Sorites Paradox (Heap)

The puzzle in one paragraph

10,000 grains of sand piled on a table is a heap. Remove one grain — still a heap. Remove another — still a heap. At no point does removing a single grain seem to convert a heap into a non-heap; the change is too small. But continue removing grains one at a time and you reach 1, then 0. Zero grains is plainly not a heap. So somewhere along the way the heap stopped being a heap — but if no single removal did the work, when?

The argument can be put as a formal induction:

1. 10,000 grains forms a heap. (Premise.)
2. For all n: if n grains is a heap, then n − 1 grains is a heap. (Tolerance principle.)
3. By induction, 1 grain is a heap. (Conclusion.)
4. 1 grain is not a heap. (Contradiction.)

Either step 1 is false (10,000 grains isn't a heap), step 2 is false (some single grain matters), or the rules of inference are wrong. None is comfortable.

Eubulides and the Megarians

The paradox is attributed to Eubulides of Miletus, a 4th-century-BCE Megarian logician and rival of Aristotle. Eubulides also gave us the Liar Paradox ("this sentence is false") and the Masked Man. Diogenes Laertius lists the Sorites in his life of Eubulides; Cicero discusses it in Academica II as a sceptical weapon against Stoic claims of certain knowledge; Sextus Empiricus elaborates it in his sceptical compendia.

The Stoics were the prime targets. They held that the wise person never assents to a vague proposition, but the Sorites suggests that almost every empirical claim is vague — when does Socrates count as bald? When does a child become an adult? When does a man cease to live? If the Stoic must withhold assent on every borderline case, his epistemology becomes paralysing.

Why this is more than a word game

Almost every predicate of human interest is vague. Bald, tall, rich, old, fat, red, warm, fast, religious, conscious, person, alive, healthy, guilty. If vague predicates are inherently paradox-prone, then ordinary reasoning is unsound on its most common materials — including the predicates that drive law (reasonable, negligent, imminent), medicine (terminal, functional), and ethics (person, autonomous).

The legal stakes are concrete. Roe v. Wade and its successors hinged on when a fetus becomes a person. Brain-death statutes hinge on when life ends. Statutory rape laws hinge on a sharp age line drawn through a vague predicate. The Sorites is the structural reason these lines look both arbitrary and necessary.

Major responses compared

Response	What it denies	Author	Cost
Epistemicism	That there's no sharp line	Williamson (1994)	Unknowable facts
Fuzzy logic	Bivalence (true/false only)	Zadeh (1965)	Higher-order vagueness
Supervaluationism	That every sentence has a truth value	Kit Fine (1975)	Truth-value gaps
Contextualism	That the predicate is fixed	Stewart Shapiro (2006)	Predicate shifts mid-argument
Nihilism	That heaps exist	Peter Unger (1979)	Almost nothing exists
Three-valued logic	Bivalence; adds "indeterminate"	Halldén, Körner	Inherits boundary problems
Tolerance denial	Step 2 of the induction	Crispin Wright (1976)	Counterintuitive grain matters

Worked example: the unmarked margin

Imagine a row of 10,000 men, each with one fewer hair than the last. Man 1 has 100,000 hairs (clearly not bald). Man 10,000 has 0 hairs (clearly bald). Walk along the row asking "is this man bald?". You'd never want to flip your answer at a single step — the man to your left has only one more hair than this one. But after enough single steps your "no" must become "yes". Where?

Man      1: 100000 hairs   not bald
Man   1000:  90001 hairs   not bald
Man   5000:  50001 hairs   borderline?
Man   8000:  20001 hairs   borderline?
Man   9500:   5001 hairs   getting bald
Man   9900:   1001 hairs   pretty bald
Man  10000:      0 hairs   bald

Most observers can confidently classify the extremes and feel uncertain in the middle. Epistemicism says one specific man is the last non-bald — say man 7,623 — and we cannot know which. Fuzzy logic says the predicate has degrees: man 7,623 is 0.43-bald. Supervaluationism says it is true that some man is the first bald, but for no specific n is it true that n is the first.

Counterarguments and modern responses

Williamson's epistemicism. In Vagueness (1994), Timothy Williamson argues that vague predicates have sharp boundaries — there is a precise number of grains that constitutes the minimum heap, and a specific second at which Socrates went bald. We cannot know these facts because our cognitive access to meaning is itself imprecise: small variations in linguistic usage shift truth values without our noticing. The view preserves classical bivalent logic, denies the tolerance principle, and pays for its cleanness with unknowable sharp facts.

Fuzzy logic. Lotfi Zadeh's 1965 paper "Fuzzy Sets" replaced binary truth with degrees in [0, 1]. "This is a heap" can be 0.6 true. The Sorites induction step says "if H(n) is fully true then H(n−1) is fully true" — but if H(n) is only 0.6 true, the conditional fires weakly, and after thousands of weak applications truth has drained to zero. Fuzzy logic is now standard in control engineering (washing machines, anti-lock brakes). Critics raise the higher-order question: where does the 0.6 come from? Setting that threshold reintroduces vagueness one level up.

Supervaluationism. Kit Fine (1975) proposed treating "heap" as admitting many sharpenings — precise predicates compatible with the vague meaning. A sentence is supertrue if true on every admissible sharpening, superfalse if false on every one, indeterminate otherwise. "This is a heap or it isn't" is supertrue (every sharpening makes the disjunction true) even when neither disjunct is determinately true. The view preserves classical inference rules at the price of truth-value gaps.

Contextualism. Stewart Shapiro (2006) and Diana Raffman argue the predicate's threshold shifts depending on conversational context. As you walk along the row, "bald" gets calibrated to the visible cases; small differences fall below the salience threshold. Sorites equivocates by holding the predicate fixed when in real use it would drift. This explains why the inductive step seems compelling locally but globally produces contradiction.

Mereological nihilism. Peter Unger (1979) bites the bullet: there are no heaps, no tables, no people. There are only fundamental simples; "heap" is a useful fiction we apply to arrangements. The paradox dissolves because nothing satisfies the predicate at all. Most philosophers regard the cost as too high — tables and people seem more secure than any premise of a paradox.

Variants

• The bald man. When does losing one hair leave you bald? Standard form, ancient.
• The tall man. A 7-foot man is tall; subtract a millimetre — still tall. Now you're 5'1".
• Forced march. Wright's variant: the speaker is asked at each step, "is this still F?" and answers "yes" to defend the tolerance principle.
• Conditional Sorites. Drop the universal premise; argue from "if H(10000) then H(9999)" through 10,000 modus ponens steps. Same conclusion, weaker premise.
• Sorites of decision. A shipwrecked sailor on a beach: "if I rest 1 more minute it makes no difference" applied repeatedly delivers death. Vagueness as akrasia.
• Continuity Sorites. A colour patch shifts continuously red → orange. At every instant the colour is the same as a moment ago. Yet it ends orange. The temporal version.

Common confusions

• It's not a slippery slope. A slippery slope is a causal claim ("legalising X will lead to Y"). Sorites is a logical claim about predicates — no causation involved.
• "Just define a number" doesn't dissolve it. Stipulating "a heap is ≥ 100 grains" gives you a non-vague predicate, but the resulting concept isn't heap — natural-language heap doesn't have that line. The puzzle is about real predicates, not stipulated ones.
• Higher-order vagueness. Even if you accept fuzzy degrees, the predicate "fully true" is itself vague. Solutions tend to push the vagueness up a meta-level rather than eliminate it.
• Sorites doesn't show concepts are useless. "Heap" works fine in normal use; the paradox arises only when we apply mathematical induction to it. The lesson is about the meta-theory, not everyday talk.