Wikipedia Deep Dive

Infinite regress

12 min read

It's turtles all the way down.

That old joke captures something philosophers have wrestled with for millennia. The story goes that a scientist gives a lecture about the Earth orbiting the sun, and an elderly woman objects: "The world is actually a flat plate resting on the back of a giant turtle." When the scientist asks what the turtle is standing on, she replies triumphantly: "It's turtles all the way down!"

This isn't just a punchline. It's a window into one of philosophy's most persistent and fascinating problems: the infinite regress. And whether you realize it or not, you've probably bumped into this problem yourself, usually around age four, when you first discovered the maddening power of asking "why?" after every answer.

What Exactly Is an Infinite Regress?

An infinite regress is a chain of explanations that never ends. Each link in the chain depends on the previous one, following the same pattern forever.

Here's the structure: Something has a property because it's connected to something else with that property. But that second thing only has the property because it's connected to a third thing. And so on. Forever.

Consider causation. Why did the window break? Because the ball hit it. Why did the ball hit the window? Because the child threw it. Why did the child throw it? Because neurons fired in her brain. Why did those neurons fire? Because of chemical reactions. Why those reactions? Because of earlier molecular states. Keep asking, and you never reach bedrock.

Or consider knowledge. How do you know the café is open? Because you read their sign. How do you know you can trust the sign? Because you've found their signs reliable before. How do you know your memory of their reliability is accurate? Because... well, because of other beliefs you hold. Each justification rests on another justification, which rests on another.

The pattern has two essential ingredients. First, there's a recursive principle—a rule that says "X has property F because Y has property F." Second, there's a trigger—something that actually has the property in question, which kicks off the chain.

When Infinity Becomes a Problem

Here's where things get interesting. Not all infinite regresses are created equal.

Some are perfectly harmless. Consider the statement "It's true that Paris is in France." That statement is itself either true or false. If true, then it's also true that "it's true that it's true that Paris is in France." And so on, forever. We've generated an infinite regress of truths. But who cares? These truths aren't doing any work. They're just piling up like reflections in parallel mirrors—infinite, yes, but completely benign.

The same goes for numbers. The Peano axioms—the foundational rules of arithmetic—guarantee infinitely many natural numbers. Each number has a successor. That's an infinite regress of sorts, but mathematicians sleep soundly at night.

The pattern seems to be this: abstract objects can regress infinitely without causing trouble. Propositions, numbers, sets—they can stack up forever. The problems arise when we're dealing with concrete things, especially explanations of why something exists or happens.

The Three Flavors of Vicious Regress

Philosophers distinguish between vicious and virtuous regresses. A vicious regress is one that creates genuine problems for a theory. There are three main ways this can happen.

The Impossibility Problem

The most dramatic form of viciousness: the regress leads to a logical contradiction or metaphysical impossibility.

Some philosophers have argued that actual infinities are simply impossible. If you accept this, then any infinite regress is automatically vicious—it posits something that cannot exist. This was a popular view historically, and it still has defenders today.

But the impossibility argument is subtler than it first appears. Not all infinities seem equally problematic. Consider three types: infinite cardinality (infinitely many things), infinite extensive magnitude (infinite extension in space or time), and infinite intensive magnitude (infinite density or concentration).

An infinite causal chain stretching back through time? That's infinite cardinality and perhaps infinite extensive magnitude. Troubling to some, but not obviously contradictory. The universe might simply have existed forever, with no first moment.

An infinite energy density concentrated at a single point? That's infinite intensive magnitude, and it does seem genuinely problematic—the kind of thing physics rules out.

So the impossibility argument, while powerful when it applies, doesn't condemn all infinite regresses equally.

The Implausibility Problem

Even if an infinite regress is possible in principle, it might be wildly implausible in practice.

This objection often targets theories about human minds. Consider the regress of justified beliefs. Suppose every belief needs to be justified by another belief. Then to have even one justified belief, you'd need infinitely many beliefs. Is that possible? Maybe. Some philosophers argue that most of our beliefs are dispositional rather than conscious—they're there in some sense, even when we're not thinking about them. Perhaps we have infinitely many of these unconscious belief-states.

But this seems like a stretch. It requires attributing to finite human minds an infinite structure that we have no independent reason to think exists. The hypothesis isn't contradictory, just implausible.

There's also the consideration William of Ockham made famous: don't multiply entities beyond necessity. An infinite regress multiplies entities infinitely. That's suspicious.

But Ockham's razor cuts in complicated ways. The cosmological argument for God claims to increase parsimony by positing a first cause—one entity instead of infinitely many. Yet it does so by introducing an entirely new kind of entity: a divine being. That's adding qualitative complexity while reducing quantitative complexity. Which matters more? Philosophers disagree.

The Explanatory Failure Problem

This is perhaps the most interesting form of viciousness, and it brings us back to those turtles.

The turtle cosmology is meant to explain why the Earth doesn't fall. But notice what happens: each turtle's stability is explained by reference to another turtle, whose stability must also be explained. The pattern never bottoms out. Have we actually explained anything?

In one sense, yes. If you ask specifically why the Earth doesn't fall, the answer is: it's supported by a turtle. That's a local explanation—it answers a specific question about a specific thing. Each step in the regress provides a perfectly good local explanation.

But if you ask why anything in this system is stable at all—why there's stability in the first place—the regress fails utterly. That's a global explanation, and the turtle theory can't provide it. At every step, it assumes exactly what it's trying to explain.

This is related to the informal logical fallacy called begging the question (which, despite common misuse, doesn't mean "raising the question" but rather "assuming what you're trying to prove"). An infinite regress can be a disguised form of question-begging, offering the appearance of explanation while merely passing the explanatory buck forever.

The Mystery of Transmission

Some philosophers have tried to identify more precisely what makes explanatory regresses vicious. One promising idea involves the concept of transmission.

Imagine explaining why your neighbor owns a bag of sugar. You discover she got it from the previous owner, who got it from someone else, who got it from someone else... forever. This regress seems viciously unsatisfying. Why?

Because ownership is being transmitted through the chain. At each step, we assume that someone already had ownership and passed it along. But this presupposes ownership rather than explaining how ownership arises in the first place. To transfer something, you first have to have it. The regress never explains the original having.

But not all explanatory chains involve transmission. Consider a Bayesian account of epistemic justification. One belief justifies another not by transferring some mystical property called "justification," but simply by raising the probability that the second belief is true. The first belief might also happen to be justified, but that's beside the point—it's not what does the justifying.

If this analysis is correct, then non-transmissive regresses might be acceptable even as global explanations. The regress exists, but it's not vicious because no property needs to be possessed before it can be passed along.

Escaping the Regress

Philosophers have developed several strategies for dealing with potentially vicious regresses.

Denial: "What Regress?"

The simplest response is to deny that any regress exists. Maybe the theory in question doesn't actually generate an infinite series. Perhaps there are hidden assumptions or subtle distinctions that block the regress from getting started.

Embrace: "Infinite Is Fine"

Infinitists bite the bullet and accept the regress while denying it's vicious. In epistemology, infinitism holds that beliefs can be justified even though the chain of reasons extends forever. This is a minority view, but it has sophisticated defenders.

The infinitist argues that justification doesn't require completing the chain—it only requires that for any given belief, a further justifying belief is available if needed. You don't have to traverse infinity; you just have to ensure that the path continues as far as anyone might want to go.

Foundationalism: "The Buck Stops Here"

The historically dominant response is foundationalism. There is a first element in the series—a foundation that supports everything above it but doesn't itself require the same kind of support.

In epistemology, foundationalism posits basic beliefs that are justified without needing support from other beliefs. What justifies them? Different theories give different answers. Acquaintance theories, for instance, say that some beliefs are justified through direct contact with their objects. You believe you're in pain not because of any inference from other beliefs but because you're directly acquainted with the pain itself. The regress stops there.

In metaphysics, foundationalism appears as the view that reality has a fundamental level—basic entities that ground everything else but aren't themselves grounded in anything. Quarks and electrons, perhaps, or spacetime points, or whatever physics eventually settles on. Everything else exists because of arrangements of fundamental stuff, but the fundamental stuff just exists, full stop.

The cosmological argument for God is a form of foundationalism about causation. Instead of an infinite regress of causes, there's a first cause—something that causes other things but wasn't itself caused. This first cause is identified with God.

Notice that foundationalism requires distinguishing two kinds of explanation or two kinds of entities: the foundational and the derivative. The same recursive principle that applies to derivative elements doesn't apply to foundational ones. The challenge for any foundationalist theory is explaining why this distinction is legitimate rather than arbitrary.

Coherentism: "It's a Circle, Not a Line"

Coherentism offers a radically different picture. Instead of a linear chain—whether infinite or terminating in a foundation—it sees the relevant entities as forming an interconnected web.

In epistemology, coherentism holds that beliefs are justified by how well they fit together with other beliefs. Justification isn't a property that flows from foundational beliefs upward; it's a property of entire systems. A belief is justified not because it sits at the end of a chain of good inferences, but because it belongs to a coherent network of mutually supporting beliefs.

Imagine a jigsaw puzzle. No single piece is primary or foundational. Each piece's place is determined by how it fits with its neighbors, and the whole picture emerges from the pattern of connections. Similarly, coherentists argue, beliefs justify each other mutually rather than hierarchically.

Critics of coherentism worry that it's too easy to construct coherent systems of beliefs that are wildly false. A paranoid conspiracy theory might be highly coherent—every piece of evidence is explained, every objection answered—while being completely detached from reality. Coherentists respond by adding constraints: the system must cohere not just internally but also with perceptual experience, or it must be the kind of system a rational person would form under idealized conditions.

The Regress and Radical Skepticism

The infinite regress problem has deep connections to skepticism about knowledge and justification.

Here's a classic skeptical argument. To know something, you need justification. But any justification can be questioned—what justifies the justifier? You face three options, and none looks good:

First, the regress might continue forever. But then you never actually have justification—you're always waiting for the next link in an endless chain.

Second, the regress might terminate in foundations. But then those foundations are unjustified—you're building knowledge on a base of mere assumption.

Third, the regress might circle back on itself. But then you're reasoning in a circle—using A to justify B and B to justify A, which proves nothing.

This "Agrippan trilemma" (named after the ancient skeptic Agrippa) has haunted epistemology for millennia. Each response to the regress—infinitism, foundationalism, coherentism—is partly an attempt to escape one of these horns without impaling itself on another.

Foundationalists argue that basic beliefs aren't unjustified—they're justified in a different way, not through inference but through some kind of direct warrant. Coherentists argue that their circles aren't vicious—holistic mutual support is different from linear circular reasoning. Infinitists argue that infinite chains aren't incomplete—potential infinity can provide all the justification we need.

The debate continues. The regress problem hasn't been solved so much as transformed into a menu of options, each with costs and benefits.

Why This Matters

You might wonder whether any of this has practical significance or whether it's just philosophers playing intellectual games.

But the regress problem shows up everywhere. In artificial intelligence, systems that learn from data face regress questions: how do you know the training data is reliable? How do you know the method for checking reliability is itself reliable? In law, questions of authority generate regresses: what authorizes the constitution? In science, the problem of induction is partly a regress problem: how do you justify using past regularities to predict the future?

More fundamentally, the regress problem reveals something about the structure of explanation itself. Every explanation takes something for granted. You can always ask for further explanation. This means that complete, self-sufficient explanation—explanation that doesn't assume anything—may be impossible in principle.

Some philosophers find this troubling. They want a world that makes sense all the way down. Others find it liberating. If complete explanation is impossible, then perhaps we should stop demanding it and learn to be satisfied with the partial, local, good-enough explanations we can actually provide.

Either way, the next time a child asks you "why?" for the fifteenth time in a row, you can console yourself with the thought that they've rediscovered one of philosophy's most profound and persistent puzzles.

It really might be turtles all the way down. The question is whether that's a problem or just the way explanations work.