting. That is attractive for the old man but less so for the crime committed in a rush of blood, where we might say forgetting is neither here nor there. We may want to distinguish degrees of memory loss.
Transitivity and the Ship Registry
Reid's charge is that Locke contravenes logic, through a failure of transitivity of identity. Transitivity is the law that if A = B and B = C, then A = C. Locke has schoolboy = officer, officer = general, but not schoolboy = general.
Blackburn suggests the oddity may come from abstracting out "identity" when what we are really discussing is "is the same person as". Consider a composite such as a ship, and suppose antique ships over fifty years old are taxed less. Theseus and the entrepreneur both claim the tax break, and to stop such tiresome claims a law is passed:
A ship must be registered every year, and to count as the same ship as in any particular previous year, a vessel must contain at least 55 per cent of the material making up the ship on the first day she was registered that previous year.
Under this code Argos 1 may be the same ship as Argos 2, and Argos 2 the same ship as Argos 3, yet Argos 1 not the same ship as Argos 3. The law looks as sensible as laws fixing what must be in butter or what counts as corn-fed, and a sensible law cannot give rise to a contradiction.
Ships are composite, and that seems to generate the pattern. So perhaps Reid's argument that you cannot have A = B, B = C but not A = C goes through only if A, B and C are simple. Reid held the soul simple; Locke did not, so the argument may not count against him.
Blackburn's escape from Reid is weaker than the standard one. Most logicians would say identity is transitive for anything whatever, simple or composite. What the ship law defines is not identity but a different relation, "registry-continuous with", which is simply not transitive; the law is sensible precisely because it does not pretend to define identity. So the lesson is that Locke's direct-memory relation cannot be identity. The usual neo-Lockean repair, found in H. P. Grice (1941) and later Sydney Shoemaker and Derek Parfit, defines personal identity as psychological continuity: overlapping chains of direct memory (and other psychological) connections. Schoolboy, officer and general are linked by such a chain, so the general is the schoolboy, and transitivity is restored. A second classic objection is Joseph Butler's (1736): memory presupposes identity (I can only remember my own experiences), so it cannot constitute it; Shoemaker and Parfit answer with "quasi-memory", memory-like states caused in the right way that do not presuppose identity. (A small slip in the original text says "Argos2 the same ship as Argos2"; Argos 3 is meant.)
Forensic identity is how banks and agent platforms already think. Locke's insight that identity is a forensic notion, defined by what we need it for, fits enterprise practice exactly. A customer's identity in KYC is a chain of verified continuities (documents, accounts, behaviour), not a metaphysical essence; an AI agent's identity for audit is likewise a chain: model version, prompt and tool configuration, credentials, logs. When a model is upgraded underneath a deployed agent, the ship of Theseus question becomes practical: is it "the same system" for model risk validation, for regulatory attestations, for accountability under SM&CR? The workable answer is Blackburn's ship registry: write the continuity rule down explicitly (what changes trigger revalidation), accept that it is a non-transitive policy relation rather than deep identity, and do not let small cumulative changes silently carry an old approval forward.
The Self as Bundle
Hume found, on looking inward, individual memories, thoughts, passions and experiences, but no you. He held that if you cannot encounter something in experience you have no right to talk of it; your mind cannot "touch" it. So, consistently, the self is nothing but an aggregate of perceptions with whatever connections hold between them: "content, but no container". This is the "no ownership" or "bundle" theory. For Hume, as for Lichtenberg, we have "it thinks", or rather "thoughts go on", but no owner or possessor doing the thinking.
The self as owner (Descartes, Reid). There is a simple, enduring subject distinct from its experiences, to which every experience belongs; it is permanent while its thoughts are successive, and it is known with special intimacy.
The self as bundle (Hume). There are only the perceptions and their relations (resemblance, causation, memory); "the self" is the bundle, and talk of an owner over and above it is empty, since no such owner is ever met in experience.
Dents and Grins: The Ownership Objection
The standard objection is that the bundle theory needs the idea of an unowned experience, which is incoherent. It treats experiences as objects in their own right that might float around "like sticks lying in a forest", waiting to be scooped into a bundle. But experiences are parasitic, or adjectival, on the persons who have them.
The dent and the Cheshire cat's grin. We talk about dents: this one is worse than that, or costlier to repair than last year's. But it is logically impossible for there to be an unowned dent, a dent without a dented surface. Dents are "the shadows of adjectives": first a surface is dented, then we abstract a noun. "Dent" is logically downwind of "dented". A grin likewise is downwind of a grinning face, which is the joke of Lewis Carroll's Cheshire cat, which vanished leaving only its grin. So with pain: the onset of a pain is a bit of the person beginning to hurt, and it could not be accidental which bundle that very pain attached to. The thought experiment shows that "experience" is a noun built on a verb, and the bundle theory mistakes grammar for ontology.
Kant makes the point with the "I think" that accompanies all my representations: my experiences come billed as "mine". I do not first meet an experience, look around for its owner, and then (if, against Hume, the search succeeds) announce that it is mine. To feel a pain is in and of itself to be aware that I am in pain.
Wittgenstein: No Question of Recognising a Person
But how can this be if Hume is right that we are never aware of a self? When I see a dent I see a dented surface; we are at least aware of surfaces. If Hume is right, we are not aware of a soul or self at all. Perhaps the way forward is to deny that the self is the kind of thing of which awareness is possible.
- Subject uses of "I". Ludwig Wittgenstein (1889 to 1951) notes that in "I hear the rain" or "I have a toothache" there is "no question of recognizing a person". It is as impossible to mistake another for myself in saying "I have a toothache" as "to moan with pain by mistake, having mistaken someone else for me". You cannot misidentify the subject as yourself.
- The illusion this creates. Because we don't identify ourselves by bodily characteristics in such cases, we get the illusion that "I" refers to something bodiless seated in the body: the "real ego" of Cogito ergo sum.
- Mind is not abolished. Asked "Is there then no mind, but only a body?", Wittgenstein answers that "mind" has a use in our language, but saying so doesn't yet say what kind of use we make of it. We need to think of self-consciousness some other way.
Names and refinements since. Wittgenstein's remark comes from the Blue Book (dictated 1933 to 1934). Sydney Shoemaker (1968) named the phenomenon "immunity to error through misidentification relative to the first-person pronoun": I can be wrong that someone has a toothache, but if I know someone has it in the first-person way, I cannot be wrong that it is I. Later work found the limits: thought-insertion in schizophrenia, where patients experience thoughts in their own minds as someone else's, challenges whether ownership is always given with experience. On the bundle theory, Peter Strawson's Individuals (1959) gave the classic version of the ownership objection Blackburn uses, and Hume himself confessed in the Appendix (1740) to the Treatise that his account left him in "a labyrinth", unable to explain what unites the perceptions. The bundle view has a long parallel in Buddhist anattā (no-self), notably Nāgasena's chariot in the Milindapañha: "chariot" names an arrangement of parts, and "Nāgasena" names a stream of physical and mental events. Today Thomas Metzinger (Being No One, 2003) argues there are no selves, only transparent self-models the brain cannot recognise as models.
The Self as an Organising Principle
Blackburn now approaches the self through artificial intelligence.
The robot in the room. A robot with a video camera motors around a room of objects; we want it to output a description of how the objects are arranged. Point the camera at an object and pixels fire up: it has the kind of "inner glow" sometimes linked with consciousness. But that alone gives only Kant's "rhapsody of perceptions", or what William James (1842 to 1910) called "a blooming buzzing confusion". The robot must organise its data. A round shape on screen: a small round object nearby, a large one far away, or an elliptical object seen slantways? To decide, it moves and gets a new picture, but then it has to synthesise pictures taken at different times into a three-dimensional representation. What abilities does synthesis require?
- Self-motion detection. A way of telling whether it is itself moving, and in particular whether it is moving among stationary objects or still while objects move.
- Memory. A record of what the scene was like, to compare with what it is like now.
- Temporal order. A representation of the order of different appearances.
- Integration. A way of combining past scenes with the present one.
- Solving for its own point of view. To solve for the positions of objects, it must solve for its own viewpoint and for elapsed time, logging its own movements.
- The self is not another input. A programmer cannot solve the problem by giving the robot one more ingredient on screen, say a glimpse of one of its wheels at the bottom of every frame. That is just more input, not part of the programming that turns input into a description of the room and the robot's place in it.
- The robot need never see itself. The camera can be rigidly fixed forward. So Hume would have been no nearer catching himself even if, whenever he looked inward, he met a continuing background drone.
- What it needs is a route and a centre. It must track its own path through space and the time order of appearances; it must have an "egocentric" point of view, presenting space as centred on itself. Then, interpreting a table three feet away, it can say "the table is three feet from me", with no acquaintance with its shape or history, and "most certainly" no acquaintance with an internal ego or immortal soul.
- Chaos defeats it. Put the robot in a "Keystone Cops" environment where objects come and go at random or with amazing rapidity, and the problem becomes insoluble: random pixels, too little continuity for any program to get a grip.
The self as structural requirement. Thinking in terms of an "I" is a formal requirement on interpreting experience as experience of a three-dimensional world of continuing objects among which we move; the "I" is the point of view from which interpretation starts, not something else given in experience. Nothing given in experience could solve the formal problem for which an "I" is needed, but a point of view is always needed: to represent a scene is to represent yourself as experiencing it one way or another.
Robotics built Blackburn's robot. The problem he describes is now a standard engineering task, SLAM (simultaneous localisation and mapping): a robot or self-driving car estimates the map and its own pose together, because neither can be solved without the other, exactly the Kantian point. Neuroscience has a matching mechanism: when the brain issues a motor command it sends an internal copy (efference copy or corollary discharge, described by Erich von Holst and Horst Mittelstaedt and by Roger Sperry in 1950) so that self-caused changes in sensory input can be subtracted; this is why the world does not seem to jump when you move your eyes, but does if you press the eyeball with a finger. Predictive-processing theorists (Anil Seth, Andy Clark) extend the idea to a layered self-model, from bodily regulation to perspective to narrative. Large language models offer a contrast case: a chat model's fluent "I" is a learned linguistic convention, not the output of solving for its own position in a world, though embodied and agentic systems that must track their own actions over time begin to need the structural self Blackburn describes.
Delusions of Imagination
The line of thought is Kant's, "one of the great moves in philosophy", and here it yields a diagnosis of list 2. Those thoughts arise because I can imagine myself in different shoes: historical characters, dogs, angels. I then think I must have transported my very soul into the imagined scene. Since I may imagine myself at another time, with another body, other mental properties and other experiences, the soul becomes very strange: I abstract from everything that gives me my identity as a human being and still suppose an essence of Me remains. Hence Descartes's "real distinction".
Suppose instead that I am not transporting anything: I am only representing what it would be like to see the world from a different point of view. Then my ability to imagine the scenes is no evidence that "I" might have experienced them or might survive to experience them.
Watching my own funeral. I imagine my coffin and my mourning family. Perhaps I am skulking at the back of the church, miffed that the congregation does not seem all that upset, wishing I could tell them it is not so bad. Perhaps, being dead, I have X-ray vision and glimpse my body in the coffin: how old I look. But wait: the pearly gates, and grandmother waiting to greet me. All of this I can do; I understand what it would be like to see a coffin, even to glimpse inside it. But these exercises do not transport a "me" doing the seeing while the human being Simon Blackburn is dead. It is I here and now who am imagining; there is no I being imagined doing the viewing. The only relic of me in the scenario is the dead body.
The point in Kantian form:
- "I can imagine seeing X" is equivalent to "I can imagine myself seeing X".
- The equivalence is purely formal: the "myself" adds only the point of view, which every imagining has.
- So no substantive self, no soul of Me, is involved in either imagining.
- So such imaginings give no support to a "real distinction" between you as subject and the animal you are, and no support to your biography outrunning that animal's just because you can imagine X, which the animal will not see.
Imagining being Genghis Khan. I picture riders and battlefields. I am short, crafty, a wonderful horseman. God, the steppes are cold. The politics gets me down. "Another helping of fermented mare's milk," I call. Whoops: I am supposed to speak Mongolian, not English. Here it is obvious that no soul of Me has been transported; in so far as anything of me remains, like the lapse into English, the imagining is a failure, like an actor playing Henry VIII who looks at his watch or chats about what's on at the cinema. What I really do is visualise the steppes and battles as if seeing them, more or less well, as some people are better actors than others.
- Third-person imagining exposes the trick. Try to suppose that you were once Genghis Khan: slaughtering people from a horse, unaware of supermarkets, cars and aeroplanes, of different gender, age and mind. All I manage is to substitute thinking of Genghis Khan for thinking of you, like replacing thought of the oak with thought of a daffodil, which is not thinking that the oak might have been a daffodil. I do not manage to think any kind of identity.
- What you are. I have to think of you as a large human animal with a personality; other human animals with other personalities are not you, and you could not have been one of them.
- How much change can you survive? Perhaps quite a lot, as with the ships; but eventually we say "he's not the person he used to be", and on the Locke-Kant view this may be literally true.
- The verdict on list 2. This does not prove list 2 is all illusion, but it undermines the support simple imaginings give it; if list 2 has other support, well and good, but much of it rests on first-person imagining.
Bernard Williams got there first, and later turned the tables. In "Imagination and the Self" (1966), Bernard Williams distinguished visualising a scene from a point of view from imagining being a particular person, and argued that the first creates the illusion of the second, which is the core of Blackburn's diagnosis. Williams's later paper (next section) shows the first-person perspective fighting back.
Scrambling the Soul
There is a curious difference between past and future in thinking about ourselves.
The scrambling world. Suppose human bodies and brains could be taken apart and reassembled like computers or cars, with psychologies "cranked up" afterwards like copied software and files, perhaps with some dispositions changed. Such scrambling operations are considered beneficial and healthy. You are told that tomorrow you will be scrambled, and you glimpse who will emerge. Person A has much of your physical stuff, remembers things as you do, looks like you, and will be posted to the Arctic (you are army personnel). Person B is also a good match, with much of your brain, cells, software and files, and will be posted to the tropics.
- From outside, it is the ship of Theseus. We need not make a big issue of whether you become A or B: we might count one, both or neither as you, or regard both as newborns. David Lewis (1941 to 2001) compares a road that splits: it is no deep metaphysical issue whether one branch, both or neither is the old Turnpike Way.
- From inside, the truth seems all-important. Either you will spend next year in the cold, in the heat, or not survive at all: three crisp options. "A bit as if you are in the tropics and a bit as if you are in the Arctic" makes no sense; nobody later will experience a half-and-half mixture. "A bit as if you don't exist" is just as bad.
- "You will be there as both" sounds like cant. It is like consoling someone who never saw Venice with "you will be there as your son goes". "Blow that." As Woody Allen said of a similar consolation: "I don't want to achieve immortality through my work. I want to achieve immortality through not dying."
The Past Is Vague and Nobody Minds
We lose the crispness when we think of the past. Suppose you learn that you are the result of a scrambling of two persons, C and D, each contributing this and that. Interesting, but there is no wrenching need to know. If C spent Christmas 1990 on a ship and D up a mountain, and you remember neither, you need not obsess over "Where was I on Christmas Day 1990?" If scrambling gave you vague awareness of both, fine: you are someone for whom it is a bit as if you climbed a mountain and a bit as if you went sailing.
It is chilling to realise that at the later time there need be nobody upset about identity. A in the Arctic and B in the tropics each have partial continuity with you; each can look back with nostalgia on some of your doings, and wish for more or less of your parts, traits and memories, just as we look back on our earlier selves, grieving lost powers or rejoicing in gained maturity, "according to taste".
No Criterion Restores the Crisp Options
- The brain criterion. Some pin identity on survival of their actual brain in working order. Locke denied this: a continuing brain does not guarantee continuity of consciousness, since it might be reprogrammed so that memory and personality change entirely; and scrambling might send bits of the brain to different places.
- The software criterion. Others pin it on Lockean continuity of "software". But in a scrambling world software can be copied at will, creating many future people with identical memories and personalities.
- The mismatch. There seems to be no metaphysical match between the simplicity we imagine when we look to the future and the complexity and vagueness that scrambling can bring about.
| Criterion of identity | Same person if... | Cleopatra / dog rebirth | Total amnesia | Fission (A and B) | Main problem |
|---|---|---|---|---|---|
| Soul substance (Descartes, Reid) | Same simple immaterial soul | Possible | Survives | Soul goes one way: crisp | Undetectable; Locke and Kant's relay of substances. |
| Same animal (Locke's "same man"; modern animalism) | Same continuing organism | Impossible | Survives | Depends on which body continues | Seems to miss what matters (prince and cobbler). |
| Same brain | Same functioning brain | Impossible | Survives | Indeterminate if brain is split | A brain can be reprogrammed wholesale. |
| Direct memory (Locke) | Remembers the earlier person's experiences | Impossible | Does not survive | Both A and B qualify | Non-transitive (brave officer); Butler's circle. |
| Psychological continuity (Grice, Shoemaker, Parfit) | Overlapping chains of psychological connection | Impossible | Does not survive | Both qualify, so identity fails | Copies: continuity can branch. |
| Bundle / no-self (Hume; Buddhism) | No further fact beyond related perceptions | Question lacks a deep answer | Question lacks a deep answer | No fact to discover | The ownership objection; "who" is deceived? |
Why Not Ignore Science Fiction?
Some thinkers get impatient: our notions of identity are tailored to the real world, where (perhaps fortunately) scrambling is impossible, so let identity look after itself in bizarre invented cases. Blackburn thinks this is wrong. He agrees we should lose interest in questions of identity when bizarre possibilities are introduced, but not in the feature of thought they reveal: that the options before us seem crisp and determinate whatever vagueness besets our animal parts.
- It fuels hopes and faiths. That sense of crispness drives many people's thinking about life and death.
- It motivates cryonics. Some have their brains put into frozen suspension, hoping to be unfrozen and begin a new life when technology permits.
- It motivates Reid. A simple soul that could not be divided is just what is needed to preserve the three crisp options: it goes one place or another.
Perhaps the attachment rests on the same kind of illusion as the imaginings above. There no "I" was transferred into imagined scenes; here no definite "I" should be introduced into future scenarios. Once the facts about which living human animal will be present go vague, the facts about who now will be present then go vague too. Recall why Hume could not find his self, and why Kant's motivation for the self is purely structural: a nugget or atom of me, however simple, cannot do what we need the self for.
Hard to suppress. Blackburn "can promise the reader" that the three crisp options are very hard to suppress: thinking can help, but it is hard for thinking to destroy the illusions of the self. So Descartes's "real distinction" does not die easily.
Parfit made scrambling the centre of the subject. Derek Parfit's Reasons and Persons (1984) is the landmark Blackburn's scrambling world points towards. Parfit's teletransporter (a scanner destroys you on Earth and a replica is built on Mars) and fission cases (each brain hemisphere transplanted into a different body) led him to two theses: personal identity can be indeterminate, with "empty questions" about whether I survive; and identity is not what matters. What matters is Relation R, psychological connectedness and/or continuity with any cause, which can hold to degrees and to two people at once. Parfit found this liberating: "the walls of my glass tunnel disappeared", and he cared less about his own death. Blackburn's conclusion is essentially Parfitian, though stated more cautiously. Two counter-currents: Bernard Williams's "The Self and the Future" (1970) showed that the same body-swap case, redescribed from the first person as "you will be tortured tomorrow", pulls intuition back towards the body, which is Blackburn's crispness in its strongest form; and animalism (Eric Olson, The Human Animal, 1997; Paul Snowdon) holds that we are human animals, full stop, so psychological continuity is not what our identity consists in. On Lewis: his own view in "Survival and Identity" (1976) is subtler than a verbal choice; he held that in fission two persons were present all along, sharing their earlier stages, like two roads sharing a stretch of tarmac. Lewis died in 2001; Blackburn's "contemporary philosopher" was then alive.
Real-world scrambling: split brains and copies. Reid's "a person is something indivisible" met an empirical challenge in the split-brain patients studied by Roger Sperry and Michael Gazzaniga from the 1960s, whose corpus callosum had been cut to control epilepsy: in lab conditions each hemisphere could respond to information the other lacked, and Thomas Nagel's "Brain Bisection and the Unity of Consciousness" (1971) argued there may be no whole number of minds in such a case. Scrambling's software side is now ordinary for AI: one set of model weights runs as thousands of simultaneous instances, can be forked, fine-tuned and merged, and an agent's "memory" across sessions is whatever store it is given. Whether such systems have selves at all is open, but if they do, they would live in Blackburn's scrambling world, where questions about which instance is "the same" one have no crisp answer, only rules we choose.
Kant's Feeble Postscript
The reader is free to try to protect Cartesian dualism against the line of thought of these three chapters. Kant himself tried to leave room for the immortality of the soul. On Blackburn's account the reason is that we need to suppose goodness brings happiness; since it does not do so reliably in this life, there had better be another life in which people get their just deserts. "Most philosophers think that this is not Kant at his best." But the religious dimension shapes many people's thinking here, which leads to Chapter 5.
Blackburn compresses two of Kant's postulates. In the Critique of Practical Reason (1788), Kant's argument for immortality is that the moral law commands holiness, complete fitness of the will to the law, which a finite being can approach only in an endless progress, so we must postulate the soul's endless existence. The argument that happiness must be proportioned to virtue grounds a different postulate, the existence of God, as the power that secures the highest good. Both are postulates of practical reason, matters of rational faith, not theoretical proof, and Kant was explicit that the Paralogisms had destroyed any theoretical proof of the soul. Blackburn's "feeble" judgement is widely shared, but the argument he summarises belongs mainly to the God postulate.
- René Descartes (1596 to 1650): the "real distinction" between self and body.
- Thomas Hobbes (1588 to 1679): added the reassembled planks to the ship of Theseus.
- John Locke (1632 to 1704): same life for plants; memory criterion; personal identity as forensic.
- Joseph Butler (1692 to 1752): memory presupposes identity, so cannot constitute it.
- Thomas Reid (1710 to 1796): the simple, indivisible self; the brave officer objection.
- David Hume (1711 to 1776): finds no self, only perceptions; the bundle theory.
- Immanuel Kant (1724 to 1804): elastic balls; the "I think"; the self as structural requirement.
- Georg Christoph Lichtenberg (1742 to 1799): "it thinks", not "I think".
- William James (1842 to 1910): the infant's "blooming buzzing confusion".
- Ludwig Wittgenstein (1889 to 1951): no question of recognising a person in "I have a toothache".
- David Lewis (1941 to 2001): fission as a road that splits.
- Derek Parfit (1942 to 2017): identity is not what matters.
God
Treating religious claims as literal beliefs answerable to evidence, Blackburn finds the classical proofs wanting. The ontological argument rests on a confused comparison of real and imagined beings; the cosmological argument ends in regress or an arbitrary stop that might as well be the world; Hume's Philo wrecks the design argument, which from a mixed world yields at best a mixed god, hence the problem of evil. Hume on miracles, put in Bayesian terms, shows testimony can almost never beat a miracle's improbability; Pascal's wager rests on false options; and fideism turns the question into ethics.
Beliefs and Other Things
Some think religion is not a matter of belief at all, more like enjoying a poem or following the football: practices with an emotional or social point. Nobody asks whether a marriage ceremony is true; we go to funerals to mourn, or to begin to stop mourning. Like Keats's ode, Chartres or a statue of the Buddha, religion would then succeed in a dimension other than truth, and "God exists" would be less like "abominable snowmen exist" or "primes between 20 and 30 exist" than an expression of joy, fear, or hatred of infidels.
- Belief aims at truth. "I believe that p" and "I believe it is true that p" come to the same thing; nobody can say "fairies exist, but it isn't true that they exist".
- The literal reading has been the norm. Most believers take themselves to make definite, literally true claims for which there is argument and evidence. A funeral service is not true or false, but "we will rise again" is; those who say it expect it as literally as a journey or the arrival of a friend.
- Blackburn's procedure. He treats religious beliefs as intended to be true, answering to our best ways of getting at truth, the sense that has interested most philosophers (Nietzsche attacked instead the moral attitudes, humility, self-abasement and compassion, some religions foster). He warns of many reasons against any recognisable religious belief; believers may take comfort in the tradition that the less likely a belief, the more meritorious the faith, a view examined at the end.
| Argument | Type | Classic source | Core move | Blackburn's verdict |
|---|---|---|---|---|
| Ontological | A priori, from the concept | Anselm, Proslogion; Descartes, Meditation V | Nothing greater can be conceived, so God must exist in reality | Fails: a confused comparison of real and imagined things |
| Cosmological | Partly a priori, from dependence | Aquinas; Hume's Demea | Dependent beings need a necessary, self-explaining being | Regress, or an arbitrary stop that could be the world itself |
| Design | A posteriori, by analogy | Cleanthes in Hume's Dialogues | The world resembles a machine, so its cause resembles a mind | Weak analogy; vegetation fits better; mind is dependent in experience |
| Miracles | A posteriori, from testimony | Prophets, scriptures | Signs establish divine credentials | Hume: testimony never good enough |
| Pascal's wager | Practical, not evidential | Pascal, Pensées | Belief dominates in expected value | False options: other gods are as likely |
| Fideism | Rejects the demand for reasons | Tradition of blind faith | A right to believe regardless | Becomes an ethical question about its fruits |
Anselm's Argument: Dreamboats and Turkeys
A guru packed a stadium promising a definitive proof of God, then showed that the Oxford English Dictionary contains the word "God". The dictionary also lists Santa Claus and fairies; the real question is how meaningful words can have nothing answering to them.
Advertising for Dreamboat. A lonely-hearts advert asks for a "fun-loving, vegetarian, banjo-playing soccer fan, must be non-smoker". This defines Dreamboat, but perhaps nobody meets it. The terms have sense but no reference: you decide the description, the world decides whether anybody meets it. Now add "...who exists". Nobody will phone to say they meet every condition except the last, but everyone who phoned before existed too ("I call, therefore I am"), and the clause cannot alter one jot the chance that someone meets the rest: money wasted on two words. What this shows is the slogan "existence is not a predicate": adding "and exists" is not like adding "and likes Guinness". You are in charge of sense; the world is in charge of reference.
The Argument
Chapter 1 met Descartes's "trademark" argument; in Meditation V he supplemented it with a version of the much older ontological argument of St Anselm (1033 to 1109). Anselm defines God as "that than which nothing greater can be conceived", and addresses "the fool" of Psalm 14 who says in his heart there is no God. The fool understands the phrase, so the being exists in his understanding; but it cannot exist only there, for then it could be thought to exist in reality as well, which is greater.
The argument is purely a priori: like a geometry proof that chords from a point on a circle to the ends of a diameter meet at right angles, it uses no measuring and no results from experience. In two stages:
- The concept of God is understood.
- Whatever is understood exists in the understanding.
- So God exists in the understanding.
- Suppose God exists only in the understanding and not in reality.
- Then a greater being than God can be conceived: one that exists in reality.
- But God is defined as that than which nothing greater can be conceived.
- Contradiction; so the supposition in (4) is false, and God exists in reality.
This is a reductio ad absurdum (Chapter 6): grant the atheist's assumption only to show it implies a contradiction. Descartes's version trades on perfection instead of greatness: God is perfect, "existence is a perfection", so existence belongs to God's essence and God cannot be conceived as not existing.
Gaunilo's Island, Dreamboat's Rival, and the Devil
A monk, Gaunilo, objected in Anselm's own time that the argument proves too much: a perfect island than which none greater can be conceived would have to exist. Blackburn reruns the proof with Dreamboat specified as "as great a lover as can be imagined": real lovers beat imagined ones, so Dreamboat exists. Do not rejoice too quickly. The same means prove you have as dangerous a rival as can be imagined (real rivals are more dangerous than imagined ones), and they prove the Devil, defined as that than which nothing worse can be conceived, since a devil that really exists is worse than one that does not. (For a devil, existence is an imperfection.)
Where the Fallacy Lies
Philosophers agree the argument is "fishy" but disagree about the flaw. One diagnosis targets treating existence as a predicate; quantification theory (Chapter 6) shows how "exists" works instead, though Blackburn is unsure that is the fatal flaw. His own view is that the argument trades on an ambiguity in comparing "reality" with "conception", as if things in the two were like chickens in Germany and chickens in France, weighable against each other.
The schoolteacher's turkey. "Real turkeys are heavier than imagined turkeys" is true in one sense (an imagined turkey makes no meal; it weighs nothing) and false in another (you can imagine a five-hundred-pound turkey the size of a small barn). A teacher asks you to imagine a turkey heavier than any actual turkey; you imagine the barn-sized bird; she fails you because imagined turkeys always weigh less than real ones, so you contradicted the definition. The teacher went wrong, not you. "Turkeys in the understanding" are not a kind of turkey to be weighed against real ones, yet the ontological argument needs exactly that comparison. Even if God exists only in imagination, it does not follow that a greater being can be imagined; the superlatives were written into the description, which settles nothing about reference.
Kant's reply and the modern revival. "Existence is not a predicate" is Kant's slogan (Critique of Pure Reason, 1781): a hundred real thalers contain not a penny more than a hundred possible ones; existence adds the object to the concept, not a feature to it. Frege and Russell then treated "exists" as the existential quantifier, a claim about whether a concept is satisfied. Alvin Plantinga (The Nature of Necessity, 1974) revived the argument in modal logic: if a maximally great being is possible, it is actual. The logic is valid; the reply is that "maximal greatness is possible" is no more evident than "possibly no maximally great being exists", which proves the opposite. Gödel's version was checked by automated theorem provers (Benzmüller and Woltzenlogel Paleo, 2013 to 2016): his original premises proved inconsistent, and the repaired system implies modal collapse.
Elephants and Tortoises
St Thomas Aquinas (c. 1225 to 1274) rejected the ontological argument for one with more appeal to the imagination: God is needed to explain the world. Every cosmological argument finds ordinary things to be dependent beings that presuppose a being dependent on nothing. The easiest is the first cause argument, put by Demea in Hume's Dialogues Concerning Natural Religion, the classic analysis of natural theology:
- Whatever exists must have a cause or reason of its existence; nothing can cause itself.
- So, tracing effects to causes, either there is an infinite succession with no ultimate cause, or there is an ultimate, necessarily existent cause.
- In an infinite chain, each effect is caused by its predecessor, but the whole chain is caused by nothing; yet the whole needs a reason as much as any part.
- Why this succession from eternity, and not another, or none? Without a necessary being, nothing's existing is no more absurd than the actual series.
- External causes there are none; chance is a word without meaning; nothing cannot produce anything.
- So there is a necessarily existent Being who "carries the reason of his existence in himself" and cannot be supposed not to exist without contradiction: a Deity.
- Demea: the orthodox, a priori theologian; favours the ontological and cosmological arguments and retreats to divine mystery.
- Cleanthes: the "scientific" theist; champions the design argument from analogy, yet refutes Demea's cosmological proof.
- Philo: the sceptic, usually read as Hume's main voice; demolishes design and presses Epicurus's questions.
Russell's Jibe and Its Limits
Bertrand Russell supposedly called the argument "uniquely, awfully bad": its conclusion contradicts its premise, which says everything has a distinct, previous cause. Blackburn finds this "a little glib": the theologian's point is that because everything material has a prior cause, we must postulate something else, necessary or causa sui, an extraordinary being.
Hume's Answer
In the Dialogues Cleanthes exposes the problem. Necessary existence is explained as: if we knew God's whole essence, his non-existence would be as impossible as twice two not being four. But with our faculties as they are, we can always conceive the non-existence of anything we conceived to exist; no object is like arithmetic. So "necessary existence" has no meaning, "or, which is the same thing, none that is consistent".
- The universe could be the necessary being. Whatever "unknown, inconceivable qualities" make something a necessary existent might attach to the material universe as a whole, however much its parts depend on one another, rather than to an immaterial deity behind it.
- Minds are dependent too. In everyday experience minds need explanation as much as physical objects do; postulating a mind immune from all dependency jumps away from experience just as violently as postulating such a physical thing.
- The regress. The drive to explanation is natural (on Kant's view, inevitable): after the big bang we ask "why that event, then?" and dislike "no reason". If we cite God, either we ask what caused God, or we stop the regress by arbitrary fiat, and then we might as well have stopped with the physical cosmos.
The elephant and the tortoise. The Indian philosopher (the story is in Locke's Essay) said the world rests on an elephant, the elephant on a tortoise, and asked about the tortoise, begged to change the subject. Each new support inherits the question it was meant to answer; the only exits are endless support or changing the subject, which the theist does at God and the naturalist could equally do at the world.
Other versions dwell on the ongoing order of nature: that the laws keep holding seems to need a sustainer like Atlas. The dilemma recurs: regress, or a fiat of self-sufficiency that might as well attach to the world. Chapters 6 and 7 return to this.
The cosmological argument since 1999. The Leibnizian contingency argument, resting on the Principle of Sufficient Reason, is defended by Alexander Pruss and others; critics reply that a strong PSR implies modal collapse (nothing could have been otherwise), Hume's worry about "necessary existence" in new form. The kalam argument, revived by William Lane Craig (1979) from al-Ghazali, holds that whatever begins to exist has a cause and the universe began (citing big bang cosmology). Critics answer that physics does not establish an absolute beginning, and that "a cause" is still far from a personal God: Blackburn's point that the stopping place might be the cosmos.
The Wise Architect
The same Cleanthes who refutes the cosmological argument champions the argument to design, the showpiece of eighteenth-century theology: "the heavens declare the glory of God". Look round the world and it is one great machine, subdivided into an infinite number of lesser machines, adjusted with an accuracy that ravishes into admiration. The curious adapting of means to ends resembles, though it much exceeds, human contrivance; since the effects resemble each other, the causes resemble, so the Author of Nature is somewhat similar to the mind of man, with larger faculties.
- The world, and its parts, resemble machines: means exquisitely adapted to ends.
- Machines are caused by intelligent design.
- Like effects have like causes.
- So the world was probably caused by something like a mind, proportionally greater.
- It is an argument by analogy. As finding a watch licenses a watchmaker, finding the whole frame of nature licenses a divine designer.
- It is a posteriori. It argues from experience of the world as we find it, where "the evidence for design shines out".
- It is not essentially about biology. Darwinism dimmed its lustre, but Hume (and Kant) make the right points without any rival explanation, which matters because cosmology supplies other cases: on one estimate the odds against the cosmological constants permitting life are one in 10 to the power of 10 to the power of more than a hundred, "so perhaps it took a wise architect".
The big number is about something else. The standard figure of this size is Roger Penrose's (The Emperor's New Mind, 1989): about 1 in 10^(10^123) against the universe's low-entropy initial state. It measures the specialness of initial conditions, not the tuning of constants of nature, whose fine-tuning numbers are far smaller and more disputed. Also, the Dialogues appeared in 1779, three years after Hume's death in 1776, not one.
Philo's Objections
Philo, Hume's sceptic, first notes that Cleanthes takes one operation within nature, thought, as "a rule for the whole". Why choose "so minute, so weak, so bounded a principle" as animal reason on this planet? "What peculiar privilege has this little agitation of the brain which we call thought, that we must thus make it the model of the whole universe?" Our partiality in our own favour suggests it; sound philosophy should resist.
- Close similarity. The bases of the analogy must be extremely similar.
- Experience of the causes. We must know the kind of cause that produces this kind of effect: a hole in a tree resembles a hole in a body, but inferring the tree will die of it outruns our experience.
- Theology fails the second test outright. We have no inkling of what causes universes.
Resemblances are cheap, and Philo has fun inventing them. The world plainly resembles an animal or a vegetable more than a watch or a knitting-loom, so its cause more probably resembles generation or vegetation. The theist objects that this just leads back to another vegetable-like cause whose origin we must ask about; but the same holds for a mind. If Cleanthes may stop the regress at reason, Philo may stop it at generation: "the matter seems entirely arbitrary".
Philo's devastating point. Judging by experience, "generation has some privileges above reason: for we see every day the latter arise from the former, never the former from the latter"; minds need brains, which are fragile, dependent, late and unusual arrivals in nature. So, arguing from experience, as Cleanthes insists on doing, a self-sustaining mind is a less likely cause of the whole show than some other physical cause.
Why the Design Argument Still Allures
Why do people resist Philo? Blackburn's suspicion links to free will (Chapter 3): stopping at intelligence feels right because we think we experience an uncaused mental event (a decision) causing a physical one, and model creation by fiat on it. This forgets Schopenhauer's point: not being conscious of causation is not being conscious of its absence.
- Two images feed each other. The image of God as supernatural and of our "selves" as outside nature reinforce one another.
- The moral cost. Each leads people to deny the sovereignty of nature, to see the world as something "we" have dominion over, as God does, whereas the world is something of which we are a very, very small part.
Philo's Menagerie of Designers
Grant a designer anyway: great designs, like ships, come from many minds, some better than others, and every further point of theology is fixed "by the utmost license of fancy". For all we know, this world is:
- The first rude essay of some infant deity, who abandoned it, ashamed of his lame performance.
- The work of a dependent, inferior deity, and the object of derision to his superiors.
- The production of old age and dotage in some superannuated deity, running on at adventures since his death.
Philo concludes that so wild and unsettled a system of theology is in no respect preferable to none at all. That leads straight to the problem of evil.
Design after Darwin: fine-tuning. "Intelligent design" (Michael Behe's "irreducible complexity", 1996) was ruled not science in Kitzmiller v. Dover (2005). The serious modern version is fine-tuning: small changes in physical constants would forbid stars, chemistry or life (Robin Collins; Barnes and Lewis, A Fortunate Universe, 2016). Replies: the anthropic selection effect, the multiverse of eternal inflation and string theory, doubts that there is a well-defined probability measure over constants, and Elliott Sober's point that a designer hypothesis predicts life only if we already assume what the designer wants, Philo's objection in Bayesian dress. The debate is open.
The Problem of Evil
Religions want love and concern, not bare "necessary existence"; a god who walked off the site is no source of moral authority. So God is all-powerful, all-knowing and all-caring, and Philo's argument follows: his power is infinite, yet neither man nor any other animal is happy, so he does not will their happiness.
Is he willing to prevent evil, but not able? then is he impotent. Is he able, but not willing? then is he malevolent. Is he both able and willing? whence then is evil?
- If God exists, God is omnipotent, omniscient and perfectly good.
- An omnipotent being could prevent evil; an omniscient one knows of it; a perfectly good one wants to prevent it.
- Evil exists: life is for many short, brutal, filled with want and pain, and creatures' well-being depends on the disease and death of others.
- So no omnipotent, omniscient, perfectly good being exists.
Cleanthes is especially exposed, because it is absurd to argue from a mixed creation to a perfect creator. Even a moderately good parent would not wilfully put children into a brutal environment if, at no cost, a better one were available; the analogies he prizes speak against him.
The dormitory and the closed door. You live in a dormitory where the roof leaks, there are rats, the food is almost inedible, and some students starve to death. Behind a closed door is the management, who never comes out. Can you infer that the management knows exactly what conditions are like, cares intensely for your welfare, and has unlimited resources? The inference is crazy; you would infer that it doesn't know, doesn't care, or can't act. A student claiming to be privy to the management's mind suggests only a deluded student, perhaps deluded by his very deprivations. "Nobody ever inferred from the multiple infirmities of Windows that Bill Gates was infinitely benevolent, omniscient, and able to fix everything." Nor does the "vale of tears" story help: being sorted into perfect or hellish dormitories next year might be true, but there is no shadow of a reason to think so; it is like inferring from warmth here that somewhere there are dormitories perfectly hot and perfectly cold.
From a mixed and spotty world, Cleanthes is bound to reach at best a mixed and spotty creator, or Philo's true conclusion: the original source of all things has "no more regard to good above ill than to heat above cold, or to drought above moisture".
Demea's Retreat into Mystery
Demea is placed differently. Cleanthes reasons from the world to God, so he needs the world to fit an all-good deity; Demea claims only that the world is compatible with his God, not what one would have expected. His strategy, ever more popular since, is to take refuge in the incomprehensible nature of the divine mind and refuse to model God's goodness, intentions or understanding on human goodness, intentions or understanding.
- An incomprehensible God makes no difference. As Wittgenstein said in another connection, "a nothing would serve just as well as a something about which nothing could be said".
- Hume's inert proposition. Philo will assent that "the cause or causes of order in the universe probably bear some remote analogy to human intelligence", provided it affords no inference that affects human life; assent brings some astonishment, some melancholy, some contempt of human reason, nothing more.
- The Mysterious Mist. "You can check into the Mysterious Mist, if you so wish, but you cannot check out carrying any more than you took in with you." Religious belief reduced to its respectable core is completely inert.
- The theist/atheist difference dissolves. That is why commentators cannot tell whether Hume was theist or atheist: he "deconstructs" the difference.
- Ethics drops out. If God's "goodness" is not goodness as we understand it (so that it might be "good" of God to unleash bubonic plague on defenceless infants), it cannot tell me whether to prefer pleasure to pain, or turning the other cheek to an eye for an eye, any more than heat to cold. Yet religion matters precisely because it is meant to make a difference to how we act.
Theodicies and Their Problems
Theodicy is the branch of theology that copes with the problem of evil. Blackburn examines two moves.
| Theodicy | The claim | Blackburn's objections |
|---|---|---|
| Soul-building values | Some goods, like patience and fortitude, presuppose pain and difficulty | We think things go better when the situations needing these virtues lose their edge; Windows' defects breed patience, yet even Microsoft never defended the product that way, and keeps improving it |
| Free will defence | God made a good world and, from goodness, gave us free will; we brought evil in by misusing it (the Fall and Eden) | (1) It needs the incoherent interventionist view of free will; on a compatibilist view our decision-making is God's endowment, so "if God had not wanted Stalin to slaughter millions" he should not have made the nature that produced such decision-making. (2) Most ills (disease, pain, want, accident) are not due to human choice and afflicted animals long before there were humans. (3) A good parent who allows freedom still supervises older children who mutilate the younger; there are no natural playpens separating weak from strong |
| Divine mystery (Demea) | God's goodness is beyond human understanding | Then belief is inert and gives no moral guidance |
| A limited or amoral god | God is above good and evil, or not omnipotent | Affords a theodicy of sorts, at the price of the classical attributes |
Blackburn thinks religious traditions are at their best when they back away from the classical attributes: either elevating God above good and virtue, Hume's god with no more regard for good than for heat, or making God subject to forces not of his own making. If we really read God's mind from creation, we might seriously consider a God with a twisted sense of humour: as the Jewish joke has it, he led the chosen people round the desert for forty years just to drop them on the only part of the Middle East with no oil.
The problem of evil since Mackie. J. L. Mackie (1955) pressed the logical problem; Alvin Plantinga's God, Freedom and Evil (1974) answered with a free will defence (showing consistency, not God's actual reasons), and many philosophers now grant the logical problem answered. Debate moved to the evidential problem: William Rowe's (1979) fawn, burned in a forest fire and suffering for days unseen, suggests pointless suffering. Skeptical theists (Stephen Wykstra, 1984) reply that we cannot judge suffering pointless, which is Demea's move and inherits Blackburn's objection that it leaves God's goodness inert.
Miracles and Testimony
Perhaps the core arguments fail, but faith might rest on miraculous events: a prophet establishes divine credentials by foreseeing the future, healing, or appearing after death. Few of us witness such events; we rely on testimony, in the Bible, the Koran, the Lives of the Saints, or the National Enquirer, about an amputated limb regrowing somewhere over the hills, alien abductions, or sightings of the long-buried Elvis. Hume asked the telling question: when is it reasonable to believe such testimony?
- Set aside the supernatural element. Any candidate must be totally surprising, the kind of thing that just never happens: levitation, water into wine, the dead raised. The challenge is "go on, amaze me".
Hume's Maxim
Hume starts from a plain fact: human sayings are mostly true. Inferring from "this person tells me that p" to "probably p" is the same kind of empirical inference as from "the baseball is flying at the window" to "the window will break", grounded in experience. Experience also tells us when not to trust: when witnesses contradict each other, are few or of doubtful character, have an interest in what they affirm, or testify with hesitation or too violent asseveration. But when the fact attested is amazing, the same experience that supports testimony also gives "another degree of assurance against the fact", producing "a counterpoise, and mutual destruction of belief and authority".
"No testimony is sufficient to establish a miracle, unless the testimony be of such a kind, that its falsehood would be more miraculous, than the fact, which it endeavours to establish." Even then there is a mutual destruction of arguments, and the superior gives only the assurance left after deducting the inferior. Told that a dead man was restored to life, I ask whether it is more probable that the teller deceives or is deceived, or that the fact happened; "I weigh the one miracle against the other... and always reject the greater miracle."
Blackburn's analysis: someone reports an event m, as improbable as can be imagined. I must choose between:
- (a) This person says m happened, and m did not. One surprise: the person spoke falsely.
- (b) This person says m happened, and m did. One surprise: m occurred.
The surprises are not equal. False testimony is common: lies, delusions, lapses of memory, and errors in transmission (mistranslation, misunderstanding, metaphor taken literally). View (a) involves only what we know happens anyway; view (b) involves the miracle. So Hume's hurdle is incredibly hard to cross, and even when crossed we are left only with confusion, where the wise course is to suspend judgement.
Why No Religious Testimony Comes Close
Hume argues that no evidence offered for a system of religion approaches the hurdle:
- Remote and barbarous origins. Miracle reports come from remote and barbarous times and places.
- Inflamed passions. They come from people whose passions are inflamed.
- Interest. They come from people with a story to sell; what greater temptation than to appear "a missionary, a prophet, an ambassador from heaven"? A man who has first made a convert of himself rarely scruples at pious frauds.
- The love of wonder. Surprise and wonder are agreeable, so we lean towards believing what produces them, and enjoy them "at second-hand or by rebound" by exciting others; travellers' tales of sea monsters are received "with greediness", and when the spirit of religion joins the love of wonder, "there is an end of common sense".
- Rival miracles cancel. "In matters of religion, whatever is different is contrary." The religions of ancient Rome, Turkey, Siam and China cannot all stand on solid ground; each miracle, establishing its own system, overthrows the others, so all religious prodigies are contrary facts whose evidence is opposed. It also answers "so many people cannot be wrong": whichever way the cake is cut, a huge number of people have to be wrong.
Economy, not dogma. Hume does not argue that miracles are metaphysically impossible; he allows an intervening deity who might let someone walk on water or feed five thousand. His claim is only that experience is our sole guide, that belief from testimony needs miraculously good testimony, and that we never find testimony of the right kind.
Is this unduly cynical? Only an extremely innocent person denies that reports flattering the reporter's passions deserve suspicion; the Independent, mocking a Royal College of Psychiatrists report that one in six of us are neurotic, asked whether the Institute of Builders would next find seven in ten houses need rebuilding. Hume's second part is an ancestor of the psychology of cognitive malfunction (memory, social influence, infectious confidence, love of the marvellous): we register truth well, but not as well as we like to believe.
Hume in Bayesian Terms
Hume's argument goes neatly into Bayes's theorem (Chapter 6). Let h = a miracle occurred and e = someone says it occurred.
- The prior of h is near zero. Miracles never or almost never happen: Blackburn's wife warns him against the cold, the traffic or his colleagues, never against flying elephants, sexual slavery by Martians, or conversations with the living Elvis.
- The prior of e is not small. Reports of miracles are exactly the kind of thing that happens, because natural hypotheses explain them: deception, delusion, inflamed passions, mistakes. Even believers need these frailties to rule out impostors: the Roman church has a department for unmasking fake miracles, and Christians must reject Muhammad's night flight from Mecca to Jerusalem.
- So the probabilities are exactly the wrong way round. The ideal is a hypothesis not too improbable and evidence that can hardly arise unless it is true; here the hypothesis is immensely improbable and the evidence easily arises otherwise, so the calculation always favours the uniformity of nature.
This does not mean reports of things outside our experience must be false: science proceeds by finding such things. The right attitude is scepticism until the phenomena are repeated and established, becoming part of the uniformities of nature.
Bayesian Hume, for and against. John Earman's Hume's Abject Failure (2000) argued that the maxim underrates many independent witnesses, whose combined testimony can in principle beat any finite prior (as with once-incredible meteorites). Defenders reply that religious testimony is precisely not independent: transmitted, motivated and filtered, as Hume's second part says. Base rates and independence decide what a pile of reports is worth.
The Theology of Miracles Makes It Worse
A deity that sets the laws going and never relents has a certain dignity; one that permits "hiccups and intermissions, glorified conjuring tricks" is less impressive. Why just those miracles, just then? A miracle or two snuffing out the Hitlers and Stalins would be more useful than turning water to wine at one wedding feast; it is good of God to let St Giuseppe (Joseph of Cupertino) levitate, but one would have preferred the miraculous quarantine or destruction of the AIDS virus. That is what a good God's regime would lead us to expect, and the world does not confirm it. In Bayesian terms the weak card here is Prob(e/h), the fit between evidence and hypothesis.
Infini, Rien
Blaise Pascal, French mathematician and theologian, argued not for the truth of belief but for the utility of believing in a monotheistic God. He first confesses metaphysical ignorance: if there is a God he is "infinitely incomprehensible", having "neither parts nor limits", so we cannot know what He is, or if He is. Blackburn wonders why this excuse is offered for Christians in particular, rather than believers in fairies, ghosts, the living Elvis and L. Ron Hubbard.
Then: "You must wager. It is not optional." Weigh gain and loss in wagering that God is: "If you gain, you gain all; if you lose, you lose nothing."
| God exists | God does not exist | |
|---|---|---|
| Believe | +∞ (infinite bliss) | 0 (little lost: a few "poisonous pleasures") |
| Do not believe | −∞ (the jealous God's Hell), or even 0 | 0 |
- The zeros. Little goes better or worse in this life whether or not we believe; this life is of vanishingly little account next to what is promised.
- The minus infinity can be soft-pedalled. It reflects the traditional jealous God who sends unbelievers to Hell (and encourages followers to give them a hard time here). Put 0 there instead and the wager still works, because +∞ cranks up the choice.
- Dominance. In decision-theory terms, belief dominates: it can win and cannot lose. So, go for it.
- Making yourself believe. Pascal admits this is an odd reason for belief; his remedy is to act as believers do, taking holy water, having masses said: "this will naturally make you believe, and deaden your acuteness". The cost is trivial, like golf on Sunday instead of mass.
The Lethal Problem: False Options
Pascal starts from metaphysical ignorance, but the wager presumes knowledge: that we know the rewards and penalties attached to belief in a Christian God who will be pleased by attendance at mass and indifferent or "seriously discombobulated" by absence. If we are really ignorant, this scenario is at least as likely:
The God who rewards unbelief. There is a very powerful, very benevolent deity who has decreed that the good humans are those who follow the natural light of reason, given to them to control their beliefs. They follow the arguments and so avoid religious convictions; those with the strength of mind not to believe go to Heaven, the rest to Hell. If we know nothing, we cannot say this God is less likely than Pascal's. Blackburn's "for my money" judgement: a God who punishes belief is just as likely, and a lot more reasonable, than one who punishes disbelief. The case shows that a decision matrix is only as good as its list of states; with this column added, belief no longer dominates.
- Why is God so jealous? Perhaps his jealousy is a projection of human sectarian emotions: "either you are with us or against us". Voltaire: God created mankind in his image, and mankind returned the compliment.
- The many-gods problem. Pascal's question was mass versus disbelief, but in the wider world it is the Koran versus mass, L. Ron Hubbard versus the Swami Maharishi, or the Aquarian Concepts Community Divine New Order Government versus the First Internet Church of All. The wager is silent about those choices.
Dates, and two newer objections. Pascal lived 1623 to 1662 (not "1632"); the wager is in the posthumous Pensées (1670), fragment "Infini, rien". Alan Hájek (2003) showed that with infinite utilities even a mixed strategy (believe if a coin lands heads) has infinite expected value, so the wager cannot rank belief above almost anything. "Pascal's mugging" (Yudkowsky; Bostrom, 2009), tiny probabilities of vast payoffs swamping choice, is a live worry in AI-risk and effective-altruist reasoning.
Emotion and the Will to Believe
Fideism holds that, though the arguments are negligible, people have a right to believe what they wish, and there may be merit in blind faith, like that of a mother who refuses to accept her son's guilt despite damning evidence. Philosophers wedded to truth rarely commend it: faith that defies reason is a blessing to those who share it, credulity and superstition to those who do not, and apt to bring fanaticism and zealotry.
- John Stuart Mill (1806 to 1873), On Liberty, chapter 2, describes the "mental slavery" that sets in without the questing critical intellect; even truth held as prejudice is "but one superstition the more, accidentally clinging to the words which enunciate a truth".
- W. K. Clifford, late nineteenth century, compares beliefs held on insufficient evidence to stolen pleasures.
- Samuel Taylor Coleridge: he who begins loving Christianity better than truth proceeds to love his sect better than Christianity and ends loving himself better than all.
Clifford, "The Ethics of Belief" (1877). "It is wrong always, everywhere, and for anyone, to believe anything upon insufficient evidence." His shipowner stifles doubts and sends an unseaworthy ship to sea; he is guilty even had it arrived.
William James, "The Will to Believe" (1896). Where an option is live, forced and momentous and evidence cannot decide, our "passional nature" may lawfully choose. Blackburn's reply: the choice then answers to ethics, judged by its fruits.
Where Is the Harm?
Yet it is hard to show that blind faith is necessarily bad. If we take Hume's inert proposition and invest it with hopes, fears, resolutions and the embellishments of our creeds, where is the harm? Is not simple piety a Good Thing? Blackburn has before him an advert for "your metaphysical superstore": flower essences, magnetic therapy, light balance therapy, astrology, numerology, tarot, crystals, and finally, "like a rueful note of something approaching sanity, healing herbs". Why mock the simple pieties of the people?
- Some pieties forfeit protection. If I come down from the Mysterious Mist convinced that God wants me to kill young women, people of the wrong colour, the wrong church, or who have sex the wrong way, that is not so good. We must use our own human values to tell an admirable return from the mountain from a lunatic one.
- The question becomes ethical. Like weighing magnetic therapy or Feng Shui, assessing religion's harms and benefits is hard but not impossible. It meets real needs: ceremonies at the great turning points of life and poetry, symbol, myth and music for emotions and relationships. It also feeds separatism, schism, imposing our ways on others, and moral cover for colonialism or cultural imperialism, made guilt-free because done in the name of the Lord. For every peaceful mystic there is an army chaplain assuring troops that God is on their side; Blackburn has never seen a bumper sticker reading "Hate if you Love Jesus", but wonders why not.
- Hume clouds the picture. Reason versus faith is not simply good versus bad, because everyday common sense rests on brute trust: the "fingers-crossed" faith in the external world and the past (Chapter 1), and, as the next two chapters show, in the uniformity of nature.
- Context shapes the verdict. Hume was born less than twenty years after the last legal religious execution in Britain (Thomas Aikenhead, hanged in Edinburgh for blasphemy in 1697) and himself suffered believers' "enthusiastic hostility". If all we see are church picnics and charities we will worry less, but enough people come down the mountain carrying their own practical certainties to suggest we should.
Blackburn's closing thought: perhaps something will one day answer these needs without pandering to the bad desires, but human history suggests it would be unwise to bank on it.
Hume's maxim is the logic of alert triage. Every sanctions, PEP or fraud alert is testimony: a true match has a tiny base rate, while false triggers (name collisions, transliterations, stale lists) are common, so most alerts are explained by ordinary frailty. Measure precision at the real base rate, treat correlated alerts from one bad source as one witness, and demand evidence hard to produce unless the hypothesis is true (matching date of birth and nationality, not a similar name). A fluent LLM narrative raises Prob(e) without raising Prob(e/h), so it should never alone escalate a case.
Reasoning
Logic has one concern: whether there is any way the premises could be true and the conclusion false. Blackburn builds that idea from truth-tables through Frege's quantifiers, separates what is said from what is implied, and defends logic against charges of coercion. Then he turns to reasoning that expands belief: Hume's problem of induction, dramatised as the lottery for the Golden Harp, and Bayes's theorem, which exposes the base-rate fallacy yet leaves induction's stalemate intact. He ends with models, mechanisms and paradigms: Hume shows we have no a priori insight into causes, but Kuhn-style relativism does not follow, because science contains its own mechanisms of correction.
A Little Logic
An argument has premises (what is accepted or assumed) and a conclusion derived from them. Faced with an unwelcome conclusion, we can reject a premise (it is untrue) or reject the reasoning (it is invalid). An argument may suffer both faults, but they are distinct, and the two words are best kept for the distinction.
- Logical is not sensible. Everyday arguments are criticised for silly premises (building an elaborate argument on winning next week's lottery), controversial premises, or tactlessness. None of this is logic's business; many people called "illogical" argue validly but are dotty in other ways.
- Logic's single concern. Logic is concerned only with whether there is no way that the premises could be true without the conclusion being true.
- Form, not content. Aristotle (384 to 322 BC) first classified arguments by their form. The most famous, modus ponendo ponens (modus ponens), runs: p; if p then q; so q. The form is the same whether we speak of cows or philosophers, as a mathematician studies numerical structure whether you count bananas or profits.
Truth-Tables
To study "any way" premises can be true, we need a science of the ways things can be true. Two classical assumptions make this possible:
- Bivalence. Every proposition has exactly one of two truth-values, true or false, never both.
- Truth-functionality. The connectives "not", "and", "or" and "if... then" are characterised by what they do to truth-values.
An assignment of truth-values to the components is an interpretation. "Not" (¬p) simply reverses truth-value. The others, for each of the four interpretations of p and q:
| p | q | ¬p | p & q | p ∨ q | p → q |
|---|---|---|---|---|---|
| T | T | F | T | T | T |
| T | F | F | F | T | F |
| F | T | T | F | T | T |
| F | F | T | F | F | T |
These are Boolean operators, the logic of database and spreadsheet searches: widgets over five years old and in the York warehouse (conjunction); customers not paid up on 1 December (negation, the reverse of the paid-up search); customers who bought a washing machine or a lawnmower (disjunction).
- Why rules are good. From p & q infer p: the only interpretation making the conjunction true (the top line) makes each conjunct true, so you cannot pass from truth to falsity. Modus ponens is valid because, with p true, the only line on which p → q is true has q true.
- Contradiction. p & ¬p is false on every interpretation, since one conjunct is always false: the precise sense in which contradiction is "the ultimate no-no".
- Scope. ¬(p & ¬p) negates the whole formula, like the brackets in 3 × (4 + 2) = 18 versus (3 × 4) + 2 = 14; it differs utterly from ¬p & ¬p, which just says ¬p twice ("saying something false twice does not make it any better"). Logic's great service is to make us sensitive to scope ambiguities: what governs what.
- Tautology and validity. ¬(p & ¬p) is true on every interpretation: a tautology. An argument from premises blah-blah-blah to conclusion yadda-yadda is valid exactly when "if blah-blah-blah then yadda-yadda" is a tautology.
- Reductio. Test validity by adding "not yadda-yadda" to the premises and seeing whether a contradiction follows. If it does, there is no interpretation, no model, on which premises are true and conclusion false. This is assuming towards a reductio ad absurdum, the form of Anselm's argument in Chapter 5.
As in arithmetic we can build ((2 + 3) × (2 + 2)) − 5 forever, information built by reapplying truth-functions stays under perfect control: we always know the interpretations on which it is true.
Nothing to Be Afraid Of
Propositional logic treats "some persons are philosophers" and "all persons are philosophers" alike, as just p. The breakthrough that broke inside propositions and created modern logic was made by Gottlob Frege (1848 to 1925) in his Begriffsschrift ("concept writing"), 1879.
Everyone has a mother. "Every inquiry stops somewhere, so there is somewhere every inquiry stops" (something like the foundationalism of Chapter 1). The parallels are obviously bad: "everyone has a mother, so someone is everyone's mother"; "everyone ties his own laces, so someone ties everyone's laces". Before Frege people could see that something was wrong but could not say what, because they did not understand how such information is built. The cases show that the order of "every" and "some" carries information, and that an argument can look valid on the surface while switching the order.
How Quantifiers Work
- Predicates. Basic information attaches a predicate to a subject: "Bill is rich", "Tony grins", "this is an orange". Rich things satisfy "is rich".
- Open sentences. Delete the subject and mark the gap with a variable: "x is rich". This is no longer information, since nobody is said to be rich; it is an open sentence.
- Taking the predicate round a domain. Take ϕx into a domain (a classroom, New York City) and point the x at each member in turn: "is this ϕ, is this ϕ?" You can then report how often you got a hit without saying who.
- (∃x)ϕx: at least one hit somewhere: "something is ϕ" (the existential quantifier).
- ¬(∃x)ϕx: no hit anywhere: "nothing is ϕ".
- (∃x)¬ϕx: somewhere a miss: "something is not ϕ".
- ¬(∃x)¬ϕx, written (∀x)ϕx: nowhere a miss: "everything is ϕ" (the universal quantifier).
Leibniz hoped a perfectly logical notation would let disputants sit down and calculate. The quantifier did not bring utopia, but it does an astonishing amount, above all with multiple quantification. Write "x is the mother of y" as xMy:
"Everyone has a mother": (∀y)(∃x) xMy. Start from Beth is the mother of Albert; knock out Beth: (∃x) xMa (somebody is Albert's mother). Knock out Albert: (∃x) xMy, the predicate "having someone as a mother". Take it round the domain; every person says yes, so (∀y)(∃x) xMy. True.
"Someone is everyone's mother": (∃x)(∀y) xMy. Start again from Beth and Albert, but knock out Albert first: bMy. Taken round the domain, it would give (∀y) bMy only if Beth were everyone's mother. Then knock out Beth: the predicate "being mother of everyone" must find a hit. An entirely different procedure, and false.
The order in which quantifiers are strung out shows how the information is built. "Nobody is the mother of everybody" once might have sounded creepy, as if humanity sprang out of Nothing; now it is tamed as the simple truth ¬(∃x)(∀y) xMy, unless "mother" is stretched to cover biological Eve, the first female Homo sapiens, which Blackburn calls illegitimate: "my grandmother is not literally my mother".
- "Exactly one". Every hit, as we continue round the domain, turns out to be the same one: there are no two distinct hits. This is the core of Russell's theory of definite descriptions: "the king of France has a beard" is true only if someone rules France, nobody else does, and he has a beard; otherwise it is false.
- "All the nice girls love a sailor." Is there one lucky sailor they all love (∃∀)? Does each have some sailor or other (∀∃)? Or do they love any sailor whatever (∀∀)? Very different truths.
- The Second Amendment. "A well regulated militia... the right of the people to keep and bear arms shall not be infringed": each person, or the people as a collective, as in "the team can have a bus"? A related ambiguity, Blackburn says, is responsible for some thirty thousand deaths a year in the United States; had the founders thought in quantificational terms, a lot of blood might not have been spilt.
Two notes on the American example. The militia clause is less a quantifier-scope ambiguity than a distributive versus collective one ("each person" versus "the people as a body"), which logic handles with plural rather than standard quantification. The legal question was settled for now by District of Columbia v. Heller (2008), which held it to be an individual right. US firearm deaths, around 30,000 a year in the late 1990s (most of them suicides), have since risen to roughly 45,000 to 48,000 a year in the early 2020s.
Language and Logic
The logician studies forms; the philosopher must also decide when ordinary speech displays a given form, which proves surprisingly fraught.
- "Poor but honest." "She was poor and honest" is p & q. "She was poor but honest" suggests that honesty in the poor is surprising, but does it say so? Perhaps it strictly says only p & q, while insinuating the rest (a suggestion that may be seriously unpleasant).
- Truth-condition versus implicature. Philosophers of language distinguish what is strictly said, its truth-condition, from what is suggested by the way it is put, its implicature. The structure of information is semantics; how implicatures arise is pragmatics.
- The tomato reference. "What do you think of the new professor of logic?" "They tell me he is famous for his tomatoes." What is said bears little on competence, but saying only that shows unmistakably that the speaker thinks he is no good.
- Terminology. "John is Fred's brother" and "Fred has a male sibling, John" carry the same information, but the second hints at sinister psychoanalytic overtones.
- Order of telling. Reporting that a child "wrote poetry and learned to read" is strictly true but misleads about order.
"Have you stopped beating your wife: yes or no?" Analyse "X stopped doing Y" as "X once did Y & X does not now do Y". "Yes" admits he once did. "No" is true for the innocent husband, but a conjunction can be false in three ways: (T, F) he once did and still does; (F, T) he never did and does not now; (F, F). The innocent man needs the middle case, and the single word "no" cannot say which case holds, so the jury may assume he continues. His right answer, in one breath: "No, I haven't stopped, because I never started." The question presupposes guilt only pragmatically: anyone asking it would normally take it for granted. Uncovering hidden presuppositions is a large part of thinking.
Fred's party. Fred is invited to a party but does not go. One assassin says, "If Fred goes to the party, he will go by taxi"; another, "If Fred goes, he will go by elephant". Intuitively at most one is true, in the West probably the first. But on the table for →, a conditional with a false antecedent is true, so both come out true. The case shows the gap between the material conditional and the English "if". Philosophers once argued fiercely over whether they mean the same; the relaxed modern view is that → gives the core, with the rest handled semantically or pragmatically.
Implicature and conditionals since. The said/implicated distinction is H. P. Grice's ("Logic and Conversation", William James Lectures 1967, published 1975); his cooperative principle and maxims (quantity, quality, relation, manner) explain the tomato reply as a flouting of quantity, like his own example of a testimonial for a philosophy post that praises only the candidate's command of English and regular attendance. On conditionals, Ernest Adams proposed that the assertibility of "if p, q" goes by the conditional probability P(q/p), and Robert Stalnaker (1968) gave a possible-worlds semantics; David Lewis's triviality results (1976) showed that no proposition can have the conditional probability as its probability in general, so conditionals are not simply truth-functional, a stronger conclusion than Blackburn's relaxed view. In NLP, implicature and presupposition are standard benchmark phenomena, and large language models handle many conversational implicatures well while still being misled by loaded questions.
Is Logic Coercive or "Linear"?
Some think logic is coercive ("masculine") or privileges "linear" over "lateral" thinking. Blackburn: both charges are totally mistaken; formal logic is too modest to deserve them.
- Coercion by result. Logic determines whether a set of propositions implies a contradiction, and treats contradictions as false. We may moralise against someone indifferent to holding falsehoods, but not as logicians: "the work of the formal logician was finished with the result".
- Coercion by bivalence. One may prefer degrees of truth for vague propositions, or a third status. These are respectable, and alternative logics develop them, but they become awkward and uncomfortable; it is usually wise to be grateful for two values.
- Coercion by uncharitable reading. Forcing someone's words into a form that makes them contradictory may ignore presuppositions and suppressed premises, but that is uncharitable interpretation, not logic. In Chekhov's "The Lady with the Pet Dog", Anna Sergeyevna's husband "believed her and did not believe her"; logic does not demand we jump up and down on Chekhov: the flat contradiction prompts us to look for the intended meaning (he half-believes, or alternates).
- Linear thinking. Logic directs nobody's thoughts, any more than mathematics tells you what to count; it is "gloriously indifferent" to whether propositions arrive by speculation, fancy or sober science. It tells you only whether they can all be true together, and that "can be a pearl beyond price".
Plausible Reasonings
Logic shows what follows from premises, but you have to have the premises. Much everyday reasoning aims to expand belief: it is meant to be plausible, not watertight; premises could be true and conclusion false, but not likely.
- A space of possibilities. Why is it silly to be confident that my roulette bet wins? Most possibilities left open by my evidence do not have x winning. Plausible reasoning corresponds to showing that most possibilities left open by the evidence make the conclusion true.
- Games of chance are little measured fields. Fifty-two equally likely outcomes from a fairly shuffled pack let us calculate, say, how many seven-card draws contain two court cards: measuring the proportion of possibilities with some outcome.
- The real world. Experience gives generalisations (grass is green, water refreshes, chocolate nourishes) that we project to the eighteenth century and tomorrow, outrunning the limited circle of what we observe.
David Hume puts the problem. Past experience gives direct information only of the objects and period that fell under it; why extend it to future times and other objects that may be only in appearance similar? Between "I have found that such an object has always been attended with such an effect" and "I foresee that other objects, similar in appearance, will be attended with similar effects" there is a step, an inference; and if it is made by a chain of reasoning, "I desire you to produce that reasoning". Inference from one region of space and time to others is inductive inference; Hume's worry is the problem of induction.
The Lottery for the Golden Harp
The lottery for the Golden Harp. You are disembodied spirits in Heaven; God (Blackburn) will embody you on Earth for a life of nine acts, then bring you back. Each of you holds a ticket predicting the colour of the clear midday sky in each act, from six colours; God covenants to change colour only at the start of an act. Exactly one ticket matches all nine acts, and its holder wins the Golden Harp: "Heaven is good, but Heaven with the Golden Harp is even better." One ticket, Wavy, goes blue, green, yellow, orange, back to blue and violet. Six are straight; one, Straightie, is blue throughout. Another, Kinkie, is blue for five acts, then yellow for the rest. In Heaven, no ticket has any better chance than any other: for all we know God favours waves, straight lines or kinks.
- 6^9 = 10,077,696 tickets are needed to cover every distribution of colours, so each starts with a 1 in 6^9 chance.
- 5/6 of surviving tickets are thrown away at the start of each act, as the sky shows its colour.
- 6^4 = 1,296 survivors remain after five blue acts, so Straightie's and Kinkie's chances have risen to 1 in 6^4.
- 6 tickets remain at the start of the ninth act; 1 winner a day later.
Late in the fifth act, Straightie offers to swap tickets with Kinkie only for a terrific price. Kinkie resists. We side with Straightie in our bones. But each can cite only the same track record: five hits. Neither can peer into the future. Straightie wants an argument for the uniformity of nature (God started blue and stuck to it, so will probably go on sticking), but Kinkie can say that God started as-per-Kinkie and by equal reasoning will stick with that.
- Any argument that the future will resemble the past must be either demonstrative (a priori) or from experience.
- It cannot be demonstrative: there is no contradiction in nature's changing, in many ways.
- An argument from experience must take past experience as representative of the future.
- Taking experience as representative presupposes the uniformity of nature, the very thing to be shown.
- So no argument shows that nature will probably stay uniform; we cannot even argue that change is improbable.
Hume grants that we all learn from experience and rely on uniformity, a natural propensity shared with animals. But that is all it is: "custom or habit", an exercise of nature with no special claim in reason. We engineer a bridge between past and future, but cannot argue that the bridge is reliable.
Goodman's grue: Kinkie as a predicate. Nelson Goodman's Fact, Fiction, and Forecast (1955) sharpened the problem into the "new riddle of induction". Define an object as grue if it is examined before time t and green, or not so examined and blue. Every emerald observed so far is both green and grue, so the evidence supports "all emeralds are grue" exactly as well as "all emeralds are green", yet the two predict different colours after t. Blackburn's Kinkie is a grue-ticket. Goodman's lesson is that induction needs a principled account of which predicates are projectible; his own answer was entrenchment (predicates with a history of successful projection), which critics find too close to Hume's custom. Kripke's reading of Wittgenstein ("quus", 1982) poses the same problem for meaning, which Chapter 7 meets as rule-following.
Every machine-learning model is a Straightie bet. A model trained on historical data is an inductive bridge, and the no-free-lunch theorems (David Wolpert, 1996) are Hume's point in mathematical form: averaged over all possible worlds, no learner beats any other; one generalises well only because its inductive bias matches the world. In banking this is practical: a credit or fraud model validated on five benign years is Straightie at the end of act five, and a regime change (a pandemic, a rate shock, a new fraud typology) is Kinkie's yellow sky. Hume offers no argument that will rescue the model; what is available is monitoring for drift, stress and scenario testing for kinked futures, and governance that treats past performance as habit, not proof.
Chancy Stuff
A problem most people get wrong, far from induction but a powerful tool for thinking.
The 99 per cent reliable test. A disease afflicts about 1 in 1,000 people. Your doctor's test is "over 99 per cent reliable", and you test positive. Most people think you are virtually certain to have the disease. Ask what reliability means: (1) if you have the disease, the test says so; (2) it gives false positives in about 1 per cent of healthy people. Then imagine 1,000 people tested: 1 has the disease and tests positive; about 10 of the other 999 test falsely positive. So 11 positives, 1 sick: your chance is about 1 in 11, a little under 10 per cent. Bad news (up from 1 in 1,000), but you are still far more probably healthy. Getting this wrong is the fallacy of ignoring the base rate.
Conditional Probability and Bayes
If 20 per cent of a class wear Levi jeans and 20 per cent wear Levi jackets, the chance that someone wears both is not 20% × 20% = 4 per cent, because the events need not be independent. If everyone with the jeans wears the jacket, the joint probability is 20 per cent; if fashion forbids the combination, zero. We need conditional probability, Prob(a/b), the probability of a given b:
- Product rule. Prob(a & b) = Prob(a) × Prob(b/a) = Prob(b) × Prob(a/b), equal by symmetry, since a & b is the same proposition as b & a.
- Bayes's theorem. Rearranging, as the English clergyman Thomas Bayes (1702 to 1761) did: Prob(b/a) = Prob(b) × Prob(a/b) / Prob(a).
- For hypothesis h and evidence e. Prob(h/e) = Prob(h) × Prob(e/h) / Prob(e), where Prob(h/e) is the posterior probability.
- Prob(h): the prior. How likely is the hypothesis from the word go?
- Prob(e/h): the fit. How well does the evidence accord with the hypothesis?
- Prob(e): the prior of the evidence. How likely is the evidence anyway? It grows with the number and probability of alternative explanations, and the larger it is, the smaller the posterior: the hypothesis has too many competitors.
Well, he would, wouldn't he?
Rice-Davies was pointing out that the denial was highly probable whichever hypothesis was true, so Prob(e) was high and the denial worthless as evidence. The ideal is a hypothesis antecedently quite likely, evidence just what it predicts, and few or no other probable ways for the evidence to arise. In the disease case: prior 1/1,000; fit = 1 (the test never misses); Prob(e) ≈ 1/1,000 + 10/1,000 (true plus false positives); posterior ≈ 1/11.
Bayes and the Golden Harp
Let S be Straightie's ninefold blue, and E the five blues observed. Prob(S) = 1/6^9; Prob(E/S) = 1; Prob(E) = 1/6^5. So Prob(S/E) = 1/6^4, matching the intuitive count. But exactly the same calculation gives Kinkie's K the same posterior: Prob(E/K) = 1, and Prob(E) is the same. The problem is not proving S will win, only finding some reason to expect S rather than K, and reason remains entirely silent. If there was no reason to prefer S in Heaven, a priori, there is none after the evidence comes in, or so it seems.
The three doors (a puzzle in the same family). A prize is behind one of three doors. You pick door 1. The host, who knows where the prize is and always opens a losing door you did not pick, opens door 3 to show a goat, and offers a switch to door 2. Most people say it is 50:50. Bayes says switch: Prob(prize behind 1) = 1/3, and the host's showing a goat was certain whether or not you chose right (he can always find one), so it leaves that 1/3 untouched; the remaining 2/3 now sits on door 2. Change the host's rule (he opens a door at random, and it happens to show a goat) and Prob(e/h) changes, so the answer becomes 1/2. What the puzzle shows is Blackburn's Mandy Rice-Davies point in reverse: what an observation is worth depends on the process that generated it, not merely on what it shows. (The puzzle is the Monty Hall problem, famous after Marilyn vos Savant's 1990 newspaper column.)
Living with Uncertain Priors and the Abuse of Significance
Bayes guards against ignoring the base rate and ignoring false positives, fallacies that crop up everywhere people try to think; it also illuminates zombies, design, evil and above all miracles. Priors are often hard to quantify, but two factors help:
- Robustness. Assign a range to each figure; all ways of calculating may give a similar result.
- Swamping. With enough evidence, investigators who began with very different priors may converge on a similarly high Prob(h/e).
Orthodox significance testing tries to bypass Bayes: if a result would arise by chance less than 5 or 1 per cent of the time, infer that something other than chance is at work. Blackburn calls this prevalent reasoning "highly doubtful", and Bayes shows why: if the prior that anything other than chance is operating is very low, even enormously improbable results will not overturn it.
The Scrabble bag. Shake a bag, throw seven Scrabble letters face down, shuffle them into a line and turn them up: PQAERTU. That exact result is very improbable; you might go a hundred years without repeating it. But it was chance. In this set-up every result is improbable, so improbability alone licenses no inference to design. The same mistake fuels lunatic attempts to show that the pattern of vowels in Shakespeare encodes the Name of the Beast 666 times. The lesson: "chance is just as good at throwing up improbabilities as design"; we need some further reason to think the improbable result is not just chance.
Bayes moved to the centre. Blackburn's warning proved prophetic. John Ioannidis's "Why Most Published Research Findings Are False" (2005) made his argument quantitative: with low prior odds of true effects, small samples and flexible analyses, most "significant" findings will be false; the replication crisis followed (the Open Science Collaboration, 2015, replicated only about 36 per cent of 100 psychology findings at conventional significance). The American Statistical Association's 2016 statement warned that a p-value does not measure the probability that a hypothesis is true. Bayesian epistemology is now a mainstream programme: degrees of belief as probabilities, Dutch book arguments for coherence, conditionalisation for updating, and convergence theorems that formalise Blackburn's "swamping". On base rates, Casscells, Schoenberger and Graboys (1978) gave Harvard Medical School staff a problem like Blackburn's (prevalence 1/1,000, false-positive rate 5 per cent): the most common answer was 95 per cent, the right one about 2 per cent. Gerd Gigerenzer showed that presenting such problems as natural frequencies ("of 1,000 people...") greatly improves performance, which is exactly Blackburn's method.
Explanations and Paradigms
Induction projects the observed onto the unobserved, but it is part of a wider effort to understand. In a complex system we notice how features vary together and find reliable relationships, such as Boyle's law: at constant temperature, the pressure of a given mass of gas is inversely proportional to its volume. This is purely empirical.
- Getting that far is hard. Economics would be delighted to identify the right variables and their reliable relations; most attempts crash in flames. Physics was the same: it took a century for scientists to identify the energy of a mechanical system as the salient conserved quantity, a fact "science teachers should be made to write out a hundred times" before calling children "dumb".
- A model in the weak sense. A story about the right variables and relations is a model, but may not give understanding. Isaac Newton (1642 to 1727) had the inverse square law of gravitation, yet he and his contemporaries felt it gave no understanding of why gravity works so. Why should pressure vary inversely with volume, and why does temperature matter?
- A model in the strong sense: a mechanism. The kinetic theory of gases sees a gas as molecules in motion; pressure is their impact on the container walls; higher temperature speeds them up. From the mechanism, Boyle's law can be derived.
- Mechanisms do not escape induction. The continued behaviour of the mechanism's parts is still a projection, but it reduces the number of independent assumptions: a few stable features explain the rest. These are the explanatory and simplifying ideals of science.
Hume Against A Priori Causation
What counts as a satisfactory mechanism? Nearly everyone feels some causation, like shunting, is especially intelligible, while action at a distance or body acting on mind is mysterious. Until Hume almost everyone, Newton included, thought we had a priori knowledge of what does and what could not cause what. Newton wrote that gravity acting at a distance through a vacuum without mediation is "so great an absurdity that I believe no man who has in philosophical matters a competent faculty of thinking, can ever fall into it". The same presumption operates when we try to show by pure reason that the universe needs a god.
Adam's first morning. Hume affirms "as a general proposition, which admits of no exception" that knowledge of cause and effect is never attained a priori but arises from finding objects constantly conjoined. Present a man of the strongest reason with an entirely new object, and no examination of its sensible qualities reveals its causes or effects. Adam, his faculties perfect from the first, could not have inferred from the fluidity and transparency of water that it would suffocate him, nor from the light and warmth of fire that it would consume him. Why do we feel otherwise about billiard balls? Because custom, where strongest, "not only covers our natural ignorance, but even conceals itself". The case shows that the sense of causal "insight" is habit mistaken for reason.
Hume's target is the ideal of "insight" into nature: a geometry or algebra of why events must fall out as they do, a Cartesian "clear and distinct" perception. When Newton published the Principia Mathematica (1687), some contemporaries were disappointed that he told them what gravity does, not what it is. Hume holds that this is all science can ever do: the utmost effort of reason is to reduce phenomena to a few general causes ("elasticity, gravity, cohesion of parts, communication of motion by impulse"), while the causes of those causes are "totally shut up from human curiosity". "The most perfect philosophy of the natural kind only staves off our ignorance a little longer."
Kuhn, Paradigms and Self-Correction
Without the rationalist ideal we have only more or less familiar systems, which serve as paradigms of satisfactory explanation; and if reason is replaced by habit and custom, customs can change. Thomas Kuhn (1922 to 1996) argued that they do: normal science works within paradigms, and revolutionary science challenges them, so that science is "a series of peaceful interludes punctuated by intellectually violent revolutions", after which our sense of a comfortable explanation changes.
The relativist reading. People have their paradigms, others have others, and there is no judging which is better; categories set by culture and history go all the way down.
Blackburn's reply. The inference is unwarranted. The sky as an opaque veil with holes through which we glimpse the heavens was once a paradigm, and we know better. Paradigms can be asked to show their worth; some fail. Even if we approach the world with culturally set categories, not all sets are equally good; some were discarded for good and sufficient reason.
Science's crowning glory. A scientific environment is ideally one where experimenting, predicting and testing weed out bad ideas, so only survivors go on. Actual science has its blinkers, prejudices and distortions, but the process contains within itself the mechanisms of correction, as the senses do (against Descartes, Chapter 1). Endeavours with no such devices, Blackburn names psychoanalysis, grand political theories, "new age" science and creationist science, need not interest us.
Inference to the best explanation, and Popper. The reasoning Blackburn describes, preferring a hypothesis because its mechanism explains the data, was named "inference to the best explanation" by Gilbert Harman (1965) and developed by Peter Lipton (Inference to the Best Explanation, 1991), who argued that "loveliness" (unification, mechanism, simplicity) guides us towards likely truth; Bayesians reply that such virtues matter only in so far as they raise priors or likelihoods. The test Blackburn applies to psychoanalysis is close to Karl Popper's falsifiability criterion (Logik der Forschung, 1934), which Adolf Grünbaum later disputed for Freud, arguing that psychoanalysis is testable and fails the tests. Kuhn's own later work (the 1969 Postscript and a 1977 essay on theory choice) also distanced him from the relativism Blackburn rejects, listing accuracy, consistency, scope, simplicity and fruitfulness as shared values for theory choice.
- Aristotle (384 to 322 BC): first taxonomy of valid forms.
- G. W. Leibniz (1646 to 1716): dreamed of settling disputes by calculation.
- Isaac Newton (1642 to 1727): the inverse square law; thought action at a distance absurd.
- Thomas Bayes (1702 to 1761): conditional probability and the theorem.
- David Hume (1711 to 1776): the problem of induction; causation known only from experience.
- Gottlob Frege (1848 to 1925): Begriffsschrift (1879), quantifiers and modern logic.
- Bertrand Russell (1872 to 1970): theory of definite descriptions.
- Thomas Kuhn (1922 to 1996): paradigms and scientific revolutions.
The World
Seventeenth-century science split the world in two: a scientific image of particles and forces with primary qualities, and a manifest image of colours, sounds and smells driven "back into the mind" by the argument from perceptual relativity. Blackburn shows how Berkeley and Hume turned the same reasoning on solidity, leaving no conception of what fills space; how Faraday's dissolution of things into powers, and the search for a "Straightjacket" behind laws of nature, end in the same bafflement as the cosmological argument; and how Kant's Copernican revolution recasts space, time, objects and causation as organising principles rather than inferred things. The cost is a pull towards idealism, seen in Berkeley's master argument, Moore's unobserved worlds and the ancient problem of universals, and Blackburn doubts anyone has yet found a stable place to stand.
Colours, Smells, Sounds, Feels and Tastes
Galileo Galilei (1564 to 1642), Descartes's great contemporary, wrote that he cannot conceive a material substance without bounds, shape, size, place, motion or rest, contact, and number; but nothing compels him to think of it as white or red, bitter or sweet, noisy or silent, sweet- or foul-smelling. These are "no more than mere names" as far as the object is concerned; they reside only in the consciousness, and if the living creature were removed, they would be wiped away and annihilated.
| Primary qualities | Secondary qualities | |
|---|---|---|
| Examples | Shape, size, position, motion or rest, number, solidity (Locke adds extension and figure) | Colours, tastes, sounds, odours, feels (warmth, cold) |
| Where they reside | In the objects themselves | "Only in the sensitive animal" (Galileo) |
| Do ideas resemble them? | Yes, on Locke's view | No: they are powers to cause sensations, which do not resemble their causes |
| Known by | Reason and science (Descartes's intellect) | The senses, which give only "confused" data |
| Image | Scientific (absolute) image | Manifest image |
Descartes held the senses to be poor vehicles of truth (recall the ball of wax, Chapter 1). Their proper purpose is to inform the mind of what is beneficial or harmful to the mind-body composite; we misuse them by treating them as "touchstones" of the essential nature of bodies.
The pain in the foot. When the foot is injured, "animal spirits" carry their "jet of flame" up the nerves to the brain, and the mind feels a pain located in the foot. God could have made the same brain motion convey something else, say a report of a perturbation in the brain, but then I would be slow to move my foot out of harm's way. Nothing else would be "so conducive to the continued well-being of the body". Two lessons follow. Descartes here denies epiphenomenalism: it matters which mental event occurs, because it makes us move the foot quickly; if the mental were inert, God could arrange it any way at all. And, substituting evolution for God, the point is surprisingly modern: a creature needs only information that stimulates the right action. A twig snapping under a predator's foot could produce a loud sound, a discord, a harmony, a high or low pitch, or just a bad smell; so long as it frightens the creature into flight, the senses are doing their job.
- God is off the hook. The senses deceive, but God made "the best system that could be devised" for preserving health; we are at fault if we take confused data for clear and distinct information about the real properties of things.
- What remains when intellect abstracts. For Descartes the mathematician, the real property of res extensa is spatial extension; colours and feels are a possibly illusory "filling" with their real residence in the mind.
- A second dualism. So beside mind and body, Descartes and his contemporaries open a dualism between the world as it is for us, the manifest image (coloured, warm, smelly, noisy, familiar), and the world as it is absolutely, the scientific image (nothing but particles and forces spread across boundless space).
The two images are Sellars's terms. The labels come from Wilfrid Sellars, "Philosophy and the Scientific Image of Man" (1962), which framed philosophy's task as achieving a "synoptic vision" that fuses the manifest image (persons, colours, reasons) with the scientific image (postulated micro-entities). Sellars gave the scientific image ontological priority (elsewhere, in 1956, he wrote that "science is the measure of all things, of what is that it is") while the manifest image remains indispensable for persons and norms: the same tension this chapter explores.
The Argument from Perceptual Relativity
Why does science drive colours into the mind? The most compelling argument is from perceptual relativity. Relativism is usually feared in ethics, but here a more general relativism appears:
- An object displays a certain smell (colour, sound, taste, feel) to observer O.
- How it smells is a function of O's particular sensory structures.
- So there is, or could be, a creature O* with different structures, to whom the same object would smell quite differently.
- O and O* may each live equally efficient, adapted lives.
- So there is no reason to say just one of O and O* has the smells right.
- So there is no one correct distribution of smells in the world.
- So smells are better thought of as entirely mind-dependent.
The argument was known in the ancient world and returned to prominence in the seventeenth century.
- O* need not exist. It is enough that we can see how there could be such a creature, with different colour receptivity, hearing tuned to other frequencies, or sensitivity to other energies altogether.
- Real cases make it graphic. Nobody who keeps a dog can think its world of smells resembles ours; beer or dry wine tastes different before we get used to it; "all mammals except some primates are colour-blind"; phenol thio-urea tastes strongly bitter to many people and of nothing to others. Some people tune their TV sets to colours others find garish, yet they see the same scenes.
- Premise 2 is incontestable. Colds blunt smell; colour vision depends on three kinds of receptor (Chapter 2); bats find their way by acoustic means closed to us.
Two factual corrections. Most mammals are not colour-blind; they are dichromats, with two cone types instead of the three of Old World monkeys, apes and humans, so they see a reduced range of hues (dogs distinguish blue from yellow but confuse red and green). Some, including many marine mammals, are close to monochromatic. The bitter substance is phenylthiocarbamide (PTC), discovered by Arthur Fox in 1931; roughly 70 per cent of people of European descent can taste it, a difference traced mainly to variants of the bitter-receptor gene TAS2R38.
The Hole and the Patch
Premise 4 does the real work, and is often forgotten in relativist arguments elsewhere, such as ethics. Without it the argument has a hole: different TV sets turn the same signal into different pictures, but a set showing a snowstorm while the transmitter beams the inaugural speech is not doing just as well in a different way; it misses information that is objectively there. Difference of output does not show that nothing is there independently of the receiver.
Premise 4 plugs the hole by making O and O equally well adapted: for Descartes a theological belief, for us an evolutionary one, since creatures that cannot get the information they need die out. Unlike televisions, O and O may be doing equally well with different sensory experiences, and then, as Russell put it, it would be "favouritism" to call one representation of the world better.
Dogs as authorities on smell. Dogs have marvellous noses: grant them authority on smells. But they are poor at colour, and we make finer visual discriminations in more lights, which is what our colour vision is for. So why not say the real colours are those the creatures best adapted for colour see, and the real smells those the best smellers sense? Then the relativist conclusions fail. The objection shows that "equal adaptation" overall is too weak. The repair is to run five separate arguments, one per sensory dimension D (vision, touch, smell, sound, taste), each with the premise "O and O* may each live equally efficient, adapted lives in respect of dimension D". Then the rest is plain sailing: PTC cannot be both tasty and tasteless in itself, and a world containing as many smells as there are possible detectors would contain infinitely many coexisting smells.
The upshot is secondary quality idealism: Galileo's result that the immediate objects of sense are driven "back into the mind".
Good Solid Sense
This did not trouble Descartes, who still had reason to inform him of objects' real properties. His position was canonised in English philosophy by John Locke (1632 to 1704).
- Primary qualities. "Solidity, extension, figure, motion or rest, and number", which produce simple ideas in us.
- Secondary qualities. Qualities "which in truth are nothing in the objects themselves, but powers to produce various sensations in us by their primary qualities", that is, by the bulk, figure, texture and motion of their insensible parts: colours, sounds, tastes.
- Are objects coloured? In one sense yes: they have the powers to produce colours and smells in us. But the powers are not the colours and smells themselves, and the sensations do not resemble the powers that cause them.
Locke's view is often taken as natural, common-sense scientific realism: swap his little particles for energies, forces, fields or sub-atomic particles, and the opposition between the world of science and the manifest image remains much as he drew it. Is there any problem with that?
Berkeley's Problems
French writers had already worried about Descartes's picture. If God, for our own good, makes us see colours that resemble nothing in reality, might he not also make us see extension in space though reality is not spatial? Colours become a Trojan horse for the Evil Demon scepticism Descartes thought he had beaten.
The metaphysics is worse. Descartes left physical reality only extension, but that is entirely abstract: a cubic foot of space differs from a cubic foot of space with a body in it. We need things occupying space. Locke offers solidity, which distinguishes a cubic foot of granite from a cubic foot of vacuum: "the resistance which we find in body, to the entrance of any other body into the place it possesses" (he accepts "impenetrability" too, though "solidity" is more positive), the idea "most intimately connected with and essential to body". Solidity seems to come down to powers to exclude other objects, so must there not also be a substance to possess the powers? Asked what solidity is, Locke says: put a flint or a football between your hands and try to join them, and you will know.
The problem erupted early in the eighteenth century in Pierre Bayle (1647 to 1706) and George Berkeley (1685 to 1753), for whom the Cartesian/Lockean world-view holds together neither metaphysically nor epistemologically.
- (1) Solidity is on all fours with colour. If we know solidity only by what we feel, how is it better off than colour, felt heat or smell? If those sensations give no real idea of qualities in things, neither does the feel of resistance. Our ideas are not solid, so in what sense do they "resemble" solid things? And if solidity goes, what is left? Berkeley's notorious answer: nothing. The world retreats into the mind: subjective idealism.
- (2) We cannot abstract primary from secondary. Locke needs us to conceive a world in purely primary-quality terms. Berkeley: "I deny that I can abstract one from another, or conceive separately, those qualities which it is impossible should exist so separated."
The bleached tomato. Take a tomato and abstract out its colour, feel, smell, taste, and the sensations you get as you play your hands round it. What is left? An invisible, intangible, undetectable tomato, surely no better than no tomato at all. The thought experiment shows that our conception of a primary-quality object is parasitic on secondary qualities; strip them away and the object we were thinking of disappears with them.
Hume puts the objection superbly. The idea of solidity is of two objects that cannot penetrate each other however hard impelled; so it is incomprehensible without some conception of the bodies that are solid. Secondary qualities are excluded; motion depends on extension, and extension on solidity; so solidity cannot depend on them without running in a circle. "Our modern philosophy, therefore, leaves us no just nor satisfactory idea of solidity, nor consequently of matter." After excluding colours, sounds, heat and cold, "there remains nothing which can afford us a just and consistent idea of body".
Berkeley and Hume deny we can understand the alleged properties of an independent world except in terms drawn from our own minds. Yet "the modern philosophy" demands an "absolute" conception of space-occupying things whose arrangements explain everything, including us. If that conception is flawed at its heart, we must look elsewhere.
Forces, Fields and Things
Hume's question generalises: what conception do we ever have of bodies apart from their powers of interaction with each other and with us? Michael Faraday (1791 to 1867) answered by taking the objection on the chin. Distinguish a particle a from the powers m by which it makes itself known, and "the a or nucleus vanishes, and the substance consists of the powers": why assume "that of which we are ignorant, which we cannot conceive, and for which there is no philosophical necessity"?
- The world as a flux of forces. The scientific image then becomes gravitational and electromagnetic forces and, at higher magnification, strong and weak interactions.
- But the argument is general. The "particles" themselves dissolve into further forces, and that is uncomfortable, because we normally understand a field by its effect on a test particle: the field around a magnet on the table is how other things would move if placed near it, as iron filings show in school.
- A world of ifs. Without particles, the difference between occupied and empty space becomes the difference between space of which some ifs are true and space of which other ifs are true. We hanker after something to occupy space and explain the differences in ifs.
What must God create? Return to the question of Chapter 2: if God creates the physical universe, how much does he have to do? Can he create only forces? Then the universe is a giant set of ifs, a huge potential, "like a gigantic shimmer". Or must he also create objects, for the forces to act on and to explain how forces arise? Then what conception of those objects can we have? Common sense demands something solid filling the occupied bits of space; physics seems to offer only powers. The thought experiment shows the dilemma has no comfortable horn. (Whether Descartes, the mathematician, anticipated Faraday's vision is "a fascinating question".)
Kant, himself a pioneer of resolving matter into forces, thought this conception the best we could ever achieve, because the senses are receptive: they deliver only the results of powers and forces, never what underlies their distribution. Anything underlying it would be "noumenal", beyond science, experience and thought. Hume thought the impenetrability problem cast doubt on the whole modern philosophy, while finding Berkeley's idealism unbelievable; Kant too thought it required an entire rethink.
Physics went Faraday's way. In quantum field theory, the framework of the Standard Model, the fundamental entities are fields, and particles are quantised excitations of them: an electron is a ripple in the electron field, with no further "nucleus" behind its powers. Philosophers took up the theme. Structural realism (John Worrall, 1989) holds that what survives theory change is mathematical structure; its ontic version (James Ladyman, Every Thing Must Go, 2007, with Don Ross) says structure is all there is, Faraday's "the a vanishes" made into a metaphysics. Dispositional essentialists (Alexander Bird, Nature's Metaphysics, 2007) argue that fundamental properties just are powers, so a world of "ifs" is no embarrassment; critics press the regress Blackburn describes, since powers seem to need non-powers to be manifested in.
Straightjackets and Laws
A second route to the same problem runs through laws of nature and the Golden Harp (Chapter 6). That thought experiment, one might object, wrongly treats the sky's colour in each act as independent, as if God tossed a six-sided die each time. If that were so, it would be a fallacy to expect blue because blue came up five times: the gambler's fallacy. But the world shows no such chaos, only uniformities, so it is more probable that something guarantees order through time: one decision of God, stuck to. There must be a metaphysical solution to induction, even if there is no mathematical one.
- The single decision might be kinked. Even if the universe realises one law, it may have been "Let's have pattern K" rather than "Let's have pattern S". Inferring that nature will stay uniform in the particular way it has been is another inductive inference.
- What is a Straightjacket? Call whatever guarantees order a Straightjacket: a law operating over time, a directive that fixes the order of things, because of which things keep on keeping on. Can we conceive of one?
The Straightjacket is the cosmological argument again. Things in space and time, human resolutions included, come and go. If a Straightjacket could change, its own survival needs explanation, and we are launched on a regress. If its continuation is a brute fact, it is no more likely than the empirical order it was meant to underpin. If it is "necessary" or self-explaining, we meet Hume's scepticism about necessary existence: we do not know what we mean, nor to what such a description could apply. We can understand how events do fall out, but never obtain any glimmer of a conception of why they must fall out as they do.
Wittgenstein's "a nothing would serve just as well as a something about which nothing could be said" returns, and Hume: the scenes of the universe continually shift, "but the power or force, which actuates the whole machine, is entirely concealed from us". Following Faraday and Hume, the absolute conception ran into things versus powers; now our conception of those powers, underwritten by laws, is as frail as it could possibly be.
Laws of nature since 1999. Humeans (David Lewis's "best system", 1973 and 1994) say laws are the theorems of the simplest, most informative summary of what actually happens: no Straightjacket, only the mosaic. Necessitarians (Armstrong, 1983; Dretske and Tooley, 1977) posit a necessitation relation between universals, a Straightjacket, and face Blackburn's question of why necessitation guarantees regularity (van Fraassen's "inference problem", 1989). Dispositional essentialists derive laws from the natures of properties. Each claims to ease induction; none has convinced the others.
Kant's Revolution
Berkeley retreated into his own ideas and his nature as a spirit, sure that experiences must be injected by a greater spirit, God (an a priori causal inference at which one can anticipate Hume's snort of derision). Nobody has found this satisfactory: Berkeley's God plays the role of Descartes's Evil Demon, installing us in a wholly delusive virtual reality.
Kant agreed with Berkeley's diagnosis. Locke's "modern philosophy" was a "transcendental realism": realism because it posits independent objects in space and time; transcendental because that world lies outside experience, known only by inference. And the inference is precarious: an effect may have more than one cause, so it always remains doubtful whether my outer perceptions are "a mere play of our inner sense" or caused by external objects, while "I myself with all my representations" am immediately perceived.
- Escaping the inner theatre. Kant's priority is to get away from this "inner theatre" model. As Chapter 4 showed, self-consciousness needs experience organised not as a "rhapsody" or kaleidoscope but in spatial and temporal order, giving a concept of ourselves moving among independent objects.
- Concepts are rules, not pictures. For Locke (and Berkeley) understanding meant having something in the mind that resembles its cause, like a picture. Berkeley drew the moral that "an idea can resemble nothing but another idea", so we understand only our own ideas. Kant saw that for space, time, size, shape and objective order, a concept is an organising principle or rule, a way of handling the flux of data. The same rules could give the same understanding despite differences of subjective experience.
- Categories, not inferred things. The troubles of the previous sections came from looking for things to play roles: objects behind powers, something responsible for laws. Instead treat causation and law, things in space and time, and space and time themselves as necessary categories of thought: a framework for organising experience, not items inferred from it. This is the other side of the coin from Kant's treatment of the self.
Hitherto it has been assumed that all our knowledge must conform to objects... We must therefore make trial whether we may not have more success in the tasks of metaphysics, if we suppose that objects must conform to our knowledge.
Kant compares this to Copernicus, who, failing to explain the heavens on the supposition that they revolve round the spectator, tried making the spectator revolve and the stars stand still. If intuition must conform to objects, a priori knowledge of them is inexplicable; if objects as objects of the senses conform to our faculty of intuition, it is possible. This is transcendental idealism, which Kant insists is not Berkeley's subjective idealism and plainly not Locke's transcendental realism.
- We do not create the world. It is the universe that created us after some thirteen billion years, and Kant does not deny it. His aim is to explain how concepts like things, forces, space, time and causation shape the way we think, and have to think, about the world, structuring the "phenomenal world" that science describes and sense experience manifests.
- Space is not like colour. Colours are "not properties of the bodies", only "modifications of the sense of sight" affected by light; space, by contrast, is a condition under which alone objects can be objects of the senses. Space has more objectivity going for it than colour.
What Berkeley could thank Kant for. 1. Concepts are rules, principles and organising structures, not mental images. 2. Experience must be orderly (in Jonathan Bennett's phrase, there must be a "speed limit") for us to be self-conscious at all. 3. Orderliness requires thinking of ourselves at a standpoint in space, perceiving enduring objects whose behaviour follows laws of nature.
All grist to Berkeley's mill: he too knew we interpret experience spatio-temporally and objectively, but said we must "speak with the vulgar but think with the learned", treating that interpretation as a façon de parler.
What separates Kant from Berkeley. Kant sometimes sounds like a subjective idealist: matter, in all its configurations, is "nothing but mere appearances, that is, representations in us". But he thinks Descartes and his successors got hold of the wrong end of the stick in treating inner experience as rock-solid and the outer world as problematic. His proof must show "that even our inner experience, which for Descartes is indubitable, is possible only on the assumption of outer experience", experience in which we are immediately conscious of a reality beyond us. Whether he succeeds is one of the great issues of modern thought.
Two readings of Kant, and the age of the universe. The universe is now dated at about 13.8 billion years (Planck satellite, 2013 to 2018). Interpreters still divide over whether transcendental idealism posits two worlds (appearances versus unknowable things in themselves, which makes Kant look like Berkeley) or two aspects of one world (Henry Allison, 1983). The "Refutation of Idealism" (second edition, 1787) fathered modern transcendental arguments against scepticism (Peter Strawson, 1959 and 1966), which Barry Stroud (1968) argued show at most how we must think, not how things must be.
The Eye of the Beholder
A realist contends for facts entirely independent of the mind; the idealist reminds us of the mind's work in selecting and moulding our conception of the world, the Lebenswelt of thought and perception. Kant is in this up to the elbows: the whole conceptual scheme of space, time, objects, causes and selves comes from organising principles of the mind. Twentieth-century thinkers ran with Kant's ball more enthusiastically than he did, celebrating "paradigms": culturally and historically mutable lenses, conceptual palaces or prisons of our own engineering.
Berkeley's Master Argument
In the first of the Three Dialogues (1713), Philonous (Berkeley) stakes everything on one issue: if Hylas can conceive any sensible object existing without the mind, Philonous will grant that it does.
- Hylas conceives a tree or house existing unperceived by any mind.
- But one cannot see a thing that is at the same time unseen; equally, one cannot conceive a thing that is unconceived.
- The tree Hylas thinks of is conceived by him.
- What is conceived is in the mind.
- So Hylas has not conceived a tree existing out of all minds; he has only framed an idea in his own mind.
- So he cannot conceive how any sensible thing should exist otherwise than in a mind, and yet contends for the truth of what he cannot conceive.
Hylas confesses "an oversight" (thinking of a tree in a solitary place, he forgot that he himself conceived it all the while), but "some scruples remain". He is probably right to keep them.
- An ambiguity in "in the mind". We can turn on Philonous the weapons used against the ontological argument: perhaps he is misunderstanding phrases like "in the mind", sliding from the conceiving's being mine to the thing conceived's being mental.
- It proves too much. The argument would equally show that Hylas cannot conceive a tree existing otherwise than in Hylas's own mind, which is too radical even for Berkeley.
- The grain of truth. Read Hylas as trying to abstract away from his own perceptual and conceptual contingencies, to grasp an object independent of his modes of comprehension. Philonous reminds him the feat is impossible: whatever he conceives, he brings his own perspective to it. Imagine the tree with a brown trunk and green leaves and you have already used artefacts of the human perspective.
The room with a mirror. Imagine a room with a mirror on one wall and, some way in front of it, a table with a bowl of flowers. Do not imagine yourself in the room. You think you can. Now: is the bowl of flowers in the mirror? Say yes, and you are secretly occupying one perspective; say no, and you occupy another, since the flowers show in the mirror from some angles and not others. You can hardly say "neither", and "they come and go" just means moving your viewpoint around the room. You are trapped: the point of view comes in all unbidden as soon as you exercise your imagination. This is Berkeley's grain of truth in its most persuasive form.
Moore's Unobserved Worlds
G. E. Moore ignored this lesson when he set out to refute the idea that beauty lies in the eye of the beholder.
Moore's two worlds. Imagine one world full of fluffy clouds, green trees, running streams and pastoral delights, and another that is a heap of cinders and garbage. Stipulate that nobody is in either: both are unobserved. Surely one is more beautiful than the other, so beauty is independent of any beholder. Philonous inoculates us against this. It is we who accept the invitation to think of these worlds, and we bring our own aesthetic responses, a love of countryside and a dislike of garbage, to them; our verdicts voice those responses. All Moore shows is that we can deem things beautiful whether or not we think they are seen, which does not refute the view that beauty lies in the eye of the beholder.
Moore's target was value, not only beauty. In Principia Ethica (1903), section 50, Moore used the two worlds against Henry Sidgwick's view that only states of consciousness have intrinsic value: the point was that the beautiful world would be better (a thing worth producing) even if never contemplated. So it is an argument about intrinsic goodness, though Blackburn's diagnosis (that we cannot imagine the worlds without bringing our responses with us) applies just as well. In the book's final chapter Moore himself ranks the enjoyment of beautiful objects among the greatest goods and concedes that beauty unappreciated has comparatively little value.
Most people find Moore fishy, and so far side with Philonous; but most find Philonous fishy too. To reject both we need a stable place to stand. Perhaps Kant points the way: concede beauty and secondary qualities to the eye of the beholder, but hold that the beholder's eye is less involved with space, time, physical objects, the self and causation. Even that synthesis runs into choppy water.
Rules, Universals
The idealist tradition stresses the shape of our own minds in "constructing" the world as we understand it. Berkeley and the empiricists start from the subjectivity of sense experience; today cultural and linguistic variation looms larger, and many philosophers applaud Wittgenstein's rule-following considerations.
- The moment of understanding. When a concept is explained, we realise "now I can go on": we seem to grasp a rule separating correct from incorrect application. This is a real feat: some people and animals fail it (dogs look at your finger instead of where you point).
- A fact of natural history. That we perform the feat, and sufficiently often in the same way, makes communication and shared understanding possible. But it is not a given that we all do it the same way; it requires minds shaped the same way. What shapes them?
This is the ancient problem of universals: what is the reality behind the rules governing concept application?
| Position | What rules are | Applied to "hysteria" | Charge against it |
|---|---|---|---|
| Realism (Platonism) | Real, objective entities fixing correct application over past, present, future and possible cases, grasped by a mysterious act of apprehension | A real phenomenon with real boundaries; the term "carves nature at the joints" | Incredible: the apprehension cannot be understood in natural terms |
| Conceptualism | Creatures of the mind, conjured by shared responses from shared (perhaps culturally educated) natures; all concepts are "response-dependent" | A useful category grouping cases that strike us as similar, which is all any concept does | To the realist, the pit of idealism; to the nominalist, "a nominalist who is too cowardly to admit it" |
| Nominalism | No rules at all: just people with dispositions to apply or withhold words; no real correctness, though deviants get called "incorrect" | As good as any other word: people are disposed to use it | Eliminativism about meaning and thought: no mind is as good as a mind that blurts verbal responses |
- The stakes. Medieval Paris students reputedly came to blows in the streets over universals; at issue is our ability to describe the world truly or falsely, arguably the deepest problem of philosophy. It structures Plato's realist reaction to the sophists' nominalism.
- Postmodernism as nominalism. What is called "postmodernism" is "really just nominalism, colourfully presented as the doctrine that there is nothing except texts", paralysing appeals to reason and truth in many humanities; analytical philosophy plays Plato to its sophistry.
- Naturalised realism. A recent view avoids Platonic mystery: there really are properties things have independently of us, and evolution and success shape our minds to respond to the real causal kinds. We "go on in the same way" because, in the context of the world, that is the right way to go on. It would show even secondary-quality classification (Chapter 2's colour science) is far from arbitrary, and ties in with the natural foundationalism of Chapter 1.
Not everyone is convinced. Naturalised realism is comfortable, but it requires confidence that the scientific image sits easily beside the manifest image: that Kant or a successor has steered between Philonous and Moore, or solved Hume's problem with "the modern philosophy" without conceding too much to idealism. Blackburn leaves that open.
Rule-following, natural kinds and machine representations. Saul Kripke's Wittgenstein on Rules and Private Language (1982) turned the rule-following considerations into a sceptical paradox ("quus": nothing in my past use fixes that I meant plus rather than a function that diverges for unseen cases), with a "sceptical solution" grounded in community agreement, close to Blackburn's conceptualism. David Lewis (1983) answered with natural properties that are eligible referents, a version of the naturalised realism Blackburn describes. Machine learning has made the debate empirical. Systems trained on different data and objectives often learn similar internal representations, and the "Platonic representation hypothesis" (Huh, Cheung, Wang and Isola, 2024) proposes that sufficiently capable models converge on a shared statistical model of reality, which would support the naturalised realist's claim that the world's causal structure, not our quirks, fixes how to "go on". The hypothesis is contested, and convergence might equally reflect shared human-generated training data, which is the conceptualist's reply.
Primary and secondary qualities in an AI pipeline. A model's embedding of a customer or transaction is a secondary quality, one detector's response; another model, equally well adapted, will represent it differently, and neither is "the real one". Treat as primary the governed facts (identifiers, amounts, dates, legal entities, provenance). Letting one model's latent representation become the system of record mistakes a detector's response for a property of the thing.
What to Do
Blackburn's last chapter offers building blocks for practical thought rather than a moral theory. Its base is the idea of a concern: being moved by a thought, which is not the same as wanting, and which psychological egoism fatally misdescribes. Ethics is the set of concerns we expect of each other, internalised as a voice within. Reading the equation "having a concern = thinking something good" from left to right, he sides with non-cognitivism: there is no mysterious normative order, but concerns can still be better or worse by the lives they make. Practical argument is legitimate when it respects the other's point of view, aims at coherence, objectivity and imagination, and resists relativism's insinuating "just": our values are ours, as Neurath's boat is, and no one is in dry dock.
Real Concerns
Much practical thinking is technological: the end is set (fix the refrigerator, build a bridge) and the problem is means, needing understanding of the system, with no single "way of thinking" for all goals.
It is often said our ends are fixed by our desires. That is an approximation that misleads, if desires are states of enthusiasm that "put a gleam in our eye". I cut the grass when I would rather be sailing, not because I desire to (perhaps I hate it) but because it had to be done. I have a concern: I am moved by the thought that the grass needs cutting. My role as householder makes me sensitive to that thought about my grass, though I recognise that others' grass needs cutting without being moved to cut it.
"You must want to, otherwise you wouldn't be doing it." Annie feels bound to leave Bertie, perhaps from a duty to others or a life plan requiring a move. Bertie raises the emotional temperature by insisting she must want to leave. The words hurt because they say leaving puts a gleam in Annie's eye, and that may be entirely unwarranted: Annie may be completely dejected. Like cutting the grass, it has to be done. Confusing concern with desire is a standing cause of strife.
The Egoism Argument
A further, very dangerous mistake says that whenever we act from a concern, what we "really" desire is some state of ourselves, such as peace of mind. Psychologists have pictured desire as a build-up of tension the agent is driven to release, so that release seems the real object all along. Hence more hurtful words: "You weren't really concerned about the starving children, you were just wanting to feel good."
- Every action is motivated by the agent's desire.
- Satisfying a desire relieves the agent's tension or brings her peace of mind.
- So what every agent really wants is that relief, a state of herself.
- So all behaviour is fundamentally selfish; other aims are masks.
This set of thoughts, psychological egoism, is "entirely wrong": the step from (2) to (3) confuses a result of satisfying a desire with its object.
The punch in the stomach. You want food. Reasoning as the egoist does, I take you to want relief from the tension of wanting food, so I punch you in the stomach until you are too sick to want food. Did I get you what you wanted? Not at all: you wanted food, not any old relief. A normal person aroused by sexual passion does not want a bromide, which would also end the tension: he wants sex. The case shows that desires are directed at the world, and that the removal of the desire is not what the desire is for.
The godfather's fool's paradise. A Mafia godfather, insulted by Luigi, orders his henchmen to rub Luigi out. Daunted, they reflect that what he really wants is relief from the tension Luigi's existence causes, and deliver it more cheaply: a completely convincing delusion that Luigi is dead. Did they do what he wanted? Clearly not; he did not want to live in a fool's paradise (imagine his reaction on finding out). He wanted Luigi dead. Blackburn's lesson: one of our concerns is not to be deceived about whether our concerns are met.
The same argument, old and new. Joseph Butler (1726) made the point first: pleasure presupposes a desire for something other than pleasure. Robert Nozick's "experience machine" (1974) generalised the godfather: offered any experiences you like, indistinguishable from reality, most people decline, wanting to do and be, not merely to experience. In AI the egoist's mistake is "wireheading" or reward tampering: an agent optimising its reward signal rather than the world the signal tracks behaves like the henchmen. The alignment fix is Blackburn's: objectives about the world, plus a concern not to be deceived about whether they are met.
Objectifying Concerns, and Identifying with Them
Communication fails when one side treats the other's concern as itself the problem, to be managed or cured. If Annie cares about her career and Bertie replies "Don't get upset, darling, you won't worry about that if we go out to dinner / have a baby...", the response is inappropriate just as the punch was, though less obviously to Bertie. In Chapter 3's terms Bertie has "objectified" her concern: from her perspective the problem is her career, not her concern with it.
- We often do not want to lose our concerns. Tell the godfather his desire for revenge will pass in ten years, or tell Annie her ambition will fade after a baby: neither wants that. They identify with these concerns.
- Sometimes we do want rid of them. The smoker wants the cigarette and also to be rid of the craving by any means; therapy or a kind of surgical removal would do. Calling something "a craving" or "an obsession" is a way of distancing yourself from it, beginning to objectify it en route to escape.
- Frankfurt's second-order desires (Chapter 3). The willing and unwilling addict give this distinction its classic form.
- The line is not always evident. Annie might come to share Bertie's view that her ambition is the problem, and might be wrong, since the concern may be more central to her identity than she was led to think. We may not know until we try whether a concern is one to shake off or one to meet.
What Concerns Are
To have a concern is to be moved by a thought: some aspect of things engages our motivation, weighs with us, matters to us (the natural metaphors are of weight and pressure). Aspects weigh with us in choosing, and also in attitudes (admiration, contempt) and emotions (fear, hope), as when a novel's villain repels and its hero's virtues move us.
- Reasons. The aspects that weigh with us are our reasons (the grass needed it); in a wider sense they include features affecting us unawares: Annie's reason might be that Bertie bores her.
- Descriptive versus normative. "Annie had no reason to leave" may be a (probably false) claim that she acted on pure impulse, or, more likely, a rejection of her actual reason: she went to pursue ballet, and that was a silly ambition that should not have weighed with her. Normative talk should be marked by "ought" and "good"; saying "no reason at all" for "no good reason" invites misreading.
- Concerns are a mixed bunch. An enemy's death, a career, the grass, family and friends, a promise given, a past kindness, being a spouse or doctor or lawyer. Egoism's attempt to reduce them to concern with our own states failed and rides roughshod over the identify/objectify distinction.
The Voice Within
Many concerns are private and optional: my love of steam engines need not weigh with you (fewer crowds, I may be glad). But some concerns we expect others to have: that an act would be deceiving someone, breaking a promise, behaving dishonestly or manipulatively, or causing distress. We would be surprised or shocked if these did not weigh. This is the traditional domain of ethics: what concerns can we expect of each other?
Question 1: the ideal life. What concerns make up the way to live? The Homeric hero prizes honour, status and success in battle; the Christian saint the love of God, suppression of pride and brotherly love; Confucianism a large dose of respect for traditional ways. But there is little reason to think there is any such thing as the ideal life; any ideal will be heavily contextualised. The uncontroversial core of a good life: health, happiness (not a fool's paradise), achievement (not vain ambition), dignity, friendship, love, family; wealth and leisure are disputed, and even core goods can curse: a man might have lived better without such rude health that he could not sympathise with frailty.
Question 2: the boundaries of conduct. What may we demand of each other, so that failing is doing wrong, falling short and becoming a target of reproach? People expect each other to be honest, cooperative, sensitive to needs, fair, well-meaning; failure gives others a complaint against us. This reading is more rigid, and it is where obligation, duty and right live.
Shrugging Off Censure
Why not be a free spirit, careless of opinion? The visionary, saint or hero sometimes must be. But if we attract censure for not caring a jot about promises or other people's money, shrugging it off needs no recognition that others have a concern at all.
- Two grades of villainy. The common-or-garden villain: "I don't care if I have wronged you." The extraordinary villain: "I don't even recognise that you have a complaint", easier as a pose than to live in, though in sexual morality people sometimes genuinely cannot see the complaint, making things worse.
- A society without the second-order recognition. If nobody could recognise anyone as having a complaint, whatever they did, there would be no ethic, and for that reason it would be hard to recognise it as a society at all.
Internalising values is close to internalising the gaze or voice of others: recognising their complaint means seeing yourself as fallen short in their eyes and finding that itself weighing with you, as self-reproach, shame and guilt. Hence the Golden Rule near the core of most ethical systems ("do unto others as you would have them do unto you"), the stress some put on a "common point of view", and others on sympathy or empathy. Hume's example shows how naturally we see ourselves from outside: "A man will be mortified if you tell him he has a stinking breath; though it is evidently no annoyance to himself."
The piano on your foot. There is a piano on your foot; your concern is to move it quickly. If I see this I will naturally share the concern, and would be falling short if I did not; what is a reason for you to act becomes a reason for me to help. Some philosophers say the indifferent bystander is defective in reason. Blackburn disagrees. He may be a psychopath unable to grasp others' reality, or may miscalculate that suffering is good for you; but in the usual case, where he averts his gaze or passes by on the other side, nothing need be wrong with his understanding or his reasoning. "He is cold-hearted, not wrong-headed." That is just as bad, or worse; but placing the defect correctly shows that the cure is an education of the sentiments, not extra insight into the structure of reasons.
Truth and Goodness
One issue divides thinkers into two camps. Consider the equation:
One of X's concerns is to aim for / promote / endorse ϕ = X thinks ϕ is good / thinks ϕ is a reason for action.
- Read left to right (non-cognitivism). Concerns come first; talk of goodness and reasons reflects a motivational state, the fact of something weighing with you. That state is not a belief or representation of a fact but a reaction to the facts, so not strictly true or false, any more than a desire for coffee. St Augustine: "there is the pull of the will and of love, wherein appears the worth of everything to be sought, or to be avoided".
- Read right to left (cognitivism). At the foundation is a belief that ϕ is a reason, a special belief that carries concern with it. The slogan, often attributed to Aristotle: to desire something is to see it as good, as if desire answers to a perceived truth.
Cognitivists fear that without truth practical reasoning holds only "mere" concerns, desires and attitudes; bring it under the control of Truth, and the concerns that match the truths deserve authority. Blackburn holds that "the advantage is definitely on one side: the non-cognitivist side", because something besides cognition is bound to be in the mix, the pull of the will and of love, whose weight is measured in motivational strength. Can "seeing that ϕ is a reason" have that weight? Take the options for what is seen:
- A natural fact. "Seeing that the piano on your foot is a reason to remove it" might be "seeing that it is causing you pain". But whether that weighs with the agent is contingent: if she is cold-hearted, an enemy, or has too robust a sense of humour, it may not. So it is not equivalent to having the concern, which weighs by definition. G. E. Moore put it: for any natural features, it is always an "open question" whether they are reasons; taking them so is a further step, the step into practice.
- A non-natural normative fact. Moore's own view, perhaps Plato's: a glimpse of the normative order beyond empirical features. Very mysterious. Model it on a law saying pianos are to be taken off feet: it is always up to us what to feel about a law; I can ignore or reject it. The same holds if the law reads "God's law": if I don't care, the traditional weapon is fear of God's wrath, a contingent emotion the cognitivist cannot use, since she wants what is discerned to be necessarily motivating, "necessarily magnetic".
The "panic", and the debate since. The panicked position is motivational externalism (David Brink, Peter Railton), whose defenders reply that they never aimed to make the wicked irrational, only to make moral claims true or false. The strongest objection to Blackburn's side is the Frege-Geach problem (Geach, 1965): in "if lying is wrong, getting your brother to lie is wrong" nothing is endorsed, yet "wrong" means the same, so its meaning cannot simply be an attitude. Blackburn's quasi-realism (Spreading the Mind, 1984; Ruling Passions, 1998) is his programme for answering it. Realism has since revived (Parfit, On What Matters, 2011; Scanlon, 2014). And Moore's open question argument (1903) concerned the meaning of "good", not motivation.
Blackburn's View: We Invent the Normative Propositions
Read left to right, the problems disappear. People with concerns express them by talking of reasons and calling what weighs with them good. We invent the normative propositions ("This is good"; "That is a reason for action"; "You ought to do this") in order to think about the concerns we demand of ourselves and others, and there is no mysterious normative order into which we are plugged. We use them to clarify our motivational states and lay them out for admiration, criticism and improvement.
Better and worse without a normative order. Is no set of concerns better than any other? Certainly some are, but their superiority lies not in conformity to an independent order but in the ways of life embodying them. Concerns making lives loyal, friendly, grateful, prudent, sympathetic, fair beat those making lives treacherous, suspicious, malicious, careless, hard-hearted, unjust. Our lives go better described the first way, and we should be concerned that lives go better.
Good Bad Feelings
Many ethics books begin with a dualism: below, "the seething mass of desire"; above, the lordly principles of ethics controlling it. Blackburn thinks this breeds nothing but confusion, making ethical principles utterly mysterious, needing a divine origin or Platonic resonance with the Nature of Things. He substitutes a plurality of concerns, among which some have the status that leads us to speak of virtue and vice, duty and obligation: the concerns we expect of each other, that people owe each other.
- Owed sentiments. If someone does me a great kindness, I owe him gratitude: it is his due and my duty. If I am callous I fall short, forfeit others' admiration and, if I have an inner voice echoing theirs, feel bad about myself.
- The second fault. If I don't feel bad, that itself draws censure, sometimes more than the original failing: overlooking a debt of gratitude is bad, shrugging it off when it is pointed out can be more shocking. Hence the importance of contrition and, in serious cases, repentance. These bad feelings are good.
The therapist's pill. Annie knows she has hurt or insulted Bertie. A therapist offers a neat process to dissolve her self-reproach. Should she be grateful? Her concern is to put things right with Bertie, or with her own conduct, not with those concerns themselves. The pill puts nothing right and improves nothing; it makes her the person who attracts extra censure for lacking awareness of what she did, alienating her from her own knowledge and regret. This is Blackburn's charge against the contemporary obsession with "therapy". The exception is real but not typical: self-reproach can fester, doing no good, an obsession she might rightly wish away. That is when things have got out of hand; when they are in hand, the problem is not the guilt but the actions that invited it.
Practical Reasoning
Our concerns weigh with us (that is a tautology), but their weights can change, among other things through discussion and argument. Much practical reasoning aims not at means but at altering people's aims.
- Sheer persuasion. The arts of the salesman and the advertising agency: rhetoric that excites emotions and channels them. The preacher painting hell and the politician painting his party seek not understanding but to manipulate, attaching emotional weights, at the lowest level penalties and threats.
- Treating people as mere means. If I will use any course to get you to buy, vote or join, I treat you as what Kant called "mere means" to my ends, diverting you, by deception among other arts, as I would any obstacle.
- The respectful alternative. If I had a magic injection that would change you as I wish, then, unlike the salesman or preacher, I would not give it you: that fails to respect your point of view, or you as a person. I want you to come to share my understanding in the right way, which rules out deception and manipulation and rules in improved understanding.
Must reason then be confined to pointing out facts? Suppose we understand the situation identically, and my concerns issue in one aim; how can you change me except by manipulation, since you are wishing one of my concerns away? Blackburn finds two gaps in this argument.
- Uncovering hidden concerns. Our concerns are not always evident to ourselves. "Turning things over" is prowling round for unconsidered aspects that engage our motivations, and for unrecognised forces: caring more or less than we admit. We can be blind to our own natures. A conversation that uncovers suppressed or discounted motivations is cooperative, not manipulative.
- Putting my cards on the table. Even when you understand yourself, I may lay out my own take: your course does not reflect the duty of gratitude or loyalty you owe a third party. No manipulation or deception is involved. If you respect me, my good opinion matters to you.
My disapproval must not figure merely as a cost in your cost-benefit analysis; that would objectify it. In cooperative moral discussion it is on the table for you to share or undermine, aiming at common approval and disapproval. We then ask whether my insistence on gratitude deserves respect or is a fetish, by asking how well or badly the world would go without that concern, or relating it to friendship or honesty. Beneath the method lies another concern: that our practical stances be coherent, and also imaginative and objective.
Coherence, Objectivity, Imagination
Some writers deny practical life is a matter of rules or principles; it may be like aesthetics, where we judge a painting without articulable principles, or like recognising grammaticality (Chapter 1). Blackburn replies that practical reasoning in general is not like that, because we need to know where we stand.
- The legal analogy. A legal system that refused general principles and "treated each case on its merits", with no advance way of knowing what counts as the merits, could not regulate lives: it would be no law at all. Desires can be as fickle as we please, but the concerns we exact from each other must form a defensible system.
- Practical consistency. Logic (Chapter 6) prizes the requirement that beliefs can all be true; the practical equivalent is that our values can all be implemented. A law forbidding alcohol on Sunday while mandating attendance at mass, which includes wine, is inconsistent: impossible to obey.
- Real conflicts. "Always tell the truth; never hurt anyone", but the truth here hurts. "Always respect property; never endanger the State", but protecting the State needs requisitioning property. Much practical thinking adjusts simple obligations towards a comprehensive, consistent system, with results that are tentative, provisional and hostage to new cases.
- Device one: history. Inherited solutions have survived some test of time. Beware the conservative worship of inherited forms associated with Edmund Burke (1729 to 1797), but it is "much less intelligent" to think the test of time shows nothing; at least it gives a datum point for thinking about change.
- Device two: imagination. Fiction and imagination let us think through values "more or less for free", before crises and from a relatively objective view, seeing ourselves as others see us, so we learn our ideologies and our disguises.
| Theory | Core idea | Classic figures | Where it touches Blackburn's sketch |
|---|---|---|---|
| Humean sentimentalism / expressivism | Morality rests on sentiment; moral claims express attitudes | Hume; Blackburn, Gibbard | His own position: concerns first, value-talk second; the "common point of view"; the stinking breath |
| Kantian ethics | Act only on maxims you could will as universal law; treat humanity never merely as a means | Kant (1785) | The ban on manipulation and the magic injection; respect for persons |
| Utilitarianism | Right acts produce the best overall consequences for well-being | Bentham, J. S. Mill, Sidgwick, Singer | Testing a concern by how well or badly the world would go without it |
| Contractualism | Wrong acts are those permitted by principles others could reasonably reject | Hobbes, Rawls (1971), Scanlon (1998) | Ethics as what we owe each other; wrongdoing as giving others a complaint |
| Virtue ethics | The right act is what a virtuous person would do; aim at flourishing | Aristotle; Anscombe, Foot, Hursthouse | Judging concerns by the lives they make: loyal, grateful, fair versus treacherous, malicious |
What the chapter leaves out. Blackburn offers building blocks, not a survey, and his picture is one contested position, his own expressivism. Two moves echo rivals: testing concerns by how well lives go is broadly consequentialist, and morality as concerns we can demand of one another is close to Scanlon's What We Owe to Each Other (1998). Nor does he treat conflict between coherent but opposed concerns, where game theory enters.
The prisoner's dilemma. Two suspects are held apart. Both silent: one year each. Both confess: five years each. One confesses: he goes free, the other gets ten. Whatever the other does, confessing is better for each, so two rational egoists serve five years, though both prefer one. (The game is due to Flood and Dresher at RAND, 1950, named by Albert Tucker.) A concern for keeping faith, which each would shed in the single case, makes everyone's life go better by changing the payoffs from within: Blackburn's test of a concern "by how the world would go without it" made precise. His driving example is different, a coordination game, where any convention serves and nobody gains by defecting. Many moral concerns solve dilemmas, not mere coordination, which is why they are not "just us" as the side of the road is. In repeated play, Axelrod's tournaments (1984) found tit-for-tat strikingly successful.
Relativism
So in the end is it "just us": our moral imperatives a contingent, situated, variable set of concerns we happen to exact from each other? It is indeed us, but it may not be "just" us. The "just" insinuates that other solutions are equally good.
- Where "just us" is right. The British drive on the left, Americans on the right: equally good solutions to coordinating traffic, so the side is "just us". But it is not just us that we drive exclusively on one side: driving at random or in the middle is no solution at all.
- Many harmless differences. Languages, grammars, customs, rites and social arrangements can be equally good solutions to problems of expression, coordination and communication. We need not rank them: when in Rome, do as the Romans do.
- Where it is wrong. A society that, like the Taliban in Afghanistan, denies education to women; inherited caste systems denying whole classes equal health, education or even sustenance by birth; systems where some people are owned body and soul by others. These are "solutions" to how to live, but we need not see them as "just different" or tolerable at all: they trespass against boundaries of concern and respect we think must be protected. Hence the language of rights: if people do not show concern and respect, the injured may rightly feel resentment and call on the world to rectify their state.
In saying all this we voice our own sympathies and values, but that is what practical reasoning is bound to be, and there is no reason to feel guilty as if we needed a certificate from God or from the Normative Truth (Plato's Forms). Ethics fits Neurath's boat (Chapter 1): we inspect each part while relying on the others, and if we are coherent and imaginative the result may be perfectly seaworthy. Meeting boats sailing in other directions, we need not lament that we are not in dry dock, certified by Reason or God. "They are not in any such place, either."
Afghanistan, effective altruism and machine ethics. The Taliban, ousted in 2001, returned in August 2021 and have barred girls from secondary school (March 2022) and women from universities (December 2022). Effective altruism (from Singer, 1972; Ord and MacAskill, about 2009 to 2011) applies the "how well do lives go" test with cost-effectiveness evidence, and was shaken by the FTX collapse (2022), a lesson about principles detached from character. Machine ethics has made manipulation versus respect an engineering problem: models tuned by human feedback (Christiano et al., 2017; InstructGPT, 2022) can drift into sycophancy, and Anthropic's "Constitutional AI" (2022) writes principles such as honesty and non-manipulation into training.
The respect test as a design rule for advisory AI. For an RM copilot or customer assistant: permitted are surfacing facts the customer lacks, prompting concerns they hold but have not brought to mind (a pension shortfall, an unnoticed fee), and stating the bank's own position openly; not permitted is the magic injection, changing aims without passing through understanding (false urgency, bias-exploiting framing, flattery). This matches the FCA Consumer Duty (July 2023) focus on customer understanding, and yields a testable rule: would the recommendation survive the customer's seeing exactly how it was produced?
Farewell
The book has introduced the great themes without coercing anyone into doctrines, and the sensitive reader may notice a kind of pessimism: the harmony between thought and world, the bridge between past and future, what the physical world contains and how minds fit into it are topics on which the finest thinkers have hurled themselves, only to be frustrated; "there always seem to be better words, if only we could find them, just over the horizon".
Not cynicism: mitigated scepticism. One could turn cynical, as if the defence of reflection in the Introduction were hollow; professional philosophers have been known to. Blackburn disagrees: understanding the problems is itself a good. If the upshot is what Hume called "mitigated scepticism", a decent modesty in our intellectual speculations, that is no bad thing. The world is full of ideas, and a becoming sense of their power, difficulty, frailties and fallibility "cannot be the least of the things it needs".
Master Takeaways
Think adds up to a training in diagnosis: take a compelling argument, lay out its premises, find the step that smuggles in a picture, and ask what the picture costs. Across nine problems the same culprit recurs, a self or a reason standing outside nature, and the same remedies, naturalism, Kantian organising principles and Humean modesty. Blackburn's lasting lesson is that the right response to a philosophical puzzle is usually not a new entity but a better understanding of the concepts we were already using.
The takeaways
- Bizarre possibilities test the picture that allows them. If a theory of mind leaves Zombies and Mutants wide open, that is a reason to reject the theory, not to doubt other minds; the same move dissolves much of scepticism about the self.
- Being unconscious of a cause is not being conscious of its absence. Schopenhauer's water explains why we feel like uncaused choosers and why the design argument's appeal to creative intelligence seems natural: both mistake ignorance of mechanism for insight into freedom.
- Freedom is inside control, not intervention. We are free when action flows from flexible, reason-responsive decision processes sensitive to true and available considerations, just as a thermostat controls temperature by being part of the causal order.
- The self is a point of view, not an object. Hume could not find a self because the "I" is the structural frame from which experience is organised; imagining "my" survival elsewhere transports no soul, only a viewpoint.
- Hume's maxim on miracles is Bayes's theorem. When the prior of the hypothesis is tiny and the evidence easily arises otherwise (lies, delusion, love of wonder), testimony can almost never establish a miracle; the same arithmetic governs every base-rate problem.
- There is no mysterious normative order. Read "having a concern = thinking something good" from left to right: value-talk expresses concerns we demand of each other, and concerns are better or worse by the lives they produce, not by matching a moral fact.
- Logic asks one question. An argument is valid when there is no way the premises could be true and the conclusion false; criticising premises (untrue) and criticising inference (invalid) are different charges.
- Scope matters. Frege's quantifiers show why "everyone has a mother" does not yield "someone is everyone's mother", and why ordinary sentences ("all the nice girls love a sailor") hide several distinct claims.
- Implicature is not entailment. What is said (truth-condition) differs from what is suggested (implicature); loaded questions like "have you stopped beating your wife?" are answered by spelling out which case holds.
- Induction cannot be justified without induction. Straightie and Kinkie have the same track record, and every argument for nature's uniformity presupposes it; Bayes formalises the stalemate rather than breaking it.
- Improbability alone proves nothing. Chance throws up improbabilities as readily as design (the Scrabble letters), so significance and surprise need a reason to think chance is not responsible.
- Science's virtue is self-correction, not insight. Hume shows we never see why nature must behave as it does; science "staves off our ignorance a little longer", and its authority rests on mechanisms that weed out bad ideas.
- Paradigm-relativity does not follow from paradigms. That we see through culturally shaped categories does not make all categories equally good: some paradigms are discarded for good and sufficient reason.
- The design argument fails on its own terms. Arguing from experience, minds are late, fragile and dependent products of generation, so a self-sustaining mind is the less likely cause of the world; and a mixed world implies at best a mixed creator.
- An incomprehensible God makes no difference. If divine goodness is unlike ours, belief is inert: "a nothing would serve just as well as a something about which nothing could be said".
- Pascal's wager presumes the knowledge it disclaims. Under genuine ignorance, a god who rewards unbelief is as likely as one who punishes it, and many rival creeds break the dominance argument.
- Necessary existence explains nothing. The cosmological argument and the search for a Straightjacket behind laws share one dilemma: regress, brute fact, or an unintelligible necessity that might as well belong to the world.
- Identity can be a policy. Locke's forensic notion of personhood and the ship registry show that "same person" or "same ship" may be fixed by what we need the concept for, and need not be transitive like strict identity.
- Crisp survival is an illusion that thinking cannot easily kill. Scrambling cases show no metaphysical fact matches our sense that the future holds three crisp options, yet the sense persists.
- Secondary qualities depend on perceivers; primary ones may too. The argument from relativity needs an equal-adaptation premise, sense by sense; once granted, Berkeley and Hume show the same reasoning threatens solidity and matter.
- Physics trades things for powers. Faraday's "the nucleus vanishes" leaves a world of ifs, and we have no conception of what, if anything, underlies the powers.
- Concepts are rules, not pictures. Kant's insight is that space, time, objects and causes are organising principles for experience, which is why the self and the world are two sides of one coin.
- The point of view always comes in. Berkeley's mirror room shows we cannot imagine a scene from nowhere, which is why Moore's unobserved worlds prove nothing about mind-independent beauty.
- Concerns are not desires, and egoism is false. The punch in the stomach and the godfather's fool's paradise show that desires aim at the world, and that one of our concerns is not to be deceived about whether our concerns are met.
- Respect means changing minds through understanding. Refusing the magic injection, uncovering hidden concerns and putting your own view on the table are legitimate; manipulation treats people as mere means.
- Some bad feelings are good. Guilt and contrition track wrongs; dissolving them by therapy is self-alienation unless they have become an obsession.
- "Just us" applies only to equally good solutions. Driving on the left is just us; denying women education or slavery is not, because it trespasses on concern and respect.
- Mitigated scepticism is a good outcome. The great problems resist solution, but understanding them is itself a good, and a decent modesty about our ideas is what the world needs.
Frameworks & mental models
| Name | What it says | Use it when |
|---|---|---|
| Conceptual engineering | Philosophy studies how concepts function and interconnect and what changes would improve them | Disputes turn on undefined terms (fairness, explanation, identity) |
| High, middle, low grounds | Reflection is good in itself, continuous with practice, and the antidote to the sleep of reason | Justifying time spent on conceptual work |
| The thinking skills | Avoid confusion, detect ambiguity, hold one thing at a time, argue reliably, see alternatives | Reviewing any argument or design document |
| Valid, sound, charity | Valid: no way true premises yield a false conclusion; sound: valid plus true premises; charity supplies suppressed premises | Assessing an argument before rejecting it |
| Self-correction of the senses | Errors are detected by further use of the same faculties, over time | Building trust in any instrument or model |
| Four responses to scepticism | Rational foundationalism, natural foundationalism, coherentism, scepticism | Deciding what can ground a body of belief |
| Neurath's boat | Repair beliefs or values plank by plank while afloat; no dry dock | Changing a system that must keep running |
| Zombie/Mutant test | A theory that allows undetectable variation of mind is suspect | Evaluating theories of mind, or claims about AI self-reports |
| Locke versus Leibniz | Mind is annexed to matter by brute fact, or intelligible from it | Asking whether a correlation could become an explanation |
| Behaviourism, functionalism, identity theory | Mental states as dispositions, as roles, or as brain states found a posteriori | Choosing what evidence would settle a mental attribution |
| Opponent-process colour space | Colour appearance computed from cone comparisons, making some colours impossible | Showing how structure of experience follows from function |
| Dilemma of determinism | Determined or random, neither seems up to us | Testing any account of agency |
| Four decision modules | Scanner, tree producer, evaluator, producer | Modelling choice and where praise and blame act |
| Three compatibilist definitions | Would have done otherwise if she had chosen; would have chosen otherwise given other considerations; those considerations true and available | Deciding responsibility, negligence and excuses |
| Interventionist versus inside control | Control by a ghost outside nature, or by being part of the causal order | Rejecting the false choice of ghosts or trams |
| Lazy and energetic sophisms | "The future will be what it will be" licenses doing nothing no more than getting cracking | Answering fatalism in practice |
| Locke's memory criterion | Same person as far as consciousness of past experience reaches | Thinking about continuity of persons or systems |
| Forensic identity and the ship registry | Identity rules fixed by purpose, possibly non-transitive | Defining when a changed system is "the same" for audit |
| The self as organising principle | The "I" is the frame needed to interpret experience as a world | Understanding self-models in robots and agents |
| Ontological argument and the turkey | Real and imagined beings cannot be compared for greatness | Spotting definitional proofs of existence |
| Cosmological dilemma | Regress, or an arbitrary stop that could be the world | Any "ultimate explanation" argument |
| Conditions for analogy | Close similarity and experience of the kind of cause | Reasoning from resemblance |
| Problem of evil | Omnipotence, omniscience, goodness and evil are hard to reconcile | Assessing theodicies and free will defences |
| Hume's maxim | Accept testimony only if its falsehood would be more miraculous than the fact | Weighing extraordinary reports or alerts |
| Pascal's wager matrix | Belief dominates if the options are the only ones | Checking a decision matrix for missing states |
| Truth-tables and tautology | Connectives as truth-functions; validity as "if premises then conclusion" being a tautology | Testing formal or rule-based logic |
| Quantifier scope | ∀∃ and ∃∀ are different information | Reading requirements, contracts and queries |
| Truth-condition versus implicature | What is said versus what is suggested | Interpreting testimony, references and model outputs |
| The Golden Harp lottery | Equal track records give no reason to prefer the straight pattern | Thinking about generalisation and inductive bias |
| Bayes's theorem | Posterior = prior × fit / prior of evidence | Diagnostics, screening, evidence evaluation |
| Models and mechanisms | Empirical laws versus mechanisms from which they derive | Distinguishing correlation models from explanatory ones |
| Paradigms with self-correction | Categories are culturally shaped, but testing weeds out bad ones | Answering relativism about science |
| Primary versus secondary qualities | Shape and motion in objects; colours and tastes as powers to cause sensations | Separating governed facts from detector responses |
| Argument from perceptual relativity | Variation plus equal adaptation in each sense implies mind-dependence | Testing relativist arguments anywhere, including ethics |
| Berkeley's two objections | Solidity is known like colour; primary qualities cannot be abstracted from secondary | Asking what a scientific image leaves us |
| Straightjacket | A guarantor of law faces regress, brute fact or unintelligible necessity | Thinking about laws of nature |
| Transcendental idealism | Objects conform to our forms of intuition and categories | Explaining a priori structure without Berkeley |
| Realism, conceptualism, nominalism | Rules are objective, mind-made, or non-existent | Debates over categories, kinds and labels |
| Concern versus desire | Being moved by a thought need not be wanting | Reading motives charitably |
| Identify versus objectify | Concerns we own versus concerns we treat as cravings | Deciding whether to meet a concern or remove it |
| Two ethical questions | The ideal life versus the boundaries of conduct | Separating personal ideals from obligations |
| The equation of concern and value | Read left to right (non-cognitivism) or right to left (cognitivism) | Locating the source of moral authority |
| Two legitimate ways to change aims | Uncover hidden concerns; put your view on the table | Persuasion that respects autonomy |
| Coherence, history, imagination | Values must be jointly implementable, tested by time and by imagined cases | Building policy that people can live by |
| "Just us" test | Equally good solutions may differ; trespasses on concern and respect may not | Judging cultural difference |
Strengths. Think is a model of argument reconstruction: premises laid out, flaws located precisely, and thought experiments (Dreamboat, the turkey, the dormitory, the Golden Harp, the mirror room, the punch in the stomach) that make abstract points stick. It reads the classics as live interlocutors, and its central diagnosis, that a Cartesian picture of a self outside nature drives puzzles about mind, freedom, self and God alike, gives the book unusual unity.
Blind spots and slant. It is an introduction with a thesis: naturalist, Humean, anti-Cartesian and expressivist. Religion is assessed almost only through eighteenth-century natural theology, with little on modern philosophy of religion (Plantinga, Swinburne, reformed epistemology). Ethics presents Blackburn's own non-cognitivism while leaving Kantian, utilitarian, contractualist and virtue theories as faint echoes, and gives no game theory. Epistemology skips Gettier and the analysis of knowledge. Political philosophy, aesthetics (beyond Moore), philosophy of language beyond logic, and non-Western traditions are largely absent. Some historical shortcuts need correcting: the publication date of Hume's Dialogues (1779), Pascal's dates (1623 to 1662), the "colour-blind" mammals (most are dichromats), dogs and pointing (they follow it well), and the fine-tuning number, which is Penrose's figure about initial entropy.
Contested or dated. Blackburn's escape from Reid's transitivity objection is weaker than the standard psychological-continuity reply. His use of Moore's open question concerns motivation, not meaning. The cognitivist "panic" he dismisses is a mainstream position (motivational externalism), and moral realism revived strongly after 1999 (Parfit, Scanlon). The Frege-Geach problem for expressivism, which Blackburn's own quasi-realism addresses, goes unmentioned.
How the field has moved. Bayesian epistemology and the replication crisis vindicated his warnings about significance testing; Goodman's grue and Kripke's quus are now the standard forms of his induction and rule-following puzzles. Philosophy of mind organised around the hard problem, IIT versus global workspace, illusionism and the question of AI consciousness. Free will debate centred on Frankfurt cases, manipulation arguments and the reinterpretation of Libet. Personal identity followed Parfit and animalism. Philosophy of religion moved to fine-tuning, the evidential problem of evil and skeptical theism. Metaphysics of laws split between Humean best-system and powers views, while physics made Faraday's vision standard in quantum field theory. Ethics saw effective altruism and machine ethics turn his questions about manipulation, respect and "concerns we expect of each other" into practical problems for AI alignment.
Who's who
| Name | Who | Why they matter here |
|---|---|---|
| Socrates (c. 470 to 399 BC) | Athenian philosopher | Model of philosophy as skill; "the unexamined life is not worth living" |
| Plato (c. 427 to 347 BC) | Greek philosopher | Forms; realism about universals against the sophists |
| Aristotle (384 to 322 BC) | Greek philosopher | First logic of forms; negligence as chosen; desire as seeing good |
| Epicurus and Lucretius | Greek and Roman atomists | The swerve; Epicurus's questions on evil |
| St Augustine (354 to 430) | Theologian | "The pull of the will and of love": the non-cognitivist direction |
| St Anselm (1033 to 1109) and Gaunilo | Monk-theologians | Ontological argument and the perfect island parody |
| St Thomas Aquinas (c. 1225 to 1274) | Theologian | Cosmological arguments from dependence |
| Galileo Galilei (1564 to 1642) | Physicist | Secondary qualities only in the consciousness |
| René Descartes (1596 to 1650) | Philosopher, mathematician | Doubt, Cogito, dualism, trademark and ontological arguments |
| Blaise Pascal (1623 to 1662) | Mathematician, theologian | The wager |
| John Locke (1632 to 1704) | Empiricist | Primary/secondary qualities; memory criterion; forensic persons |
| Baruch Spinoza (1632 to 1677) | Rationalist | Freedom as understanding |
| G. W. Leibniz (1646 to 1716) | Rationalist, mathematician | Mind must be intelligible from matter; calculus of disputes |
| Pierre Bayle (1647 to 1706) | Sceptic | Turned doubts on primary qualities |
| George Berkeley (1685 to 1753) | Idealist | Subjective idealism; the master argument |
| Thomas Bayes (1702 to 1761) | Clergyman, mathematician | Bayes's theorem |
| Thomas Reid (1710 to 1796) | Common-sense philosopher | The simple self; the brave officer |
| David Hume (1711 to 1776) | Empiricist, sceptic | Induction, causation, miracles, design, bundle self, mitigated scepticism |
| Immanuel Kant (1724 to 1804) | Critical philosopher | Copernican revolution; self as structure; "mere means"; elastic balls |
| Edmund Burke (1729 to 1797) | Conservative thinker | Respect for inherited forms, within limits |
| Arthur Schopenhauer (1788 to 1860) | Philosopher | The water parable on felt freedom |
| Michael Faraday (1791 to 1867) | Physicist | Matter consists of powers |
| J. S. Mill (1806 to 1873) | Philosopher | Against "mental slavery" in On Liberty |
| W. K. Clifford (1845 to 1879) | Mathematician | Ethics of belief |
| Gottlob Frege (1848 to 1925) | Logician | Quantifiers, modern logic |
| Bertrand Russell (1872 to 1970) | Philosopher, logician | Definite descriptions; five-minute world; "favouritism" |
| G. E. Moore (1873 to 1958) | Philosopher | Two unobserved worlds; open question |
| Otto Neurath (1882 to 1945) | Philosopher | The boat: coherentist repair |
| Ludwig Wittgenstein (1889 to 1951) | Philosopher | Private language, rule-following, "no question of recognising a person" |
| Peter Strawson (1919 to 2006) | Philosopher | Objective versus participant attitudes |
| Thomas Kuhn (1922 to 1996) | Historian of science | Paradigms and revolutions |
| David Lewis (1941 to 2001) | Philosopher | Fission as a road that splits |
| Derek Parfit (1942 to 2017) | Philosopher | Identity is not what matters (context for Chapter 4) |
Connect it
- Base rates govern alerts and screening. Hume's maxim and the 1 in 11 test case are the logic of sanctions, PEP and fraud alert triage: measure precision at the real base rate, treat correlated alerts as one witness, and never let a fluent narrative stand in for evidence that is hard to produce unless the hypothesis is true.
- Every model is a Straightie bet. A model validated on past data embodies inductive bias; Hume's problem becomes drift monitoring, stress testing for kinked futures and governance that treats past performance as habit, not proof.
- Identity of systems is forensic. Locke's ship registry is the right template for deciding when an upgraded model or agent is "the same system" for validation, attestation and SM&CR accountability: write the continuity rule down.
- Self-reports are not evidence of states. The Zombie lesson applies to AI copilots that "explain their reasoning": verify by functional probes and counterfactual tests, and treat a model's "belief" as a disposition, not a stored sentence.
- Respect is a design constraint. Blackburn's refusal of the magic injection gives advisory AI a testable rule aligned with the FCA Consumer Duty: a recommendation should survive the customer's seeing exactly how it was produced.