And are there any examples in world's languages of the one mechanism developed into the other?

  • Grammatical gender is a specific form of noun class system in which the division of noun classes forms an agreement system with another aspect of the language, such as adjectives, articles, pronouns, or verbs. [...]

    There are three main ways [...] by grouping them with other nouns that have similar form (morphological) [...]

    ("Grammatical gender", Wikipedia)

  • Agreement or concord happens when a word changes form depending on the other words to which it relates.

    ("Agreement (linguistics)", Wikipedia)

The a/an-distinction in English meets the conditions: the nouns can be classified into two groups, one with "a" as the article and the other with "an".

The nouns with "an" share all morphological similarity: start in a vowel.

  • 5
    Starting with a vowel is not generally a morphological characteristic. They would need to share a morpheme, for example. Aug 28 '19 at 19:55
  • I think you may not be distinguishing between morphology and phonology. a vs an has to do with phonology. Aug 30 '19 at 4:44

Gender is, as mentioned in your first quote, a system of agreement based on noun classes. Gender systems always have a semantic component (or in other words, there is a "semantic core" to at least one of the noun classes/genders): there are languages where gender corresponds to phonological or morphological criteria for most words--or for most non-human words specifically--but there are no languages with a gender system that is always arbitrary from a semantic point of view and based purely on phonology (Kibort & Corbett).

The trigger of gender agreement is expected to be the noun itself, not the phonological form of the first word after the article in the noun phrase.

The English use of a vs. an does not qualify as a gender system. It is not based on noun classes (as shown by my examples below, the same nouns, contribution and idea can be used with either a or an depending on what word comes after the article), and there is no "semantic core" to the phrases that take a or the phrases that take an.

The a/an distinction is phonologically (not morphologically) conditioned: as you said, an is used before vowel-initial words (not nouns--the only thing that's relevant is the pronunciation of the directly following word, which is not a noun in phrases like an important contribution or a good idea). From a historical perspective, this phonological conditioning comes from earlier differences in the rate of dropping the sound /n/ before vowels vs. before consonants: the /n/ was originally present in the article in all contexts.

I haven't heard of a language where a gender system originated from a phonologically conditioned sound change like this. However, in languages that have preexisting gender systems, phonologically conditioned sound changes can influence the gender category of nouns with a certain phonological form. For example, many Romance languages have articles that take different forms conditioned both on the gender of the associated noun and the phonological form of the following word. I know of some evidence that a form that was originally conditioned phonologically may be reinterpreted as indicating gender information instead.

Spanish nouns like agua

A specific example: in Spanish, the noun agua, which has the typically feminine ending -a and which consistently triggers feminine agreement on postposed adjectives, takes the definite article el, which is identical in form to the article used for masculine singular nouns, and different from the article used for consonant-initial feminine singular nouns (la). This use of el seems to have originated as a morpho-phonologically conditioned variant form of the article (el is used before nouns starting with a stressed /a/, whether masculine or feminine), and for the standardized form of Spanish, the rule is often described that way. (Wiktionary specifies that unlike the English a vs. an rule, the standard la vs. el rule is conditioned by both phonology and morphology, not purely by phonology, as it seems that la is used before feminine adjectives starting with stressed /a/.)

But ukemi wrote an interesting answer to this question on Spanish SE that indicates that for some speakers, gender may be involved, as apparently there are examples with such nouns of el being used even when a preposed adjective intervenes between article and the noun, and of preposed adjectives being used in the masculine form.

It seems that, since the morphophonological condition for using el before a feminine noun is so restricted (when I first wrote this post, I misremembered it as applying to all feminine words starting with a stressed vowel, but it seems that it only applies to feminine nouns starting with the stressed vowel /a/ specifically, and there are exceptions even for the small number of words in that category), the phonological basis for the use of el with agua is not very salient to Spanish speakers, so the form el is prone to being reinterpreted as an example of masculine agreement for words preceding this noun.

(One thing I'm not sure about is to what extent the confusion/variability in agreement for adjectives and other preposed words before nouns like this is a matter of spelling vs. pronunciation. One of the examples in ukemi's post is que el abundante agua caída; my understanding is that because Spanish speakers blend vowels together between words, this would when spoken sound very similar or maybe even identical to the standard form que la abundante agua caída. But the example por aquelm asa y no por éstaf is clearly distinct in pronunciation from the form with feminine agreement (which would be por aquellaf asa y no por éstaf) as l and ll represent different consonant sounds in Spanish.)

Works cited


Sumelic's answer is entirely correct, and goes into quite a lot of detail. But there's also a much simpler answer to this specific question: the article doesn't specifically depend on any feature of the noun.


A good dog.
An excellent dog.
A very excellent dog.
An extremely good dog.

Clearly there's no property of the word "dog" that's affecting the article here.

Grammatical gender is usually thought of as a morphosyntactic property: it takes effect at the morphological and syntactic level, when the sentence is still in the form of an abstract tree made of abstract morphemes.

But here, the article doesn't depend on the structure of the tree, or the head of a phrase. It depends on the phoneme that comes immediately after it, in the final, linear sentence. In other words, this effect happens at the morphophonological level, after the syntax tree has been flattened down into a linear sequence, and the morphemes are made up of phonemes instead of just being abstract units. So it happens "too late" in the production process to be considered a form of gender.

P.S. I'm simplifying things a lot in this explanation, and speech production is a bit more complicated than a linear "processing pipeline". But the different "stages" of processing are still useful in all sorts of linguistic models, even if they don't correspond perfectly to specific regions of the brain or whatever, so I have no qualms about using them here.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.