SAE phonology and grammar tests

[HazelFiver]

Remember the SAE phonology test and SAE grammar test from back in 2013? I've turned them into webpage forms that automatically calculate the score. They can be viewed here and here respectively. I don't know if you will find this useful or if someone already did it, but IIRC someone wanted something like this.

I've taken the liberty of rewording some of the questions on the grammar test to make their meaning clearer. All of these were the subject of complaints on the associated thread except for:

• verbal negation with a negative indefinite negative indefinites that preclude verbal negation (e.g. English "Nobody listened" instead of *"Nobody didn't listen")
• conjunction A, B and C "A and-B", i.e. "and" does not precede or follow the whole list and is not a suffix

On the first one, I think it's important to convey that the existence of negative indefinites isn't the issue. What Haspelmath was talking about in his paper is "NV + NI" vs. "V + NI" -- negated verb + negative indefinite vs. non-negated verb + negative indefinite, and he said the latter was the SAE structure. I also don't know how he defined negative indefinites. WALS includes words like "anybody", and Haspelmath may have as well. I don't know about anyone else, but I was under the impression that the way to not meet this condition was to not have negative indefinites at all, and that's not true.

The second one is just wrong. Haspelmath said nothing about how conjunctions are used in lists with more than two items. He said that "A and-B" was the SAE structure as opposed to "A-and B", "A-and B-and" and "A B-and". (The use of a hyphen in "A and-B" is slightly unclear given that he doesn't mean that "and" has to be a prefix, so I've added a short explanation as well.) At least one person interpreted "A, B and C" as meaning that "A and B and C" doesn't fully meet the condition, which is not at all suggested by the paper.

I've also merged the last two "one point" questions back into a single question and replaced "gerund" with "converb" and a short definition to make it clear that these are really part of the same idea.

All this confusing and misleading wording came straight out of the Wikipedia article, so we can blame whoever added it there. That article stinks IMO. I've tried to improve it a little, but it's hard. This is also where "predominantly suffixing inflectional morphology" and "nominative–accusative morphosyntactic alignment" come from, and they were unsourced in 2013 and still are. I suggest replacing them with "a prominence of anticausative verbs in inchoative-causative pairs" and "syncretism of instrumental and comitative cases", which are from Haspelmath's paper and properly sourced to it.

If you object to any of my changes, say so. I also have some thoughts on how the phonology test could be improved (my web page is unchanged from the last version by Nortaneous), and I'll post those here if there's interest.

PS -- anyone remember me? I signed up in 2016 but dropped out of conlanging for an extended period due to real life. I wasn't very interesting.

Edit: linked to the middle of the phonology thread by mistake

[Pabappa]

Thanks, this is nice.

Poswa gets a 74 for phonology. I may have misread one or two questions but they would be in the 1 point section, not the 5 point section. Khulls gets 28.5. I'll do the grammar questions later since I'd have to think more deeply to answer them, though at first blush I'd say that all of my conlangs would probably score in the low teens on the grammar test.

EDIT: Okay, the grammar test was easy. I think all my conlangs for the indefinite future will get the same score: 9%. i only checked "predominantly suffixing inflectional morphology", "conjunction "A and-B", i.e. "and" does not precede or follow the whole list and is not a suffix", and "no distinction between alienable (e.g. legal property) and inalienable (e.g. body part) possession". Most of the other questions cant even come up because i generally have no pronouns, no particles ("and" is a verb), etc, and pretty much anything that can be an inflection is (so there's tons of cases, moods, tenses, etc and theyre all fusional and stackable)

[akam chinjir]

Nice!

I agree about the quality of that wikipedia article.

Haspelmath actually has a whole book about indefinite pronouns, worth checking out. I agree with your interpretation of that point. Having distinct indefinite pronouns for negative contexts is totally normal, what's distinctive is not also needing the regular negation particle (or whatever).

I almost agree about conjunction, too, though wonder if it's worth also distinguishing "A with B" from the SAE pattern.

And I feel like "relative pronoun" could use a bit of unpacking, since the definition Haspelmath (and WALS) uses is a bit finicky, and IIRC on the old board plenty of people counted their languages as having relative pronouns when they just had a clause-initial particle (like English "that," not like English "who").

And maybe worth mentioning that (at least on the old tests) scores on the grammar test were usually significantly lower than on the phonology test---on the phonology test if you got below about 70, you were pretty un-SAE, but not really on the grammar test.

Edit: Oh, also, on #7 on the grammar test: word equivalent to "than" in comparisons of inequality instead of a preposition (e.g. English "bigger than an elephant")---that makes it sound like "than"-particles and prepositions are the only two possibilities, but there are other ones.

And #14 no marking of arguments other than the subject on the verb: as worded you get points if you've got no arguments marked on the verb, but I doubt that's intended. marking of the subject but of no other arguments on the verb, maybe?

#10 in the second part only one converb (non-finite subordinate verb), preference for finite subordinate clauses maybe should have one and only one; I'm not sure how I'd weight the two parts of the condition if I had exactly one nonfinite subordinate verb form but used it all over the place.

[bbbosborne]

nishtigian receives a 48.5 for phonology, and 23 for grammar. i assume the higher the score, the more european, right?

also, does "7 or more vowel qualities" mean 7+ vowels in total with qualities' distinctions, or 7 contrasts? like if you have length, roundedness, and creaky voice and the number of total vowels given those possible contrasts was 15 or something, then you can check the box, or if you have 3 vowel lengths, roundedness, creaky+breathy+vocal fry, and 3 tones, then you check the box?

[Aftovota]

also, does "7 or more vowel qualities" mean 7+ vowels in total with qualities' distinctions, or 7 contrasts? like if you have length, roundedness, and creaky voice and the number of total vowels given those possible contrasts was 15 or something, then you can check the box, or if you have 3 vowel lengths, roundedness, creaky+breathy+vocal fry, and 3 tones, then you check the box?

I believe it means how many distinct qualities on the height-frontness-roundness axes. Like /a i u/ is 3, /ɑ ɛ i u/ is 4, /a e i o u/ is 5, /a e i ə ɨ o u/ is 7, /ɐ ɛ e i ɔ o u/ is 7, /ɑ ɶ ø ʏ y ɛ e ɪ i o ʊ u/ is 12, etc., no matter length, phonation, nasalization and such.

[Nortaneous]

Pannonian gets either 71.5 or 82.5 depending on analysis - cf. 80 for Hungarian and 78.5 for Albanian.

My other, less serious IE lang gets 53.5, which seems reasonable given that it's in China.

[Salmoneus]

If you object to any of my changes, say so. I also have some thoughts on how the phonology test could be improved

Hello again.

It was me who originally wrote that phonology test, but I did so off the top of my head without any research or planning, and I've always meant to go back and do it properly. I even got as far as researching a bunch of the questions - comparing suggested generalities against both SAE and non-SAE languages to see which were really distinctive - but I only got a bit of the way through and got bored. I do intend to go back and finish it, but...

[mèþru]

Thank you both for making this tool!

[Vlürch]

Phenglộl got a score of 44.5 on phonology and 64 on grammar. (Hopefully no one finds it annoying that I linked to that other forum, it's just that it was where I posted about it... and I think literally 90% of the people here are also there, so...)

For phonology: I took the question about voicing distinction on "stops and fricatives" to mean "stops and/or fricatives", since it only has voicing distinction on stops but not fricatives; I checked "No non-glottal POA with only one MOA, and that MOA isn't fricative" because /ʋ/ would probably count as labial rather than specifically labiodental, and "More phonemes at a single coronal POA than at any other single POA" because the dental and alveolar consonants would probably be analysed as the same POA since there's no phonemic difference for any stops or anything except in a few affricates.

It also has barely borderline phonemic voiceless nasals and one barely borderline phonemic lateral fricative, and nasalised syllables (which would mean phonemic nasalised retroflex vowels), but I'm pretty sure all of that would be analysed as allophonic if it was a natural language, so... oh, and I checked "At least 10 vowels in total (including length and quality, but not syllable-specific tone or phonation" because it does have tone but it goes together with vowel quality in most cases.

For grammar: I honestly didn't understand half of the questions at least right now ( ), so I answered all the ones I was unsure about with either the most SAE thing or the simplest thing because that's probably exactly what I did without even realising it. One thing, though, is that verbs don't conjugate even for subject, so I'm not sure if I should've left that unchecked instead.

Yeah, I'm stupid and uneducated and conlang just for fun. Sue me.

[Frislander]

Asta got 51.5 on phonology, which seems a little strange to me, and 36 on grammar, which seems more reasonable. (Oh look I did what Vlürch did, sorry about that).

Again thanks for making this!

[Moose-tache]

There are a couple of questions that use something like "stops other than nasals." Would it be simpler to say "plosives?"
It would be nice to explain at the beginning what the test means by point of articulation. For example, would dental and alveolar count as two, or one ("coronal")?
Does the test assume language-specific rules are in place, or something more scientific and universal? For example, some European languages treat /v/ phonemically like an approximant, even though it is phonetically a fricative.

[chris_notts]

I'm not sure how to answer quite a few of these questions. Just a couple of examples, but I could highlight a lot more:

No initial velar nasal

What if your language has no velar nasal at all? Presumably you should tick, but if so this is not really a SAE feature at all. The majority of languages lack the velar nasal: in the WALS sample (https://wals.info/chapter/9) 50.1% of languages don't have this phoneme. What is common in SAE is having it but only in coda position.

Sint, my most recent conlang: 15 consonants, none of which is the velar nasal

obligatory definite and indefinite articles/suffixes

What if you have obligatory articles, but they don't mark definiteness? Plenty of verb initial languages (e.g. most Austronesian languages) have obligatory articles that really just mark nominality and/or categories like common vs proper noun. Or maybe you have articles that mark specific vs non-specific (e.g. most Salishan languages).

Sint: articles mark noun class, deixis (this/that) and specificity, inspired by Salish languages.

How should Sint score on these two categories?

[Birdlang]

Pigeonese got a 42 in phonology and a 31 for grammar. Not surprising
Bartalonian got a 40.5 in phonology and a 15 in grammar. Not surprising either that the phonology and grammar are super unSAE.
Upsaclottian got a 28.5 in phonology and a 0 in grammar. Not surprised. It’s super different grammar wise compared to European.

[ˈkuː.ɑːnˠ001]

86 for the phonology
(Not particularly surprising. One of the design criteria for it was to be reliably pronounceable by me, a native SAE speaker)
40 for the grammar
(Very much a provisional result, though, given how fluid the grammar still is)

Also:

I've also merged the last two "one point" questions back into a single question and replaced "gerund" with "converb" and a short definition to make it clear that these are really part of the same idea.

I feel like an example would also be very helpful for this question..

[mèþru]

I have a complex noun class system in some conlangs that includes masculine and feminine but also others, so what would I check for that?

[Salmoneus]

I'm not sure how to answer quite a few of these questions. Just a couple of examples, but I could highlight a lot more:

No initial velar nasal

What if your language has no velar nasal at all? Presumably you should tick, but if so this is not really a SAE feature at all. The majority of languages lack the velar nasal: in the WALS sample (https://wals.info/chapter/9) 50.1% of languages don't have this phoneme. What is common in SAE is having it but only in coda position.

Clearly, if you don't have velar nasals, then you do not have initial velar nasals. It's not a hard question.
And no, what is common in SAE is not having initial velar nasals. As that WALS map points out, most major European languages do not have phonemic velar nasals of any kind (well, most or at least very many). So having or not having /N/ is not very distinctively SAE. But having initial /N/ is extremely distinctively non-SAE. [although it does occur in Irish - I suspect they're not counting that for Reasons].

That was the point: not to specify things that only SAE had, or that all SAE had, but features where the SAE distribution was very far away from the global distribution. Initial /N/ is actually a good example - in the WALS survey, 30% of languages around the world had initial /N/, compared to 0% of European languages. So initial /N/ is a good indicator that something is not very SAE.

More generally, however, I feel I re-iterate, as I do every time this comes up: I wrote this off the top of my head, years ago, without thought or research or probably even basic spellcheck, probably late at night some time. Yes there will be questions that aren't excellently phrased, or are misguided, and other questions that are missing entirely. I do intend to write a revised, fact-based version at some point, but not immediately, and it'll be a total rewrite from scratch.

What if you have obligatory articles, but they don't mark definiteness?

I didn't write the grammar test, but again: if your articles don't mark definiteness then clearly no, you do not have definite and indefinite articles. [if they're specific vs non-specific, that's more of a judgement call. Inevitably in any situation like this, you'll have to make judgement calls, because the author can't prejudge the situation of every single language and write it into the questions. So going back to the phonology test, if a language has a fricative that's allophonically sometimes approximant, or vice versa, then there's no way I can tell you what to consider it as, you have to work that out yourself.]

Fortunately, this isn't standardised testing - there is no 'should'.
[/quote]

[mèþru]

Latest language I've worked on has a 51 on the Phonology test. It is supposed to be the parent language of a language I'm making as a "secret" personal language for someone who only knows English and Spanish, so I might have to make it way more SAE in the sound changes.

[mèþru]

And a 45 on the grammar

[Znex]

Hawntow seems to have about 26-30 on the phonology test, depending on how you analyse it (I compared a more surface analysis with my original PIE-esque building blocks structure), and then about 12 on the grammar test.

[chris_notts]

Based on best guesses / interpretation for Sint:

phonology:44
grammar: 24