[DSRP Evidence](https://dsrpevidence.org/)

# A Taxonomy of Part-Whole Relations

## Details

**Authors** Winston, Morton E. and Chaffin, Roger and Herrmann, Douglas

**Year** 1987

**Publisher** Cognitive Science

**Kind of work** article

**Discipline** Cognitive Science

**Secondary disciplines** Linguistics, Philosophy

**Applied** false

[Read it at the publisher](https://doi.org/10.1207/s15516709cog1104_2) 
10.1207/s15516709cog1104_2

## In authors' words

### What they set out to do (purpose)

To develop a classification of part-whole (meronymic) relations that explains how English speakers use "part of" and related terms, and to explain why part-whole syllogisms sometimes appear not to be transitive.

### Who or what was studied (sample)

No new participants. The paper draws on constructed example sentences judged for acceptability and validity, and on earlier sorting studies in which undergraduates sorted word pairs illustrating 31 semantic relations.

### How they did it (methods)

Linguistic analysis of example sentences using three features (functional, homeomerous, separable), together with a relation element theory and a principle of element matching. The taxonomy is compared with the clusters from the earlier sorting data.

### What they found (results)

The authors identify six types of part-whole relation: component-integral object, member-collection, portion-mass, stuff-object, feature-activity, and place-area. These differ from one another, and from spatial inclusion, class inclusion, attribution, attachment, and ownership, on the three features. Part-whole inference is transitive when the same relation type appears in both premises. Apparent failures arise when the premises mix types and so equivocate. A syllogism is valid only if its conclusion contains the elements shared by both premises.

## Commentary

### In short

This paper sorts the many things English speakers call a "part" into six distinct kinds of part-whole relation. It shows that the inferences people reject, such as part of a part not counting as part of the whole, come from mixing these kinds.

### Response

The paper bears most directly on the systems structure: it shows that ordinary part-whole talk is not one relation but at least six, each with its own logic of composition. The distinction structure also does work, since the six types and their separation from class inclusion, attribution, and ownership rest on contrasts along three features (functional, homeomerous, separable). The fit is partial in one respect: the evidence is analytic, built from constructed examples and prior sorting data, so it shows the structure in English usage rather than in a measured population. The relational character of the meronymic types, treated through relation elements, is arguably present but is not claimed here as a separate pattern.

**Patterns it shows** D, S

Shows the structure is there

**Added** 2026-10-13

**How to cite this** Winston, Morton E. and Chaffin, Roger and Herrmann, Douglas (1987). A Taxonomy of Part-Whole Relations. Cognitive Science.