David Lewis

Picture a seminar room in the 1970s. A soft-spoken man with a long beard — he looks a bit like a wizard who has wandered into a mathematics department — makes a claim, calmly, as if reporting the weather.

There are talking donkeys, he says. There are golden mountains. There are worlds where the dinosaurs never died, worlds where you were never born, worlds where you are, right now, a rock star reading a philosophy book between shows. These are not stories. They are not ideas in God's mind or convenient fictions. They are real — flesh-and-blood, rock-and-lava universes, every bit as solid as this one. There are more of them than you can count. We just can't visit, because they are completely cut off from us in space and time.

The room does what rooms always did when David Lewis said this. It stares.

Lewis knew the stare well. He gave it a name — the incredulous stare — and treated it, with complete seriousness, as the strongest objection his theory faced. Arguments he could answer; he answered hundreds, patiently, in print and in thousands of letters. But a stare is not an argument, and he admitted, disarmingly, that he did not know how to refute one.

Here is the puzzle of this chapter. The philosopher who believed the most unbelievable thing in modern philosophy was not a crank on the margins. He was the opposite: the field's gold standard. For thirty years, if you worked on almost any problem in metaphysics, philosophy of mind, or philosophy of language, you worked on Lewis — with him, against him, in machinery he built. Colleagues who thought his universes were absurd still conceded that no one reasoned more honestly, priced his assumptions more openly, or built systems that held together better.

So the question his career poses is not really "could there be talking donkeys?" It is something that should matter to anyone who reasons for a living: what do you do when your best, most carefully checked argument leads you somewhere your gut refuses to go? Lewis's answer was simple and unnerving. You go.

The Life

David Kellogg Lewis was born in Oberlin, Ohio, in September 1941, into a thoroughly academic household. His father taught government at Oberlin College; his mother was a distinguished historian of the Middle Ages. Dinner-table argument was the family sport, and the boy absorbed the assumption that ideas were things you built carefully and defended in public.

He did not start out aiming at philosophy. At Swarthmore College he began in chemistry — and it was a year abroad at Oxford, in 1959–60, that changed everything. There he was tutored by the novelist-philosopher Iris Murdoch and sat in on lectures by the giants of Oxford's golden age, including Gilbert Ryle (chapter 223), whose deflating, no-nonsense way with the mind had made him famous. The young chemist crossed the Atlantic one kind of student and came back another: he returned to Swarthmore and switched to philosophy for good. There is a pleasing irony here, because Ryle had spent his career draining metaphysics of its ghosts — and his accidental convert would go on to build the most extravagant metaphysics of the century.

Lewis went to Harvard for his doctorate under W. V. O. Quine (chapter 236), the great skeptic about meaning and necessity — another ironic pairing, since the student would make possibility the centerpiece of everything. The apprenticeship shows: Lewis inherited Quine's demand that a philosopher tally up, honestly, what their theory says exists, and never smuggle anything in unpaid-for. Harvard also supplies the era's best Lewis anecdote: he failed his preliminary examination in metaphysics — twice. He would get so absorbed in any question he found interesting, his wife later explained, that he neglected the rest of the exam. The man who twice failed a metaphysics exam became the most important metaphysician of his half-century; let that comfort you before your next test.

His dissertation became his first book, Convention (1969), and it announced his method to the world. The puzzle: how can language be based on agreements, when we'd need language to make the agreements? Lewis's answer borrowed the game theory of the economist Thomas Schelling. A convention, he argued, is a self-sustaining pattern of coordination — like everyone driving on the right — that persists because each person expects everyone else to keep to it, and no one gains by deviating alone. No founding contract required. Language, he showed, could arise the same way: signaling patterns that stabilize because they work. It was a naturalist's answer, close in spirit to what Paul Grice was doing with meaning and intention — and it was already pure Lewis: a hard old problem dissolved by patient, formal machinery.

After a few years at UCLA, he moved in 1970 to Princeton, and for the next three decades Princeton was the center of gravity of English-speaking metaphysics largely because Lewis was in it. His influence traveled by an old-fashioned medium: the letter. Lewis answered philosophical correspondence the way other people answer email — thousands of single-spaced, closely argued replies, to famous colleagues and unknown graduate students alike, often more carefully reasoned than the papers they responded to. A whole generation half-joked that the field was run by mail from a house in Princeton. (The letters were so substantial that, years after his death, scholars began publishing them in volumes of their own.)

Then there was Australia. Beginning in the early 1970s, Lewis and his wife visited almost every year — Canberra, Sydney, Melbourne — drawn by the blunt, realist, argue-about-everything style of Australian philosophy and by friendships with philosophers like D. M. Armstrong in Sydney. Australia became his acknowledged spiritual home, and Australian philosophy became, in return, deeply Lewisian.

The man himself was a bundle of endearing particulars: the beard, the encyclopedic love of trains and model railroads, the beloved cats (one of whom, Bruce, makes cameo appearances in his philosophy as a stock example). He was shy in conversation and fearless on the page. And behind it all ran a quiet, decades-long struggle with diabetes, managed with the same discipline he brought to everything. His wife Steffi was his lifelong collaborator, first reader, and editor — and, near the end, something more: when his kidneys failed in 2000, she gave him one of hers. He recovered, kept working, kept writing. On October 14, 2001, he died suddenly at home in Princeton of complications of the diabetes, at just sixty. The shock across philosophy was immediate and worldwide; the flow of publications, drawn from his papers and letters, has never really stopped.

The Big Ideas

Lewis wrote on almost everything, and everything connected to everything else — his work is less a pile of papers than one interlocking system. Four load-bearing walls hold it up.

Modal realism: all the worlds are real. Start with a kind of talk you use every day: things could have been otherwise. You could have missed the bus. The election could have gone the other way. Philosophers call this modality — the could-be, must-be, might-have-been dimension of reality — and the great modern tool for handling it is the idea of possible worlds: complete alternative ways reality could be. "You could have missed the bus" becomes "there is a possible world where you miss the bus."

Almost everyone treated this as a useful manner of speaking. Lewis's scandal was to take it literally. Possible worlds, he argued, are not abstract diagrams or sets of sentences. They are concrete universes — with matter, space, time, people — exactly as real as ours. Each is spatiotemporally isolated: no distance, no signal, no causal connection links one world to another, which is why you will never observe one. And here is the twist that makes the theory run: actuality is indexical. The word "actual," Lewis argued, works like the word "here." When you say "here," you don't name a metaphysically special place — just the place you happen to be. Likewise "the actual world" just means this world, the one we're in. The inhabitants of every world correctly call their own world actual. Ours has no cosmic badge of honor. This package is modal realism: the doctrine that every possible way a world could be is a way some world genuinely is.

Why on earth believe it? Lewis's argument was refreshingly honest: not proof, but usefulness. Take the worlds as real, he showed, and problem after problem falls open. Necessity and possibility get clean analyses (necessary = true in all worlds; possible = true in some). Counterfactuals — "if X had happened, Y would have" — get a semantics (more in a moment). Properties can be defined as sets of things across worlds; the meanings of sentences can be built from worlds; even the content of your beliefs can be modeled as the set of worlds your beliefs allow. One assumption, an entire toolbox. Philosophy, he argued, should shop the way science does: accept the theory with the best ratio of problems solved to assumptions made. For modal realism, he judged, the price is right. And he was scrupulously honest about the price — which is why he named the incredulous stare himself, weighed it as a genuine cost of his theory, and simply judged that the benefits outweighed it.

He also audited the alternatives. Most philosophers preferred ersatzism (from ersatz, "substitute"): keep possible-worlds talk but make the worlds abstract stand-ins — maximal consistent sets of sentences, say, or uninstantiated ways things might be. Lewis argued, in his masterpiece On the Plurality of Worlds (1986), that every ersatz substitute either quietly assumes the modal notions it was supposed to explain or can't tell apart possibilities that clearly differ. Cheap worlds, he argued, don't do the work. You get what you pay for.

This is where Lewis and Saul Kripke (chapter 284) — the twin giants of the possible-worlds revolution — split decisively, and the split is worth getting exactly right. For Kripke, possible worlds are models: formal points in a mathematical structure, stipulated counterfactual scenarios, "ways things could have been" that carry no exotic ontology — machinery for semantics, nothing more. For Lewis, they are universes. Kripke wanted the logic of necessity without the metaphysical bill; Lewis argued that if the worlds do real explanatory work, you must pay for them — and that Kripke's necessary-identities-discovered-empirically picture, however brilliant, left the worlds themselves unexplained. Two programs, one vocabulary: the models and the universes.

One more gear in the machine. If other worlds are isolated universes, then you exist in only one of them — so what makes it true that you could have been a doctor? Lewis's answer: counterpart theory. In other worlds there are people very similar to you — your counterparts — and "you could have been a doctor" is true because some counterpart of yours is one. Kripke fired the era's most famous jab at this: Hubert Humphrey, who lost the 1968 election, could not care less that someone resembling him wins in another world — he cares whether he could have won. The counterpart theorist's reply: he could have won precisely in virtue of his winning counterpart — that is simply what "he could have won" amounts to, and no shadowy transworld Humphrey, mysteriously one person in many universes, does the job better.

Counterfactuals and causation: the closest worlds. Lewis's Counterfactuals (1973) opens with one of philosophy's most cheerful first sentences, about what would follow if kangaroos had no tails (they'd topple over). The serious point: counterfactual "if"-sentences had resisted analysis for decades. Nelson Goodman (chapter 234) had shown why: to evaluate "if the match had been struck, it would have lit," you must hold fixed the right background facts (oxygen present) while releasing others (the match's never being struck) — and specifying which facts are "cotenable" with the supposition seemed to require counterfactuals again. Circle, unbroken.

Worlds break it. Lewis's semantics: "If A had happened, C would have happened" is true just in case, among the worlds where A happens, the ones most similar to ours are worlds where C happens too. Don't audit background facts one by one; rank whole worlds by overall similarity and check the nearest. (Robert Stalnaker had independently built a close cousin of this theory; the headline difference is that Stalnaker assumed a single closest world, while Lewis allowed ties and even endless chains of ever-closer worlds.)

Then Lewis put the machinery to work on causation itself. In a 1973 paper he proposed that causation is, at bottom, counterfactual dependence: the rock's throw caused the window's shattering because if the throw hadn't happened, the shattering wouldn't have — with longer causal chains stitched from such links. It is the "but-for" test every law student learns, upgraded into metaphysics. And then came the counterexamples, especially preemption: two assassins, one shoots first; the victim would have died anyway, so the counterfactual fails, yet the first assassin plainly caused the death. Lewis spent nearly three decades refining the analysis — from "Causation" (1973) to "Causation as Influence" (2000) — openly owning each epicycle. The result was less a finished doctrine than a forty-year public workshop in how to do philosophy: propose sharply, test brutally, revise honestly.

Humean supervenience: the mosaic. Behind all of this stood a picture, which Lewis named for David Hume (chapter 151), the great denier of hidden necessary connections. Humean supervenience is the thesis that, within each world, all there is to reality is a vast mosaic of local matters of particular fact — one little thing and then another. Point-sized bits of the world have their qualities; everything else — laws, causes, chances, minds — supervenes on (is fixed by) the total arrangement of those bits, the way a picture is fixed by its pixels. No pushings and pullings over and above the pattern.

But then what are laws of nature, if not iron rails guiding events? Here Lewis polished an idea from Hume's twentieth-century heir Frank Ramsey (chapter 227): the best-system analysis. Take all the truths about the mosaic; consider every way of systematizing them; the laws are the generalizations in the system that best combines simplicity and strength — the shortest cheat-sheet for the universe that tells you the most. Laws don't govern the pattern; they summarize it supremely well. Objective chances joined the system too, linked to rational belief by his famous Principal Principle — the rule that your confidence in an outcome should match what you know of its objective chance. And Lewis was honest about the physics: he acknowledged that quantum entanglement, where particles' states seem tied together non-locally, puts real pressure on the tidy pixel-mosaic — he defended the doctrine against philosophers, he said, while deferring to physicists on whether it ultimately survives.

The Ramsey–Lewis method and the mind. Ramsey had also left behind a technique for defining theoretical terms — "electron," say — by the total role they play in a theory. In "How to Define Theoretical Terms" (1970), Lewis perfected it into what is now called the Ramsey–Lewis method: rewrite the theory as one big sentence saying "there exist things that play these roles," and define each term as whatever occupies its role. Then came the masterstroke: apply it to the mind. Our everyday psychology — pain makes you wince, beliefs plus desires produce action — is itself a term-introducing theory, "folk psychology." Mental states are whatever states occupy those causal roles; and science shows the occupants are brain states. This yoked functionalism (mind defined by role) to the Australian identity theory of his friend D. M. Armstrong — the Sydney connection made doctrine. Lewis added famously humane refinements: the madman whose pain has all the wrong causes and effects, the Martian whose pain runs on hydraulics — pain as a role that different occupants can fill. When Frank Jackson's thought experiment about Mary, the scientist who knows every physical fact but has never seen red, threatened physicalism, Lewis answered with the ability hypothesis: what Mary gains is not new information but new abilities — to recognize, imagine, remember. Thomas Nagel's (chapter 281) "what it's like" got the same deflating treatment: know-how, not extra facts.

And around these pillars, the range: a paper using his semantics to diagnose Anselm's (chapter 82) ontological argument; the classic time-travel paper beloved of undergraduates, arguing that a time traveler can kill their grandfather in one sense of "can" and won't in another — no paradox, just two senses of ability; the compatibilist argument that we can be "free to break the laws" without ever breaking them; a dispositional theory of value (roughly: values are what we'd desire to desire under ideal conditions); and a mereology — the theory of parts and wholes — so permissive that any things whatever compose an object, including the front half of a trout plus the back half of a turkey. The trout-turkey exists, Lewis maintained; it's just not a thing worth mentioning at parties.

Before & After

Lewis is where a dozen threads of this book braid together. The deep ancestor of his plural cosmos is Leibniz (chapter 137), who imagined infinite possible worlds in the mind of God, with God actualizing the best — Lewis's audit deleted God and the favoritism, and let every world be real on its own. From Hume (chapter 151) he took the namesake vision of a world without hidden necessary glue. From Ramsey (chapter 227) he took both the best-system view of laws and the method for defining theoretical terms that became half of his philosophy of mind. From Ryle (chapter 223) came the Oxford year that converted a chemist; from Quine (chapter 236), his teacher, the discipline of the honest ontological ledger — count what your theory says exists, and pay for all of it. Goodman's (chapter 234) cotenability problem is the lock his closest-worlds semantics was cut to open. And Kripke (chapter 284), his exact contemporary, is the essential foil: together they made possible worlds the common currency of philosophy, while disagreeing utterly about what the currency was — formal models for the one, concrete universes for the other.

Forward, the legacy is an empire. The revival of confident, systematic metaphysics from the 1970s onward — after decades in which positivists and ordinary-language philosophers had declared it dead — is his revival more than anyone's, and the possible-worlds toolkit now sits in every subfield's toolbox. The Canberra Plan, the program of Frank Jackson, David Chalmers, and their Australian allies, turned the Ramsey–Lewis method into a general recipe for locating folk concepts in the scientific world. Today's liveliest metaphysics of science is a standing argument between Lewis's Humeans and the anti-Humean rebels — Tim Maudlin's primitive laws, the new theorists of powers and dispositions — over whether the mosaic needs governing after all. Formal epistemology builds on his papers, with the Principal Principle a fixture of every debate about rational credence. His closest-worlds semantics and his account of truth in fiction live on inside linguistics and the study of storytelling. And next, the road bends sharply: chapter 286 brings Julia Kristeva, born the same year as Lewis, whose questions — language, subjectivity, the speaking body — arrive from the French lineage of Beauvoir and Barthes, about as far from Princeton's ledger-book metaphysics as philosophy gets. The distance itself is the lesson: this story has more than one center.

Why It Still Matters

Start with the courtroom. The "but-for" test — would the harm have occurred but for the defendant's act? — is how the law decides causation every day, and its puzzle cases (two fires converging on one house, two assassins) are exactly Lewis's preemption cases. His forty-year workshop on counterfactual causation is, in effect, the law's own reasoning examined under a microscope; philosophers of law still mine it when the simple test breaks.

Now the machines. The current frontier of artificial intelligence and data science is causal inference — teaching systems to answer not just "what happened?" but "what would have happened if...?" Whether an algorithm discriminated, whether a drug worked, whether a policy caused a change: each is a counterfactual question, and the frameworks used to answer them are direct descendants of the possible-worlds semantics Lewis built. Every time an analyst runs a "what-if" model — of an epidemic, an election, a portfolio — they are ranking nearby possible worlds by similarity, whether they know the pedigree or not.

Lewis also equips you for a very modern temptation. Physicists speak seriously of multiverses; tech billionaires muse that we live in a simulation. Lewis's career is a master class in how to handle wild hypotheses responsibly: state the theory precisely, tally its costs honestly, accept it only if the explanatory benefits genuinely cover them — and never pretend a stare is an argument, or that an argument is refuted by a stare. In an age of hot takes, his ethic — believe where the reasoning leads, price your assumptions, answer every objection, keep your incredulity and your honesty in the same head — may matter more than any one doctrine.

And there is a quieter relevance. Regret and hope are counterfactual emotions: if only I had; imagine if I could. Lewis gives that everyday ache a logic. The you who took the other job is not nothing — there's a truth about how that would have gone, grounded in how the nearest worlds run. Thinking clearly about the might-have-beens, it turns out, is a skill; Lewis wrote its manual.

Over to You

Somewhere, on Lewis's view, there is a world where a counterpart of you made the choice you regret not making — and a truth about how it turned out. Does that thought comfort you, unsettle you, or strike you as meaningless? Why?

Lewis said the laws of nature are just the best summary of the universe's pattern — excellent bookkeeping, not commands. If that's right, does anything make the sun rise tomorrow? Does it matter?

Kripke got the worlds without the wild ontology; Lewis insisted the work isn't done until you pay for your tools. In your own thinking, when is it honest to use a framework you don't literally believe?

Where to Go Next

Start with the source: David Lewis, On the Plurality of Worlds — the modal-realist masterpiece; the opening chapter, where the worlds pay their way, is among the clearest philosophical prose of the century.

A modern way in: Daniel Nolan, David Lewis — a patient, accessible tour of the whole interlocking system, written for readers meeting it fresh.

The thread forward: From Princeton's quiet system-builder we cross to Paris and to Julia Kristeva (chapter 286), born the same year — where language is no longer a coordination game to be modeled but a drama of the speaking self, and philosophy's other great twentieth-century current reaches our story.