Loss of grammatical number

[Chengjiang]

Is there any natlang precedent for total loss of grammatical number? I know plenty of languages have lost dual or paucal numbers while retaining the plural, and some languages have lost morphological number distinctions on most nominals while still marking their number elsewhere, but are there documented cases of languages entirely losing number marking? (Highly restricted exceptions, such as continuing to mark number on personal pronouns but not anywhere else, are fine.)

(Presumably this must be able to happen, or we’d expect number-less languages to have slowly dwindled away, but I haven’t read any confirmed cases of it happening.)

[Moose-tache]

Vovin in 1994 suggested that the -na at the end of some words in Old Japanese is not a genitive but a plural or dual marker. Since Korean and Japanese have obviously recent synthetic plural marking, Altaicists have been looking for years for some evidence of ancient number and case marking that has since been lost. To my knowledge no smoking gun has been found, but that would be a good place to start looking. From the same part of the word, the Manchu plural marker is innovative, so there may have been a time when it had neither inherited Tungusic plural marking or innovated plural marking.

[Kuchigakatai]

Something I'd like to note is that presence of obligatory grammatical number is also not necessarily a yes/no thing. Many languages limit it to a few, some, or many but not all, syntactic contexts, making number marking optional elsewhere.

Mandarin, for example, generally marks number in the direct object by means of a determiner or classifier, and if a demonstrative is used it always comes with number marking too. Otherwise number marking is very much optional, especially so when NPs have definite reference (since their number is already known in the conversation).*

French has almost completely left number marking to determiners at the beginning of an NP (if we ignore its orthography), but this has not produced much ambiguity since NPs generally must start with one such determiner.** It would've been interesting if determiners had been forbidden in some contexts, producing number ambiguity in them, such as after some prepositions ('with', 'as'...), or in direct objects when the focus of a sentence is an adverbial.

I'm pretty sure that English 'with child' doesn't mean it must be one fetus. 'A woman with child' could mean she may be carrying twins, etc., even if by default we can reasonably think it is very likely one fetus...


* The most important exception is verb-object compounds, comprised of a monosyllabic verb + a monosyllabic noun, e.g. 招人 zhāo-rén lit. 'summon-person', meaning '(for a company) to be hiring; (for a disease) to be infectious', 開車 kāi-chē lit. 'open-car', meaning 'to drive (a vehicle, vehicles)'. But I'd say the noun shouldn't be counted as a full direct object NP there anyway, but as a morpheme incorporated onto the verb.

** The language even developed an article for indefinite mass nouns and indefinite plurals!, i.e. the awfully-named "partitive" article, which is notable as these things don't normally have an article in Romance or Germanic. In English we say 'I brought wood' and 'I brought matches', not 'I brought [article] wood' and 'I brought [article] matches'. But in French you have to say, J'ai apporté du bois, J'ai apporté des alumettes, respectively.

[Chengjiang]

Ser, thanks, but I’m aware of all that. I’m specifically looking for cases where a language went from generalized number marking to none or nearly none. Hell, as far as I’m concerned French still has full number marking, it just localized the nominal part of it to mostly-obligatory determiners.

[Richard W]

Just for clarity, what languages do you consider to lack number marking? A few relatively well-known languages will do.

I ask because there is no point my investigating Proto-Autronesian > Thai if you turn round and say that Thai has grammatical number, e.g. in the ancient first person pronouns and the demonstrative adjectives.

[Kuchigakatai]

Ser, thanks, but I’m aware of all that. I’m specifically looking for cases where a language went from generalized number marking to none or nearly none.

I think I meant my post more for my past self (and anyone who may find it useful) than for you... For years I mistakenly thought that having no number marking meant that nouns had no plural forms, which immediately meant there was ambiguity everywhere. I didn't realize number can be obligatorily marked by (nearly) always having a determiner that carries the distinction (as in French). Even the verb could carry number, with polypersonal agreement. I didn't realize that what really matters for ambiguity is whether the marking is obligatory vs. optional throughout.



By the way, it occurred to me that maybe some relevant languages can be found at the border of families with obligatory marking (Turkic, Tungusic, IE, Pama-Nyungan) and those with optional marking (Korean, Sino-Tibetan, Kra-Dai, Austronesian).

I found Bengali doesn't inflect nouns for number in indefinite inanimates. That is, with definite reference, there's an opposition between জুতাটা juta-ṭa ("shoe-DEF.SG") 'the shoe' vs. জুতোগুলো juto-gulo ("shoe-DEF.PL") 'the shoes', but with indefinite reference there's just জুতো juto 'a shoe, (some) shoes'. I have no idea if this is accompanied by obligatory number marking using a determiner nearby though, but it might be worth looking into.

More relevantly, according to Ekaterina Gruzdeva's Nivkh (Languages of the World / Materials 111), in Nivkh only subjects have obligatory number marking; for a non-subject NP it is optional (p. 17). This is particularly interesting in that language because it also has morphological case, with a "nominative" (really a direct case) that marks both subjects and objects (the so-called "dative-accusative" marks the animate causee of a causative verb), but obligatory number marking is tied to the syntactic function of subject, not the nominative case.

Nivkh is, of course, an isolate (or at any rate an isolate group of two or three closely-related languages), so this doesn't explicitly mean that number marking has been made optional in non-subjects from a previous all-obligatory stage. But it is notable that its speakers are bilingual with Russian, Nivkh being moribund to be replaced by it, and influence from Russian can't be brought up for a loss of number marking. The plural marker is very short in Nivkh proper / Amur too (-ku), but OTOH who knows if the Nighvng / East Sakhalin -kun(u) reflects an older longer form. I suspect that if it has long been optional then -kunu may be the older form, but OTOH Quechua has a very long -kuna obligatory plural marker... and Classical Arabic had -a:tu(n)/-a:ti(n) (feminine) and -u:na/-i:na (masculine) for most human nouns, also obligatory.

Hell, as far as I’m concerned French still has full number marking, it just localized the nominal part of it to mostly-obligatory determiners.

In that part I was basically saying, "What a missed opportunity to lose obligatory number marking...!!".

Proto-Autronesian > Thai if you turn round and say that Thai has grammatical number, e.g. in the ancient first person pronouns and the demonstrative adjectives.

I think it all depends on the extent to which using those demonstrative adjectives is mandatory. What do you mean by ">" when you write Proto-Austronesian > Thai?

[Richard W]

It's argued that Thai is an Austronesian language. The Tai-Kadai languages seem to have a lot of Austronesian in them - their non-Sinitic number system looks Austronesian (PMP to be precise). The Tai family has adopted Sinitic numerals. The good triconsonantal stem evidence seems to be limited to a few villages, and the reduction of CVCVC to CVC may be riddled with irregularities. The parallel reduction in Eastern Cham reportedly produces multiple forms - Brunelle has details.

[Curlyjimsam]

[French has almost completely left number marking to determiners at the beginning of an NP (if we ignore its orthography), but this has not produced much ambiguity since NPs generally must start with one such determiner.** It would've been interesting if determiners had been forbidden in some contexts, producing number ambiguity in them, such as after some prepositions ('with', 'as'...), or in direct objects when the focus of a sentence is an adverbial.

au(x) yields a partial case of this.

[Chengjiang]

Thanks, guys.

Just for clarity, what languages do you consider to lack number marking? A few relatively well-known languages will do.

Lacking obligatory grammatical number marking for verbs and most NPs is probably enough, e.g. for the purposes of this thread I’d consider Standard Chinese to lack number marking since the only obligatory indication of number is on pronouns.