SAE phonology and grammar tests

[bradrn]

Fun thing: a few months ago I had a small argument/discussion with zompist because he also holds this idea that a language can be said to have object agreement if object pronouns (nearly) always appear to replace an absent NP. And of course, we were discussing French. …

Trask and Millar’s Historical Linguistics (which I’m reading now) has an interesting example: Jean, il l’a achetée, la bagnole. Now, I don’t know French, but to me this certainly looks like an example of polypersonal agreement.

While I agree that in Mandarin and Cantonese it's more likely due to pronouncing most syllables with basically equal stress (aside from a few grammatical function words/affixes and the noun-marking suffixes of Mandarin), I wonder if there isn't some truth to what bradrn said. Japanese has simple phonotactics and yet it has heavily-stressed personal pronouns (part of what makes them noun-like, regarding the other thread). Does anyone know what typically happens in Polynesian and Bantu languages? It wouldn't surprise me if in both families pronouns tend to be fully stressed too (not counting the polypersonal agreement prefixes of Bantu verbs).

Don’t know about Polynesian, but Oceanic languages in general have lots of pronominal clitics, which essentially act as reduced pronouns. The subject clitics have gone a bit further and are basically agremeent markers now, but the object clitics still act as reduced pronouns. An example from NE Ambae:

Langi

wind

mo

REAL

here-gi=eu

blow-APPL=1s.O

Langi

wind

mo

REAL

here-gi

blow-APPL

i

PERS

neu

1s

*Langi

wind

mo

REAL

here-gi=eu

blow-APPL=1s.O

i

PERS

neu

1s

The wind was blowing on me.

[akam chinjir]

I'd say genitive -'s -s' still counts as a case marker...

Why not as a postposition, though? It can even appear on a verb, typically one at the end of a relative clause. A noun phrase composed of a pronoun is the only clear exception, though coordinate noun phrases (e.g. 'you and me') are near the boundaries of native speakers' competences.

"Postposition" seems right, but that doesn't mean it's not a case-marker. (Not that I have any very clear idea when you do and when you don't think of a postposition as also a case marker.)

So English has (at least?) two postpositions, "'s" and "ago."

[akam chinjir]

Trask and Millar’s Historical Linguistics (which I’m reading now) has an interesting example: Jean, il l’a achetée, la bagnole. Now, I don’t know French, but to me this certainly looks like an example of polypersonal agreement.

Note the commas, though; you can sort of get away with the same thing in English: "As for Jean, he ate it, the baguette." (Er, that's how I read it, must be hungry.) The overt arguments are somehow adjoined to the clause, it's not obviously agreement or clitic doubling or anything like that.

(There's an analysis of some so-called polysynthetic languages where NP arguments are always adjoined like this, so polypersonal agreement ends up not really being agreement, more a kind of incorporation. I think that view has fallen out of favour, though I don't know why.)

[bradrn]

Trask and Millar’s Historical Linguistics (which I’m reading now) has an interesting example: Jean, il l’a achetée, la bagnole. Now, I don’t know French, but to me this certainly looks like an example of polypersonal agreement.

Note the commas, though; you can sort of get away with the same thing in English: "As for Jean, he ate it, the baguette." (Er, that's how I read it, must be hungry.) The overt arguments are somehow adjoined to the clause, it's not obviously agreement or clitic doubling or anything like that.

This is true. As for that English example, I don’t think that’s totally acceptable for me, although you could get away with it if the object is clearly marked as an afterthought: ‘As for Jean, he ate it. —The baguette, that is.’

[Ares Land]

The interesting thing about French is that that sort of construction is very common.
Is it an example of object marking? I don't think there's really a good answer to that question.
Phonologically, je l'ai achetée is a single word.
But you can and we do say often j'ai acheté la bagnole, j'ai pris la voiture, with object marking etc.
But these still carry only one phrase stress on the last syllable ! If French was an Iroquoian language, we'd describe that as noun incorporation.
You think that's far-fetched? But as it happens, I can't add material between pris and la voiture.
*J'ai pris hier la voiture while I can say Je l'ai prise hier, la voiture.
So except for an orthography that's eight centuries out of date, j'ai pris la voiture is a single word.

All in all, though, if I had to fill a WALS survey, I'd err on the conservative side and say that French doesn't have object marking. Or noun incorporation. The formal register doesn't use those constructions much, there are plenty of cases where we drop object pronouns even in informal speech, the pronouns can still work as independant words, after a fashion...
All in all, I'd say French is undeniably SAE -- it is in fact, one of those languages that define SAE, and yet sometimes uses features we think of as very un-European. A good example of why we shouldn't take linguistic categories too seriously

[akam chinjir]

Can you say Je prends souvent la voiture?

[Ares Land]

Can you say Je prends souvent la voiture?

Ah, yes, good catch! Yes, I can. So, sorry, I can actually add material between V and O.

[quinterbeck]

Leima got 23/80

Although it's not exactly the most naturalistic conlang out there

[Hallow XIII]

Can you say Je prends souvent la voiture?

Ah, yes, good catch! Yes, I can. So, sorry, I can actually add material between V and O.

Adding on to that, what about "j'ai pris la voiture hier"?

[Ares Land]

Can you say Je prends souvent la voiture?

Ah, yes, good catch! Yes, I can. So, sorry, I can actually add material between V and O.

Adding on to that, what about "j'ai pris la voiture hier"?

Yes, absolutely.

[HazelFiver]

The original test has the same problem with the "no marking of arguments other than the subject on the verb" item. I thought it was an improvement to exclude languages with the fairly un-European situation of no person marking at all, but if English and French both have object agreement, it's not clear this should be here. I'm also confused about what it should mean to not mark objects on the verb. In fact several SAE languages are listed as having object marking on verbs by WALS (Spanish, Greek, Albanian and Hungarian).

To make it more meaningful, it could be written to specify affixes as opposed to clitics.

Meanwhile, this is how I arrived at my score for English. (This time it seems to be 88.75, not 92.5.) I think my wording is sometimes unclear, though I tried to get around this by linking to the chapters each question refers to.

More: show

Section 1

Case, tense, aspect and mood markers are all affixes or clitics This may be ambiguous. Chapter 20 doesn't seem to account for periphrastic constructions, hence English being listed as "exclusively concatenative". (The other options are "isolating", "tonal" and "ablaut".) Counting those here would also rule out every language with the have-perfect, which is an SAE structure. Maybe reword to "case, tense, aspect and mood markers are never particles, tones or ablaut".
Verbs inflect for 2-5 categories (tense, aspect, etc.) 2-3 categories
Strong preference for suffixes over other types of affix This has a very specific meaning, being defined in terms of the "prefixing/suffixing indexes" described in chapter 26. English has case suffixes (sort of), subject suffixes on verbs, tense-aspect suffixes on verbs, plural suffixes on nouns, arguably negative suffixes, and no affixes for the other listed categories. This produces a suffixing index of 8 and a prefixing index of 0, adding up to an affixing index of 8. The suffixing index is 100% of the affixing index, so English is "strongly suffixing".
Masculine/feminine-based gender system with two or three genders Listed as having three genders that are assigned semantically, but this is arguable.
Obligatory plural for all nouns
Distinction between "we" and "I", but no distinction between "inclusive we" (speaker, addressee and possibly others) and "exclusive we" (speaker and others but not addressee)
Two-way distance contrast in demonstratives, or no distance contrast at all
No distinction between pronominal and adnominal demonstratives (I don't like this vs. this book)
Differentiation between reflexive pronouns and intensifiers (e.g. German sich vs. selbst)
Person marking on verbs but not on adpositions (not counting languages with no adpositions at all)
No numeral classifiers for countable nouns (words like one glass of water)
No pronominal possessive affixes on nouns
No possessive classification (e.g. alienable vs. inalienable)
Genitives, adjectives and relative clauses highly differentiated
Separate words for "and" and "with"
Same conjunction used for noun phrases, verb phrases and clauses
Past tense with no remoteness distinctions
No optative inflection, but verbs inflect for other moods
Situational possibility marked with verbal constructions (e.g. I can open this door)
No markers of direct evidentiality
Number is marked on verbs, but no suppletion of verb forms according to number
Prepositions (adpositions precede noun phrases)
Demonstrative precedes noun
Numeral precedes noun
Degree word precedes adjective (e.g. very good)
Nominative-accusative morphosyntactic alignment I would count this even if it only applies to pronouns. Romance languages don't have case marking on nouns either AFAIK.
Person is marked on verbs, but third person is never zero-marked Not listed as having any zero marking here. I don't know what the English past tense should count as, but there are no markers for other persons either, so I don't think it's the same thing.
"Give" verb always takes a direct object and an indirect object Listed as "mixed", as it can also use a double-object construction.
No applicative construction Supposedly doesn't have any. I don't understand applicatives that well, but chapter 109 states that "Dutch and English have the double object construction but not the applicative construction". The examples given all involve some kind of special affix.
Negation with a negative particle Listed as using a particle. The full form "not" is still quite common, especially in writing. I wouldn't count something as an affix if it often acts like a particle.
Negative statements constructed the same way as affirmative statements except for the negative marker (symmetric negation) Listed as "both".
Desire expressed with a "want" verb, not a desiderative affix or particle

Section 2

No productive use of reduplication Shm-reduplication is not counted or was overlooked. It's not the kind of core feature seen in the examples on chapter 27 (e.g. plurals), and they don't list any languages as having only partial reduplication, which this is.
No associative plural
Both definite and indefinite articles, with definite article distinct from demonstrative
"T-V" politeness distinction in second person pronouns: plural is used as a polite singular (e.g. French tu vs. vous) Disappeared with the disuse of the original singular.
Syncretism of comitative and instrumental (with my friends vs. with a knife)
Distributive numerals (words with meaning like John and Bill carried three suitcases each) either marked by preceding word or completely absent
Periphrastic perfect formed with "have" verb
Prohibitive constructed like the imperative with a normal negative marker (Work! vs. Don't work!)
Epistemic possibility marked with verbal constructions (e.g. John may have arrived)
No distinction between situational and epistemic modality (You may go home now vs. You may be mistaken; You must go home now vs. You must be from Northumberland)
Suppletion of verb forms according to both tense and aspect (e.g. go vs. went) Listed as tense only.
SVO (subject-verb-object) as one of the dominant word orders
Interrogative phrases placed at the front of content questions ("wh-fronting"; e.g. Where is the cat?)
Pronominal subjects expressed with both obligatory subject pronouns and verbal affixes Feature 101 listed as "obligatory pronouns in subject position". Subject pronouns are dropped sometimes, but is this common enough to count?
Subject/agent marked on the verb, but object/patient not marked Listed as "only the A argument". I left out the detail that this refers only to person marking.
Both reflexive and non-reflexive reciprocal constructions "Distinct from reflexive"
Passive voice construction formed with a participle and a copula-like verb (e.g. I am known)
Negative indefinites that preclude verbal negation (e.g. Nobody listened instead of *Nobody didn't listen)
Polar (yes-no) questions marked with different word order
Possession expressed with a "have" verb
Predicative adjectives distinct from verbs (e.g. John is tall instead of *John talls)
Nominal and locational predication expressed the same way (John is a tailor vs. John is in Paris)
Particle (word equivalent to "than") in comparisons of inequality (e.g. bigger than an elephant)
Relative clauses constructed with inflected relative pronouns (e.g. the man who washed my car)

Probably the most dubious are gender, subject pronouns and object marking. If we flip these, the score goes down to 82.5.

[akam chinjir]

I guess the most useful way to draw the line for the purposes of this test is to distinguish languages in which there are object markers that regularly occur even in the presence of an overt object NP. You'll want to decide what to say about languages in which those object markers occur only with (say) definite or animate objects, and maybe make it explicit that the overt object NPs in question aren't dislocated ones. Something like that would at least give English, French, and German the classification they intuitively ought to get. (Depending on details Spanish could get counted as having object marking, but that's a lot less counterintuitive; in fact it's how WALS judges things.)

[KathTheDragon]

I guess the most useful way to draw the line for the purposes of this test is to distinguish languages in which there are object markers that regularly occur even in the presence of an overt object NP.

This. When there is no person marker in the presence of an overt NP, it hardly makes sense to call it agreement, as no agreeing is happening. This is where English is.

[Richard W]

Something like that would at least give English, French, and German the classification they intuitively ought to get. (Depending on details Spanish could get counted as having object marking, but that's a lot less counterintuitive; in fact it's how WALS judges things.)

Of course, written French has perfect tenses agreeing with the preceding direct object.

[HazelFiver]

Made some changes so English and French don't slip through the cracks.

• 1.1. Case, tense, aspect and mood not productively marked with particles, tones or ablaut
• 1.3. Strong preference for suffixes over other types of affix (suffixing index more than 80% of the affixing index)
• 2.1. No productive use of full-word reduplication
• 2.14. Pronominal subjects normally expressed with both subject pronouns and verbal affixes
• 2.15. Subject/agent marked on the verb, but object/patient not marked when the object/patient is a noun phrase

[Richard W]

Made some changes so English and French don't slip through the cracks.

• [
• 2.15. Subject/agent marked on the verb, but object/patient not marked when the object/patient is a noun phrase

What do "subject/agent" and "object/patient" mean? Do they mean 'agent as subject' and 'patient as object'? For example, in The dog was given a bone, is the only question whether verb agrees with a bone? Can the first "marked" be replaced by "sometimes marked" and "not marked" be replaced by "never marked"?

[HazelFiver]

Made some changes so English and French don't slip through the cracks.

• [
• 2.15. Subject/agent marked on the verb, but object/patient not marked when the object/patient is a noun phrase

What do "subject/agent" and "object/patient" mean? Do they mean 'agent as subject' and 'patient as object'? For example, in The dog was given a bone, is the only question whether verb agrees with a bone? Can the first "marked" be replaced by "sometimes marked" and "not marked" be replaced by "never marked"?

This is just how I worded it. "Subject/object" is more traditional, and the original test uses the word "subject", but chapter 102 discusses "A and P arguments". Agent/patient would be more correct here. Maybe "Verbs agree with noun-phrase agents but not NP patients" or similar.

I'm not sure about sentences like these. The verb may also agree with the dog as a promoted "subject" depending on how the language does passive voice. I would judge a language by active-voice sentences.

[Kuchigakatai]

• [
• 2.15. Subject/agent marked on the verb, but object/patient not marked when the object/patient is a noun phrase

What do "subject/agent" and "object/patient" mean? Do they mean 'agent as subject' and 'patient as object'? For example, in The dog was given a bone, is the only question whether verb agrees with a bone? Can the first "marked" be replaced by "sometimes marked" and "not marked" be replaced by "never marked"?

This is why, I think, the WALS authors also needed to introduce the concept of canonical sentences, where some specific situation would be prioritized. In this case, it'd be something like sentences in the active voice, with an animate subject, with a direct object that is directly affected, like "The dog bit the bone". What a language would do with "The dog was given a bone" would be irrelevant.

This is one of many examples of the eternal problem of how to constrain linguistic comparison in a way that the results are still meaningful and interesting.