Fallacy of Composition¶

Infer that a predicate established for one or all parts must hold of the whole they form, without a warranted preservation rule across aggregation, interaction, scale, configuration, or shared constraints.

Version: v5 · 2026-09-19 · History
Domain-specific #: 371
Domain group: Humanities
Origin domain: Rhetoric
Subdomain: Informal Fallacies → Rhetoric
Aliases: Composition Fallacy

Core Idea¶

The fallacy of composition infers that a predicate true of one or all parts must therefore be true of the whole they compose. The premise about the parts can be correct. The defect is the missing preservation rule across aggregation, interaction, scale, configuration, or shared constraints.

The error is conditional. Additive or closure-preserving predicates can pass from parts to whole when an explicit warrant applies. The Atomistic Fallacy is the multilevel-statistics subtype, adding an individual-level estimate, a group-level target, and contextual variance.

How would you explain it like I'm…

Tiny Bricks, Big Castle

Every Lego brick is tiny, but a castle made of lots of bricks is not tiny. The fallacy of composition is thinking that whatever is true of each little piece must be true of the whole big thing. Sometimes it does work, because lots of light bricks really do add up to a heavier pile. You just can't assume it always works.

Parts Don't Always Add Up

The fallacy of composition is assuming that if the parts have some quality, the whole must have it too. Each player on a team might be great, but the team can still play badly together. Things can change when parts are combined, interact or compete with each other. The mistake isn't reasoning from parts to whole at all, because some things really do add up, like weight. The mistake is assuming it carries over without a good reason.

Unwarranted Part-to-Whole Inference

The fallacy of composition is concluding that a property true of one or all parts must be true of the whole they make up. The facts about the parts can be completely correct. The flaw is assuming, without a reason, that the property survives being combined. Properties can change when parts interact, grow in scale, get arranged in a pattern, or compete for the same resources. Not all part-to-whole reasoning is wrong, though. Additive quantities like mass really do sum up. So the argument is fine when there's a demonstrated rule that the property carries over, and fallacious when that is just assumed. Its mirror image is the fallacy of division, which goes from whole to parts.

The fallacy of composition is the inferential error of concluding that a predicate true of one or all parts must therefore be true of the whole they compose. The premise about the parts may be true. The defect lies in an unlicensed preservation step, because no adequate rule is given showing that the property survives combination, interaction, scale, configuration, or competition for shared resources. The error is conditional rather than a blanket ban on part-to-whole reasoning. Mass, cost under stated accounting rules, and other additive quantities aggregate validly, and some classes are closed under specified operations. A composition inference is therefore warranted exactly when the predicate has a demonstrated preservation rule, and fallacious when preservation is treated as automatic. In multilevel statistics, the atomistic fallacy is a subtype: it promotes an individual-level relationship to a group or population while ignoring contextual variance. The general fallacy does not require data or statistical modeling, and it applies equally to qualitative predicates and configured systems.

Scope of Application¶

The fallacy appears in informal logic, economics, organizations, engineering, and system design wherever local truth is promoted to a collective conclusion.

Clarity¶

Ask: “What makes this predicate survive the operation that creates the whole?” Repeating that every part has the property does not supply the bridge.

Abstract Reasoning¶

From \(\forall i\,P(x_i)\), one may infer \(P(C(x_1,\ldots,x_n))\) only if \(P\) is preserved under composition operation \(C\). Without that additional premise, the move is unsupported.

Knowledge Transfer¶

The roles transfer across argumentative settings, but the abstraction remains domain-specific because it is a normative diagnosis of an inference. Aggregation and Emergence carry the observer-free structures that explain particular failures.

Example¶

One spectator sees better by standing; it does not follow that every spectator will see better if all stand. Universal adoption changes the reference environment and removes the local advantage.

Relationships to Other Abstractions¶

Current abstraction Fallacy of Composition Domain-specific

Parents (2) — outgoing typed relations

Fallacy of Composition is a kind of Cross-Level Inference Prime

Fallacy of Composition is the defective upward species of Cross-Level Inference, carrying a part-level predicate to a whole without a warranted preservation rule.
Fallacy of Composition is a kind of Informal Fallacy Prime

Fallacy of Composition is the Informal Fallacy species that illicitly promotes a part-level predicate to a whole.

Children (1) — abstractions linked here by typed relations

Atomistic Fallacy Domain-specific is a kind of Fallacy of Composition

Atomistic Fallacy is Fallacy of Composition specialized to projecting an individual-level statistical relationship onto a group or population.

Hierarchy paths (5) — routes to 5 parentless roots

Fallacy of Composition → is a kind of Cross-Level Inference → presupposes Hierarchy → is a kind of Network

Show alternative paths (4)

Distinction from Neighbors¶

Fallacy of Division reverses the direction.
Atomistic Fallacy is the statistical part-to-group subtype.
Ecological Fallacy is the statistical whole-to-individual subtype already carried by Fallacy of Division, not a separate node.
Emergence is a system phenomenon, not an inferential error.

Neighborhood in Abstraction Space¶

Fallacy of Composition sits in a sparse region of the domain-specific corpus (93^rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25

Notes¶

(New domain genus; queued for Claude style harmonization and independent citation review.)