What it is
A concept is a bundle, and the bundle can be taken apart. What it comes apart into are characteristics: the properties of objects, abstracted and held in common by everything the concept covers. The full set of characteristics that constitute a concept is its intension. The set of objects that fall under it is its extension.
The two move against each other, and that inverse relation is the working mechanism of the whole apparatus. Add a characteristic and the intension grows while the extension shrinks, because fewer objects now qualify. Take one away and the reverse happens.
Not every characteristic earns its place. Suonuuti's illustration is the plainest available: every plant has roots, trees included, so having roots separates a tree from nothing, whereas having a hard self-supporting trunk separates a tree from a climbing plant. Only the second is delimiting, and only delimiting characteristics should be used in a definition. The nearest neighbouring record is the one on writing definitions, and the boundary is that analysis and formulation are separate skills, which ISO 704 marks by giving them separate clauses.
Where the authorities stand
ISO 704 has given characteristics a clause of their own across editions, but the numbering does not carry across. The 2022 edition orders the material as terminological analysis, intension and extension, essential against non-essential characteristics, and shared against delimiting characteristics; the 2009 edition put the last two the other way round and titled the essential-characteristics subclause after necessary and sufficient characteristics, which the later edition dropped. Cite the edition, always.
The older text is candid about its own inheritance. It defines an essential characteristic through what is necessary and sufficient to fix an extension, then observes that terminology work, being concerned with intension and designation, does not use necessary and sufficient characteristics. The apparatus comes from logic, and in practice the delimiting against non-delimiting cut carries almost the whole load.
The real dissent lies elsewhere. Temmerman argues that many categories, in the life sciences above all, have prototype structure rather than a fixed set of characteristics, so that no bundle of delimiting characteristics can be assembled at all. Frame-based terminology reaches a similar conclusion from another direction, replacing the bundle with roles supplied by the event a domain is organized around. Both have their own records, and neither denies that the analysis works where concepts are well behaved, which in legal and administrative terminology they usually are.
Consequences for translation
Most equivalence disputes between a translator and a reviser are arguments about characteristics conducted without anyone naming one. Naming them ends the argument faster than any appeal to usage or to a dictionary.
- A definition that admits too much has too few delimiting characteristics; one that admits too little has mistaken a non-essential characteristic for a delimiting one. Both faults are found by listing, not by reading.
- Equivalence is settled by comparing intensions, not forms. Two designations that share no material at all can name one concept, and two that look identical can name concepts separated by a single characteristic with large consequences.
- Where one characteristic decides the case, put it on the term record. A note saying the source concept requires registration and the candidate does not is worth more to the next translator than a page of usage citations.
- The analysis is what makes a coined equivalent defensible. A coinage that reproduces the delimiting characteristics can be argued for in front of a reviser or a court; one chosen because it sounded right cannot.
- It affects word count and segmentation. Where the target language has no single designation carrying all the delimiting characteristics, the equivalent becomes a phrase, and that is a fact about the concept rather than a stylistic expansion.
Examples
EN>FR
- Source: river
- Target: fleuve / rivière
- Comment: The standard illustration of inclusion, used by Felber and still in circulation. French divides the field by a characteristic English does not lexicalize, whether the watercourse discharges into the sea, so one English concept covers two French ones. No amount of surrounding context resolves the choice; only the fact about the outlet does.
PL>EN
- Source: spółka jawna
- Target: registered partnership
- Comment: General partnership is the obvious candidate and Kusik rejects it on one delimiting characteristic: the Polish company is constituted by entry in the National Court Register, and the English and American general partnership is not. He proposes registered partnership because the designation carries that characteristic. The method does not merely reject a candidate, it produces one.
FR>EN
- Source: table de salon
- Target: coffee table
- Comment: A pair usually treated as unproblematic that turns out to be a thin intersection once the characteristics are listed. Vezzani uses it to argue that partial equivalence is a matter of degree. The working lesson is more modest: a pair that looks obvious can still deserve a note.
References
- Felber, H. (1984). Terminology Manual. Paris: Unesco and Infoterm.
- ISO 704:2009. Terminology work: Principles and methods, 3rd edition. Geneva: International Organization for Standardization, clause 5.4.
- ISO 704:2022. Terminology work: Principles and methods, 4th edition. Geneva: International Organization for Standardization, clause 5.4.
- Suonuuti, H. (2001). Guide to Terminology, 2nd edition. Nordterm 8. Helsinki: Tekniikan Sanastokeskus, sections 1.2 and 1.3, pp. 13 to 14.
- Kusik, P. (2022). "English Translation Equivalents of Selected Polish Partnership Types Revisited from the Perspective of Comparative Law." Lingua Legis 30, 7 to 21, at pp. 17 to 18.
- Vezzani, F. (2025). "Pour une formalisation du concept d'équivalence en terminologie." mediAzioni 46, A138 to A156.