Epistemology

Boundary of Knowledge§

Stating a scientific fact requires us to first separate philosophical views from observational facts — and epistemology, the study of knowledge, is what lets us do that. Take an example. You ask a weather-prediction system, "Will it rain today?", and it answers, "It might rain." At first glance it has handed you knowledge. But ask yourself: if it does not rain, was the system wrong? No — whether it rains or not, "it might" was correct either way. As we will make precise once we reach falsifiability below, a claim that can never be wrong tells you nothing: philosophically, the system has given you no scientific knowledge at all.

Now compare it to a system that answers: "It will rain if and only if you see gray clouds today." This is a directly falsifiable claim: if there are no gray clouds yet it still rains, the system is wrong.

Reasoning about what an answer can and cannot tell you — that is epistemology at work.

Definition of Epistemology§

Before defining epistemology precisely, consider its literal translation.

Components of the word§

• Episteme: meaning knowledge, understanding, or acquaintance
• Logia: meaning the study of, or science of

Together it means: the study of knowledge.

Definition of Knowledge§

Classically, knowledge is defined as "justified true belief", with the well-known abbreviation JTB. Karl Popper challenged this definition. To call a belief justified, the argument goes, you would have to have checked every edge case, every condition, at every time and on every day — which can never actually be completed. So justification, taken strictly, is never finished.

Popper's is not the only challenge to JTB. The more famous one, due to Edmund Gettier, runs the other way: you can hold a belief that is justified and true and it still feels like it isn't really knowledge, because it turned out true only by luck. We will not chase that thread here — it is enough to know that the classical definition is contested from more than one side.

Rather than offer a rival one-line definition, Popper reframed where knowledge lives (this answers his own worry about justification, not Gettier's luck puzzle, which we are simply setting aside):

Of these three, World 1 (rocks and trees) is the one everyday intuition reaches for first. But in physics, we take "knowledge" to mean World 3 — the objective, shareable artifacts of theories and arguments. To make the three concrete: an apple falling is World 1; your private conviction that it will fall is World 2; Newton's written law of gravitation, which anyone can test and criticize, is World 3. This is exactly what lets World 3 sidestep Popper's worry about justified: a public theory never asks any one person to have personally checked every case — the impossible, never-finished task — it only has to be laid out openly and stay exposed to testing and criticism, a standard you actually can meet. (Taking "knowledge" to mean World 3 is Popper's proposal — a choice about what the word should mean here, not a settled fact — but it is the one we adopt.) That is what we will mean from here on when we say "we know."

Whichever stance you take on that choice, when this book says "we know" it means that World-3 sense — and acquiring it is never a step you can skip.

Why do we care about Epistemology§

Most physics textbooks keep their philosophy implicit. We want to make it explicit here, as a way to demystify science and separate philosophical choices from observed facts. This makes it easier to ask speculative questions about science. Whether you want ideas for sci-fi or want to contribute to the field, having a clear framework helps.

But the deepest reason is this: stating a fact requires us to first define the limits of knowledge — even if that knowledge is purely probabilistic.

There are a lot of statements that we think are true or false where, in reality, they are unknowable.

What is Science§

Falsifiability§

A scientific statement must be falsifiable: there has to be some possible observation that would show it wrong. (You could, in principle, split "false by pure logic" from "false by physical measurement," but in physics the two travel together closely enough that we will simply say falsifiable.)

For example, Claim: "Tomorrow might rain"

This is unfalsifiable, because whether tomorrow rains or not the statement is correct — by definition "might" means either case is possible. This does not mean the statement is false; the statement is true, but it is unfalsifiable, and therefore not a statement science takes any interest in.

Now compare a claim that is falsifiable: "All swans are white." A single black swan would prove it wrong — so it passes the test, whether or not it happens to be true.

Collecting the facts is never the hard part; the work is reasoning out whether they carry any information at all — and that reasoning can fail in subtler ways than a plain "might rain." Two such cases are worth seeing, but you can safely skip them on a first read and expand a box below when you are curious.

Unfalsifiable ≠ false: the case of an almighty God

The same rule applies to religion — or at least to its most airtight claim, the existence of an almighty God, which turns out to be unfalsifiable.

First, define the properties of God: an almighty, all-knowing, non-physical being. Two of those properties do the work. Because He is non-physical, no instrument can ever point at Him directly. And because He is almighty, any indirect evidence you might raise — say, the absence of some miracle you expected — could always be something He simply chose to allow. So no observation can ever disprove Him. Notice carefully what this does and does not mean: it does not show that God exists — it shows that the claim is unfalsifiable, and "unfalsifiable" is not the same as "true." (This narrow, maximally-hedged claim is the unfalsifiable one; plenty of other religious claims — the efficacy of prayer, the age of the earth — do make testable predictions.)

Dodging falsification: claims rigged so nothing can count against them

A claim can also fail this test in a sneakier way — by being rigged so that no observation could ever count against it.

Claim: "All adult human behavior is driven by unconscious, repressed childhood desires."

It looks testable — "just collect the data and see if it fits." But it is not, because whenever the data do not fit, a defender blames the data rather than the claim — even when a rival explanation would have predicted that exact mismatch. Nothing could ever count as evidence against it, so it is unfalsifiable. (This was one of Popper's own examples of pseudoscience.)

Other unfalsifiable claims worth recognizing (so you can spot a scam)

The list goes on, but a few more are worth mentioning so that you can protect yourself from scams.

• Simulation Theory: We are living in a simulation
• Introspective Solipsism (The "Matrix" Dilemma): brain in a vat
• Omnipotent Religious and Theological Claims: existence of God or the supernatural
• Psychological Egoism (The "Selfish Motive" Claim): whether truly, purely selfless altruism exists
• Subjective Definitions of "True" Traits: A real friend would never forget a birthday.
• Fate and "Everything Happens for a Reason": You didn't get that job because the universe has a bigger, better plan for you

And a few respectable but currently untestable ideas from pop-science — these are live, disputed research questions, not scams; they share only the unfalsifiable property, not any manipulative intent:

• The Everett Many-Worlds Interpretation (Quantum Multiverse): every quantum outcome really happens, each in its own separate, unreachable universe.
• The Holographic Principle (As a Literal Reality): our three-dimensional world is really information encoded on a distant two-dimensional surface.
• Cosmic Inflation Multiverse (The "Infinite Bubble" Universes): our universe is one bubble among endlessly many we can never reach.
• Fine-Tuning and the Anthropic Principle: the constants look "tuned" for life because only a life-permitting universe could contain observers to notice.
• What Happened "Before" the Big Bang: asks about a "before" at the very point where time itself may begin.

Verifiability (and why science does not require it)§

There is a tempting alternative that sounds even stronger than falsifiability: instead of asking whether a claim can be proven false, ask whether it can be proven true — verified. Surely real knowledge is what we can confirm? But this hides a fatal problem, and the white swan exposes it.

Return to the white swan, "All swans are white":

Falsifiable?: Yes — a single black swan refutes it. Verifiable?: No — no matter how many white swans you record, you can never prove you have seen every swan that exists.

This asymmetry is the pivot of modern science: a universal claim — and every law of physics is one — can be refuted by a single observation, yet never confirmed by any number of them. If we demanded verification, no law of physics would ever qualify as knowledge. Karl Popper's decisive move was to drop verification and keep falsification: a claim earns its scientific standing by being something that could be proven wrong, not something that can be proven right. (Notice this is the same underlying problem that sank JTB's "justified" — you can never finish checking every case — resurfacing here as the reason verification fails.)

This is essentially what Popper laid out in his 1934 book, later translated in 1959 as The Logic of Scientific Discovery. Deciding what counts as science is itself a philosophical choice — but Popper was the first to insist on this asymmetry, and to point out that the scientists of his day were chasing claims that could never be refuted at all. We treat this as common sense today only because earlier scientists, taking both the right and the wrong paths, built that common sense for us. Because of his contribution to the philosophy of science, modern science is often called Popperian.

Notice the consequence: verifiability is deliberately not one of the criteria on the checklist below — falsifiability is.

Reproducibility§

If a claim is not reproducible, then the claim is useless to science. This does not necessarily mean the claim is false, but a scientific claim must be reproducible.

For example, take "Tomorrow will rain."

Falsifiable?: Yes — just wait until tomorrow and see. Reproducible?: No — tomorrow is one specific time, and the claim says nothing about other times.

To make the claim scientific, we need to find out why tomorrow will rain, which is where the fun of physics begins. Recast it as "Given these specific atmospheric pressure and humidity levels, there is an 80% probability of precipitation" and both bars are cleared at once: it is falsifiable by collecting data, and reproducible by recreating that same pressure and humidity.

A second, statistical sense of "reproducible": my uncle who smoked

Claim: "My uncle smoked a pack of cigarettes every day, drank whiskey, and lived to be 100 years old. Therefore smoking isn't actually bad for your health."

Falsifiable?: Trivially yes.

Reproducible?: No. The story may well be true, but one person proves nothing — reproduce it across a thousand people and the result would very likely reverse, showing that smoking kills rather than heals. (Note this is a slightly different sense of "reproducible" than the rain case: there it meant repeating a condition across time; here it means testing a claim across a large enough population — statistical reproducibility — so a single lucky case cannot stand in for the rule.)

Tentativeness§

Tentativeness is the requirement that "a scientific claim must have a possibility of being wrong."

This is more of an attitude than a logical property. Any scientific claim must stay open to being proven false, no matter how it is phrased or voiced, because any falsifiable statement might in fact be falsified.

Within pure theory, this is simply an attitude toward science and harms no one. Outside of theory, however, ignoring it does real harm — and starts to deserve a rebuke. Take for example

Claim: "If a child is struggling in school, it is because the teacher isn't matching their specific learning style. For example, a 'visual learner' must see pictures, while an 'auditory learner' must listen to tapes."

Unlike the earlier tests, this one is not a clean pass or fail — it is about the stance you take. You hold the claim tentatively only if you keep it open to being proven wrong; cling to it against the evidence and you have abandoned tentativeness.

And, as it happens, is the claim even true? Over the last two decades, massive global psychology reviews tracked students extensively. The data consistently showed that teaching a student specifically to their preferred "style" does not improve their test scores or information retention at all. The human brain processes information best when multiple senses are engaged together, regardless of personal preference.

Keep the two lessons apart, though: whether learning styles are real is a question of falsifiable fact (and the answer turned out to be no); tentativeness is about the humble stance you should hold while such a question is still open, whatever the answer later proves to be.

Parsimony / Occam's Razor§

In plain terms: when two explanations account for exactly the same observations, prefer the one that adds fewer unnecessary assumptions. This is a choice about which theory to pursue — not a proof that the more complex one is false.

For example, take "A thunderbolt is the weapon of Zeus, punishing wrongdoers" versus "Thunder is just nature, like everything else." Before science, with no natural mechanism known, the Zeus story was simply the best explanation on offer — not because it was parsimonious, but because there was nothing to weigh it against. After the scientific revolution, the picture flips: thunder no longer needs an explanation of its own, it is just one more consequence of electromagnetism. That unification — one law instead of many separate stories — is what Occam's Razor rewards.

A second example: retrograde motion, epicycles, and the Sun

For millennia, astronomers had noticed that planets occasionally seem to move backward across the night sky (retrograde motion).

The complex theory: The Earth is the center of the universe, and planets move in wildly complicated, looping mini-orbits called "epicycles."

The parsimonious theory: The Sun is the center, and Earth is simply overtaking the other planets as it laps them.

Explanatory and Predictive Power (The "Time Traveler" Rule)§

A scientific statement cannot just be a story that explains what already happened (hindsight). It must provide a clear mechanism for how something works, and it must predict the future under specific conditions before it happens — it has to call the shot in advance, like a time traveler who already knows the result.

For example, Claim: "Yesterday it rained"

Predictive?: None — it already happened.

But recall our earlier claim: Claim: "Given these specific atmospheric pressure and humidity levels, there is an 80% probability of precipitation."

Predictive?: Yes — anytime we find exactly this condition, we can be 80% confident it will rain.

A genuine scientific claim escapes mere philosophical assertion by predicting observable facts.

What's Knowable (the Demarcation Checklist of Science)§

Our five items do not all do the same kind of work, so it helps to sort them into two groups (the word demarcation just means the line between science and non-science).

Three hard tests — properties of the claim itself. A claim is empirically knowable only if it is:

• Falsifiable
• Reproducible
• Explanatory and predictive — it must name a mechanism and call its shot in advance (the two halves of the earlier "time traveler" rule)

Two habits — how you must hold the claim, not gates it passes. As we saw, these are a stance and a choice, not pass/fail properties of the claim:

• Tentative — you keep it open to being wrong
• Parsimonious — you prefer the leaner explanation, while remembering that is a choice, not a proof

This checklist demarcates empirical science specifically; as we will see in the next section, there is also knowledge reached mainly by reasoning, which arrives by a different route.

A claim that passes all three hard tests is empirically knowable — that much is fixed by the claim itself. The two habits are then on you: to stay scientific you must hold even a passing claim tentatively and prefer its leaner form. (Being stubborn or needlessly complicated does not make the claim any less knowable; it only makes your handling of it less scientific.)

A claim can stop being an open, active scientific question for two very different reasons, and they are worth keeping apart. It may have already passed so thoroughly, and survived so many attempts to refute it, that we no longer bother re-running the checklist on it day to day (that the Earth orbits the Sun is not something we re-litigate each morning). This is not a permanent exemption: if new, contrary evidence ever turned up, the checklist would apply again at once — nothing is settled forever, exactly as tentativeness demands. Or — the opposite case — the claim may be one that could never pass, because it fails the very first test, falsifiability. Take the claim that an almighty, non-physical God exists: because such a being is non-physical, no instrument can point at it, and because it is almighty, any missing evidence could always be something it chose to allow — so no observation could ever count against it. That safety from any possible refutation is exactly the signature of something unfalsifiable.

From Empiricism to Rationalism§

Physics constantly balances two pathways to knowledge. They are not contradictory — just two different approaches to arriving at what we know.

Empiricism§

Knowledge from observation.

Water is the arche (the ultimate origin, principle, or fundamental substance)§

Thales of Miletus claimed water is the arche for several reasons:

• Everything that nourishes life is moist, and even warmth seems to arise from and feed on moisture — so moisture looked like the substance everything else lives on.
• Life needs water.
• Water is what holds the earth (in the cosmic view of his time).

These are all observational claims, which he then used to infer that water is the arche — though notice that the leap from them to "water is the arche of everything" already runs well past what was observed, a foreshadowing of the "logic-first" path we meet next. We are not here to judge whether his claim is true. Either way, Thales of Miletus is regarded as the father of Western science and philosophy: he was the first to try to escape supernatural explanation, and is traditionally credited as the first to predict a solar eclipse — an achievement usually attributed to his use of eclipse cycles inherited from earlier Babylonian astronomy rather than observation alone (though historians still debate exactly how, or even whether, he did it).

Planetary motion§

For well over a thousand years — from antiquity until Copernicus in 1543 — the Aristotelian geocentric worldview dominated natural philosophy. Early astronomers observed that while most stars stayed fixed in position across the seasons, the planets wandered erratically across the night sky. To explain this "retrograde motion" (where a planet seems to move backward in its orbit), astronomers like Ptolemy developed a complex system of "epicycles" — small circular orbits that a planet traced while riding a larger circle around the Earth. On one common reading, Ptolemy treated this mainly as a calculating device — a piece of geometry prized because it predicted where the planets would appear, rather than a firm claim about what was physically real out there (though historians still debate how instrumentalist he really was). Either way the lesson holds: getting the prediction right and explaining the true mechanism are two different things, a distinction we will lean on again and again.

In 1543, the Polish astronomer Nicolaus Copernicus challenged the paradigm by placing the Sun at the center, with Earth and the other planets revolving around it. What actually bothered him was a bookkeeping trick of Ptolemy's called the equant: to keep a planet's motion looking steady, Ptolemy had let it move at a constant rate only as seen from a special off-center point, not from the true center of its circle. Copernicus felt that broke a long-cherished ideal — that heavenly motion should be genuinely uniform and circular, not just uniform from some contrived viewpoint — and moving the Sun to the center was his attempt to rescue it. But notice two things. First, that ideal was a choice, not a fact: rearranging the picture this way predicted nothing new, and his own model — still built from circles and epicycles — was no more accurate than Ptolemy's, and in places was actually more complicated. Second, the real predictive gain came only later, with Kepler.

The decades-long observational records of the Danish astronomer Tycho Brahe, combined with the mathematical genius of his assistant, the German astronomer Johannes Kepler, solved the prediction problem without any return to epicycles. Using Brahe's extensive data on the motion of Mars, Kepler discovered that the planets travel in ellipses, and established what are now known as Kepler's Laws of Planetary Motion:

• The Law of Ellipses: Planets move in elliptical orbits with the Sun at one focus (one of the two interior points that fix an ellipse's shape — not its center).
• The Law of Equal Areas: An imaginary line drawn from a planet to the Sun sweeps out equal areas in equal intervals of time, meaning planets move faster when they are closer to the Sun.
• The Law of Harmonies: The square of a planet's orbital period (the time for one full orbit) is proportional to the cube of its average distance from the Sun. For a feel: Earth sits one astronomical unit from the Sun and takes one year; a planet four times as far takes not four but eight years, because 8² = 4³. (That "average distance" is more precisely the semi-major axis — roughly the half-length of the ellipse's long axis.)

Logic First§

This second pathway — reasoning your way to a conclusion that the data has not yet reached — is what we call rationalism. (Note this reasoning still stands on hard-won empirical results; it runs ahead of the data, it does not float free of it.) It requires much longer and deeper chains of reasoning, and many more thinkers; taken literally, every logic-first chain would trace all the way back to ancient times. So we will give only the conclusion of each path here. But please do not mistake the people who reached these conclusions for geniuses who created answers out of nowhere — they built on everyone before them, just as much as their predecessors did.

Quantum negative energies§

Paul Dirac famously looked at a mathematical equation and noticed it allowed a negative-energy state. Instead of treating it as non-physical, and without any experiment proving negative energy exists, with simple reasoning he was forced to accept the existence of antimatter, years before it was physically detected in a lab. We will not develop the quantum machinery in this book, but the reason he was forced is this: those negative-energy states are part of a complete set of solutions, and throwing them away would leave the description incomplete — and an incomplete description quietly leaks probability. (Picture probability as a fixed 100% that has to be spread across everything that can happen; discard a whole chunk of the possibilities and the total no longer adds up to 100%.) The chance of finding the particle somewhere would stop summing to a whole, and particles could simply vanish. So the math will not let him discard them.

Once negative energy must exist, Dirac was led to the "sea": he supposed those states are all already filled, and a gap in that filled sea is what we observe as antimatter. (Notice this whole argument rests on an assumption we chose: we treat the particle as a closed system, where the total probability must stay fixed; a genuinely open system need not conserve it. That is more a modeling decision than the deep philosophical choices we met earlier — but a decision nonetheless.)

Chemistry§

Dmitri Mendeleev (the father of the periodic table) organized the 63 elements known in his time by their atomic weights, and reasoned ahead of the available data by leaving empty spaces for undiscovered elements and accurately predicting their specific properties.

Again, this is a choice, not a proof: believing in a not-yet-observed element does not make it real. It is just that, if the element did not exist, explaining "why not" would be far more complex than accepting it on the strength of his model.

Operationalism and Relativity§

Operationalism: a physical concept is only meaningful if you can state the exact operation used to measure it. (This is itself a philosophical stance — a choice about what to count as meaningful, not a proven fact — but it is the one Einstein adopted.) To make it concrete: what is a length? Operationally, it is what a ruler reads. What is a time interval? It is what a clock reads.

Before Einstein, Isaac Newton treated time as absolute — flowing evenly across the universe. Einstein realized that an "absolute time" no one could ever measure has no operational meaning, and so he defined time strictly by what clocks read.

Absolute space and time§

In Newton's view, what we measure as a length or a time interval is only a measurement of space and time — not space and time themselves, which exist absolutely, beneath any measurement. He also held that the measured value of space and the measured value of time are independent of each other, which matches everyday intuition perfectly. On this view, measurement is enough to describe space and time for all practical purposes, even though it never quite reaches the absolute reality underneath.

In Newton's defense, one can recover Einstein's results while keeping absolute space and time — but only by adding machinery Newton never needed. You would posit a single hidden, undetectable vantage point — a "frame," meaning just some observer's view of who is moving and how fast — in which moving clocks really run slow and moving objects really shrink a little along their direction of travel. (We will see in a moment why a moving clock runs slow; lengths turn out to shrink for a closely related reason, which we will not derive here.) The bookkeeping works out; it is just far more contrived — and to a strict operationalist, a frame that no experiment could ever detect has no meaning at all, which is exactly why Einstein set it aside. Einstein himself praised Newton, noting that treating space and time as absolute was, in his era, essentially his only option.

Relative spacetime§

Between Newton and Einstein came the work of James Clerk Maxwell, whose theory of electromagnetism fixes the speed of light in empty space at a single definite value. Einstein saw that if that speed must come out the same for every observer, then measurements of space and time can no longer be independent of motion.

Why would that follow? Picture a simple clock that ticks by bouncing a pulse of light straight up to a mirror and back down. Now let that clock fly past you. From where you stand, the pulse no longer goes straight up and down — it travels a longer, slanted zig-zag, because the clock moved sideways between the bounce and the return. But light cannot speed up to cover the extra distance (its speed is fixed), so from your vantage point each tick simply takes longer: the moving clock runs slow. And from that clock's own point of view, symmetrically, it is your clock that runs slow — as long as neither of you speeds up or slows down. There is no privileged mover.

This leaves two options:

1. Either measured lengths and time intervals are not enough to describe space and time, or
2. space and time simply are what those measurements say (operationalism), in which case they cannot be absolute.

Einstein and his peers took option 2 — the more parsimonious choice, though, as always with Occam's Razor, a choice and not a proof. The precise definition comes later; for now the point is that he tied space and time together into one linked thing — spacetime — defined by the literal lengths and intervals we actually measure, rather than two separate absolutes sitting beneath them. (A later chapter returns to this under the name the relational view of space and time — a closely related framing we will sharpen there.)