Number and classifiers

Whether the grammar makes you commit to how many, and how counting works. Also called Grammatical number or nominal plurality and numeral classification.

Reference/All structures

Obligatory number marking Optional number marking Numeral classifiers Dual and paucal systems

The parameter

Two questions sit behind this parameter, and they are usually asked together because the answers correlate. The first is whether a noun must carry a number value at all. The second is whether counting requires a classifier: a grammatical word standing between the numeral and the noun, chosen according to the shape, animacy or function of what is counted.

The correlation is real but not a rule. Languages that leave number unmarked tend to be the ones with obligatory classifiers, the standard explanation being that their nouns behave like mass nouns, so a unit of counting must be supplied before a numeral can attach. That is a tendency rather than a definition, and the two WALS surveys do not line up language for language.

The two-way singular and plural division is also a European habit rather than a universal. Arabic and Slovene have a grammatical dual, distinct from the plural and obligatory where exactly two are meant; others distinguish a paucal. Where plural marking exists it is often restricted by animacy, applying to humans first and to inanimates last or not at all.

How languages vary

  • Obligatory number marking: English, German, Dutch, French, Spanish, Polish, Russian, Turkish. Every countable noun carries a value.
  • Optional number marking: Mandarin, where the suffix 们 attaches only to human nouns and pronouns and is blocked by a numeral; Korean. Japanese has no nominal plural morphology at all.
  • Numeral classifiers: Mandarin, Japanese, Korean, Vietnamese, Thai, where a numeral cannot combine with a noun directly.
  • Dual and paucal systems: Standard Arabic and Slovene have a productive dual; Sorbian retains one; the dual survives in the pronouns of several Polynesian and Inuit languages.

Haspelmath's WALS chapter 34 samples 291 languages: 133 mark plural on all nouns obligatorily, 55 on all nouns but always optionally, 15 obligatorily except on inanimates, 40 only on human nouns obligatorily, 20 only on human nouns optionally, and 28 have no nominal plural. Gil's chapter 55 samples 400: numeral classifiers are absent in 260, optional in 62 and obligatory in 78. The European arrangement, obligatory plural on everything, therefore covers fewer than half the sample.

Consequences for translation

Where both languages mark number and select differently, the result is Catford's intra-system shift: Turkish suppresses the plural suffix after a numeral and English requires it, without either language lacking the category. Where the source leaves number open and the target does not, the translator supplies a value the source never carried, which is obligatory explicitation.

The second case is the consequential one, and European translators consistently underestimate how often it arises. A Japanese or Mandarin noun phrase in a contract, a certificate or a medical record frequently says nothing about how many, and the English rendering has to say.

In official documents, number is a fact. "The child" and "the children", "the account" and "the accounts", "the signatory" and "the signatories" allocate different rights and obligations, so a silent guess is a factual assertion by the translator. The correct handling is a rendering that preserves the indeterminacy where the register allows it, and otherwise a note recording that the source is silent. Where the instrument contains an interpretation clause providing that the singular includes the plural, that clause governs.

Classifiers change the shape of the segment. Counting expressions expand translating out of a classifier language and contract translating into one, so word counts diverge and term base entries have to record which classifier a noun takes.

Number errors read fluently. A wrong plural is not a mistranslation of content and does not trip a target-language reader, which is why revisers treat number, like definiteness, as a check run against the source.

Examples

TR>EN

  • Source: Üç kardeşim var.
  • Target: I have three siblings.
  • Comment: Turkish leaves the noun unmarked after a numeral, so the English plural is an intra-system shift. "Kardeş" is also unmarked for sex, so "brothers" or "sisters" would add what the source withheld.

ZH>EN

  • Source: 我买了书。
  • Target: I bought a book. / I bought books.
  • Comment: The noun is number-neutral and the sentence is complete. English cannot leave the question open, and neither rendering says what the source did.

EN>ZH

  • Source: three books
  • Target: 三本书
  • Comment: The classifier 本 is obligatory between the numeral and the noun and is selected by the noun's type. It corresponds to nothing in English, which attaches the numeral directly and marks the plural instead.

EN>AR

  • Source: the books
  • Target: الكتابان / الكتب
  • Comment: Arabic distinguishes a dual from a plural, so the translator has to decide whether exactly two are meant. English does not say, and Arabic will not let the question stand open.

References

  • Haspelmath, M. (2013). "Occurrence of Nominal Plurality." In M. S. Dryer & M. Haspelmath (eds), The World Atlas of Language Structures Online, chapter 34. Leipzig: Max Planck Institute for Evolutionary Anthropology.
  • Gil, D. (2013). "Numeral Classifiers." In WALS Online, chapter 55.
  • Corbett, G. G. (2000). Number. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press.
  • Aikhenvald, A. Y. (2000). Classifiers: A Typology of Noun Categorization Devices. Oxford: Oxford University Press.
  • Velupillai, V. (2012). An Introduction to Linguistic Typology. Amsterdam: John Benjamins.

Language pairs, understood structurally

Knowing which shifts a pair forces is what separates a defensible translation from a fluent one. It is the basis on which Translyta's language profiles are built.

Request a quote