The sorites paradox is a family of arguments in logic and the philosophy of language showing how apparently harmless assumptions about small changes can lead to an unacceptable conclusion. Its traditional example concerns a heap of sand: one grain is not a heap, and adding a single grain apparently cannot turn something that is not a heap into a heap; repeated application of that principle nevertheless implies that no number of grains makes a heap. The paradox exposes a tension between clear cases, gradual change, and the apparent absence of sharp boundaries in ordinary concepts.
The heap argument
The name derives from the Greek sōros, meaning “heap.” A standard version begins with three claims:
- One grain of sand does not constitute a heap.
- If grains do not constitute a heap, then grains do not constitute a heap.
- A sufficiently large collection of suitably arranged grains does constitute a heap.
The second claim expresses a tolerance principle: a sufficiently small change should not alter whether a predicate applies. Starting from the first claim and repeatedly applying the second yields the conclusion that even a million grains do not constitute a heap, contrary to the third claim.
The argument can also run in reverse. Begin with a clear heap and remove one grain at a time. If removing one grain never destroys a heap, the reasoning eventually implies that a single grain—or even no grains at all—constitutes a heap.
Other examples involve “bald,” “tall,” “young,” “red,” and “rich.” Removing one hair apparently cannot make a non-bald person bald, yet enough successive removals produce a clear case of baldness. The paradox does not depend on the particular starting quantity. It requires a sequence with clearly contrasting endpoints and adjacent cases that appear too similar to justify different classifications.
Logical structure
Let mean that a collection of grains, under otherwise suitable and fixed conditions, constitutes a heap. For some sufficiently large finite number , the premises are:
Repeated applications of modus ponens to the first two premises yield , contradicting the third premise.
Under classical logic, this reasoning is valid: if the premises were all true, the conclusion would follow. The difficulty is therefore not simply an invalid inferential step. It is that individually plausible commitments form an inconsistent set.
A classical treatment also shows why rejecting the universal tolerance premise has a striking consequence. If the first and last collections differ in classification, there must be at least one adjacent pair for which:
Thus, within a classical, two-valued model, a one-grain difference must somewhere separate a non-heap from a heap. That result appears difficult to reconcile with the ordinary meaning of “heap.”
A sorites argument need not concern a process unfolding in time. It can compare a sequence of simultaneously existing objects, such as differently sized people or progressively shaded colour samples. Its structure concerns adjacent classifications, not necessarily physical transformation.
Vagueness and borderline cases
The paradox is a central problem in the study of vagueness. A vague expression has borderline cases: cases in which its application is not settled in the straightforward way characteristic of clear positive and negative instances. Some collections plainly are heaps; others plainly are not; intermediate collections seem to resist definite classification.
Vagueness differs from ambiguity. An ambiguous expression has multiple distinct meanings, as “bank” can refer to a financial institution or a riverbank. “Heap” can generate a sorites argument even after one meaning has been fixed. Vagueness also differs from ordinary uncertainty about measurement: knowing the exact number and arrangement of grains does not obviously settle whether they form a heap.
Nor is continuity by itself the source of the problem. A precisely defined predicate, such as “has a mass of at least one kilogram,” has a sharp boundary even though mass can vary continuously. Conversely, grain counts are discrete, yet “heap” remains vague.
The puzzle becomes more difficult through higher-order vagueness. If cases are divided into clear heaps, borderline heaps, and clear non-heaps, questions arise about the boundaries between those categories. A third category does not automatically remove the problem; it may relocate it.
Historical development
The sorites is traditionally attributed to Eubulides of Miletus, a fourth-century BCE philosopher associated with the Megarian school. Ancient accounts associate him with several paradoxes, including the liar paradox. The attribution identifies an important figure in the ancient discussion rather than establishing that he was the first person to notice every form of the problem.
The paradox attracted particular attention among the Stoics. Chrysippus discussed soritical questioning and was associated with a strategy of withholding an answer as a sequence approached uncertain cases. Such suspension avoids committing oneself at each step, but it leaves open the question of what makes the underlying classifications true or false.
Modern discussions have made the sorites a testing ground for theories of meaning, truth, and logical consequence. It is particularly prominent in debates over whether vagueness is a feature of language, a limitation of knowledge, or potentially a feature of reality.
The term sorites also has a separate use in traditional logic for a chain of connected syllogisms. A soritical chain in that sense need not be paradoxical.
Principal approaches
Epistemicism
Epistemicism holds that vague predicates have sharp boundaries, although speakers generally cannot know where those boundaries lie. On this account, there is a precise transition between non-heaps and heaps, and the tolerance premise is false at that transition.
This preserves classical logic and the principle of bivalence, according to which every proposition is either true or false. Borderline cases are cases of ignorance rather than failures of determinate truth.
The central challenge is to explain how ordinary linguistic use could determine an exact boundary that competent speakers cannot identify. Epistemicists distinguish the existence of a boundary from its knowability and argue that reliable knowledge may require a margin for error that is unavailable near the boundary.
Supervaluationism
Supervaluationism treats a vague expression as permitting multiple admissible precise interpretations, commonly called precisifications. Different admissible precisifications of “heap” can place the threshold at different points.
A sentence is supertrue if it is true under every admissible precisification and superfalse if false under every one. A borderline claim can be neither supertrue nor superfalse.
This account preserves many classical logical principles without assigning every sentence a determinate truth value. For example, “This is a heap or this is not a heap” can be supertrue because it is true under every precisification, even when neither disjunct is supertrue.
Each precisification contains a sharp boundary, but no particular boundary is singled out as the uniquely correct interpretation. Consequently, a claim that there is some boundary can be supertrue without any particular proposed boundary being supertrue. Critics question whether this adequately explains ordinary judgments and how the admissible precisifications themselves are delimited.
Many-valued and degree-based theories
Other approaches replace the two classical truth values with three or more values, sometimes with a continuous range of degrees. In fuzzy logic, a predicate can receive values between and , allowing gradual movement from non-application to full application.
On such an account, adding one grain may change the degree to which a collection is a heap by only a very small amount. Repeating a small change many times can then produce a large overall difference.
The treatment of the argument depends on the chosen meanings of logical connectives and on which truth values count as acceptable for inference. It is not enough simply to assign intermediate numbers: the theory must explain which tolerance claims hold and whether the relevant rules preserve them.
A recurring objection is that exact degrees can introduce a new kind of precision. It may be no clearer why a collection has degree of heaphood than why a sharp boundary falls at a particular grain count. Defenders distinguish formal modelling choices from claims that ordinary language encodes uniquely exact numerical degrees.
Contextualist approaches
Contextualist theories emphasize that the standards governing a vague expression can vary with conversational circumstances. A collection may count as a heap relative to one comparison class but not another.
Some accounts argue that successive questioning changes the context. Attention to the similarity between neighbouring cases encourages identical answers, while comparison of distant endpoints encourages different ones. The standards in play may shift during the sequence, making locally plausible judgments appear to support a single, globally applicable tolerance principle.
The challenge is to explain sorites arguments considered all at once, rather than only in conversation. Contextualism must also distinguish ordinary context dependence from vagueness: fixing a comparison class does not obviously eliminate every borderline case.
Changes to inference or consequence
Some theories locate the problem partly in the inferential structure. They distinguish the acceptability of each short step from the acceptability of chaining all those steps into one argument.
Certain nonclassical accounts restrict the unrestricted transitivity of consequence or related structural rules. This allows them to treat local transitions as acceptable without licensing the endpoint conclusion.
Such approaches must explain why the altered rules capture vague reasoning while avoiding excessive restrictions on ordinary deduction. Merely observing that a long chain feels less compelling than a short one does not itself provide a logical solution.
Inconsistency-based approaches
Some proposals accept that ordinary vague concepts generate inconsistent commitments. Approaches associated with dialetheism allow some contradictions to be true, while suitable nonclassical logics prevent a contradiction from entailing every proposition.
These views preserve more of the original conflicting intuitions but face the task of explaining which contradictions are acceptable and how reasoning with vague predicates remains informative.
Significance and limits
The sorites matters wherever gradual differences meet categorical distinctions. In law, administration, and scientific classification, precise thresholds can make decisions determinate even when the underlying concept is vague. A legal age boundary, for example, can provide a clear rule without establishing that maturity itself changes abruptly on that date. Stipulating a threshold solves a practical classification problem, but it does not necessarily explain the original concept.
The paradox is also related to, but distinct from, slippery-slope arguments. A causal slippery slope predicts that one event will lead to further events. A sorites argument instead concerns a sequence of classifications and an asserted tolerance between neighbouring cases. The existence of intermediate cases alone establishes neither a causal chain nor the impossibility of a meaningful distinction.
Several tempting responses therefore leave the central issue unresolved. Identifying clear endpoints does not explain the intermediate cases; invoking gradual change does not specify an acceptable logic; and choosing a convenient cutoff does not show that ordinary language already contained it. A theory of the sorites must explain which commitment fails—or how the framework changes—while accounting for both clear classifications and the apparent harmlessness of small differences.