Linguistic "tendencies" caused by sampling

[Chengjiang]

Lately I've been wondering if a lot of the cross-linguistic tendencies that both conlangers and historical linguists tend to take for granted aren't actually the result of certain language areas or language families making up much of the sample relevant to the tendency.

Here's an example: It's been said, especially as an argument in favor of glottalic-hypothesis PIE or pre-PIE, that languages with a series of ejective consonants frequently lack a bilabial ejective, and that the bilabial ejective is the most likely to be absent. Wikipedia asserts this on its page about ejectives, for instance. However, most of the languages I'm aware of that have ejectives but lack a bilabial ejective fall into the following groups:

a) Na-Dené languages, which entire family does not distinguish manners of articulation on the bilabial stop that are distinguished on stops and affricates at other points of articulation (these make up the majority right there)
b) Languages with proportionally few labial consonants in general (mostly also found in North America)

I'm aware of plenty of other language groups where ejectives are common (all three Caucasian families, southern Bantu, "Khoisan", Pacific Northwest sprachbund, Maya languages, Aymara, Quechua, various Afro-Asiatic languages), and most of their members appear to have a full series of ejectives across most or all POAs, including bilabial. Hell, despite Proto-Semitic not having a bilabial emphatic stop, its descendants that realize emphatics as ejectives have often closed the gap by innovating a bilabial ejective. So unless there are a hell of a lot of groups I'm not aware of with this gap in the ejectives, I'm tempted to say that the "bilabial ejectives are rare" tendency doesn't hold much water.

Any comments on this idea? Any other tendencies you're aware of that seem dubiously based on a particular sample of languages?

[CatDoom]

Retroflex ejectives are extremely uncommon, although they're not particularly difficult to articulate or to perceive. They do, however, occur in a number of languges in California and the Pacific Northwest. While I haven't made an effort to seek out counterexamples, I can't think of any languages that have both retroflex stops and ejectives but lack retroflex ejectives, so it may be that the west coast of North America is simply the only area (or one of very few areas) where those two features co-occur.

[Chengjiang]

Here's one I've heard stated a lot on this board, both before and after my hiatus:

A language will have as phonemes at least three of /p t k ʔ/.

Most often I hear this stated not even as a tendency, but as an absolute rule. I am not convinced that there is anything about not having /p/, the most likely member of /p t k/ to be missing, that makes having /ʔ/ particularly more likely. For one thing, although it's more common in some areas than others, the unconditional or near-unconditional shift [p] > [f] or [ɸ] (with optional later stages such as [f] or [ɸ] >[h] > [Ø]) appears to be fairly common cross-linguistically, regardless of whether the language in question possesses /ʔ/. For instance, a shift of [p] to [ɸ] is reconstructed for Proto-Celtic and Old Japanese (and thence to [Ø] in the former and [h] in the latter), neither of which had /ʔ/. Fijian currently has only /t k/ for voiceless stops unless one counts fairly recent loanwords. So do Nubian and Songhai. I haven't encountered this one yet in actual linguistic publications, and I kind of suspect it's an artifact of certain more well-known languages that lack one of /p t k/ (e.g. Hawai'ian, Arabic) having /ʔ/.

[finlay]

I don't think japanese ever actually lost p though. It still shows up in geminates and after n

[Chengjiang]

I don't think japanese ever actually lost p though. It still shows up in geminates and after n

True. Before Japanese regained /p/ from Chinese and other loanwords, however, [p] could reasonably be regarded as an allophone of /ɸ/. (I think the re-phonemicization of [p] happened before the [ɸ] > [h] shift; correct me if I'm wrong.) Of course, [ɸ] could also be treated as an allophone of a phoneme /p/.

[Xephyr]

I do not understand/agree with your rationale for your excluding Na-Dene and other North American languages. In the former case, I literally don't understand what you're saying-- it sounds like you're pretty much saying "they don't count because their labial series lack phonation distinctions" which I thought was what we were trying to measure in the first place. In the latter case, sure you could say there are areal reasons for them lacking labials, but couldn't you say the same for the other language families-- that the reason they don't lack labials is because they aren't in linguistic areas that disfavor labials? I know that sounds kind of dumb, but what I'm saying is, you can't state a linguistic tendency and then justify it by arbitrarily excluding languages that don't exhibit that tendency, and you are never going to find a language or family that isn't affected by other forces which determine its typology (which is what I think is meant by the saying that "If there's a nuclear war tomorrow, all of a sudden OVS is the default word order in human language")-- the Platonic Ideal of a language doesn't exist.

[clawgrip]

I don't think japanese ever actually lost p though. It still shows up in geminates and after n

True. Before Japanese regained /p/ from Chinese and other loanwords, however, [p] could reasonably be regarded as an allophone of /ɸ/. (I think the re-phonemicization of [p] happened before the [ɸ] > [h] shift; correct me if I'm wrong.) Of course, [ɸ] could also be treated as an allophone of a phoneme /p/.

From what I know, in older Japanese, [p] could only occur after /ɴ/ and /Q/ (gemination, i.e. っ), while [ɸ] occurred elsewhere. Since the environment for [p] is more restricted than that for [ɸ], it makes sense to refer to the phoneme as /ɸ/. But finlay is right that [p] was never lost, as far as I know. So you could say that Old Japanese lacked /p/, but you can't say that it lacked [p]. Not sure how that affects your theory.

[Chengjiang]

I do not understand/agree with your rationale for your excluding Na-Dene and other North American languages. In the former case, I literally don't understand what you're saying-- it sounds like you're pretty much saying "they don't count because their labial series lack phonation distinctions" which I thought was what we were trying to measure in the first place.

It's not that. I'm not saying they don't count, for any reason. I'm saying that there's very little outside that family that shows such a tendency, and that therefore the tendency for /p'/ to be absent looks like it's based on overrepresentation of that family. If there were more separate groups that had ejectives but lacked /p'/, I'd think the tendency was pretty sound, but the fact that most of the examples seem to be from one family/area makes me think it's a feature of that family/area rather than a tendency in languages in general.

[Chengjiang]

So you could say that Old Japanese lacked /p/, but you can't say that it lacked [p]. Not sure how that affects your theory.

I'm aware of the environments in which [p] was retained. The "at least three of /p t k ʔ/" meme, when I've heard it, has almost always been phrased in terms of phonemes, not phones.

[finlay]

it's situations like that that make me think that phonemes are a pile of wank, though.

[Chengjiang]

it's situations like that that make me think that phonemes are a pile of wank, though.

Like what?

[Nortaneous]

Here's one I've heard stated a lot on this board, both before and after my hiatus:

A language will have as phonemes at least three of /p t k ʔ/.

Most often I hear this stated not even as a tendency, but as an absolute rule.

That meme is long gone -- someone noticed Aleut, so it got rewritten to "at least three plosive POAs" or something like that, but Abau has no other plosives than /p k/ and Northwest Mekeo may-or-may-not have no other plosives than /p b k g/. (I made a database for questions like these! Use it!)

I once saw a paper arguing that there was a language that had no stops but /?/, with everything else being fricatives allophonically realized as stops when next to /?/, but I have no idea where that paper is or what the language was.

I do not understand/agree with your rationale for your excluding Na-Dene and other North American languages.

It makes sense -- what it's saying is that the observed tendency for /p'/ to be less likely than other ejectives is [near-?]fully explained by an areal feature. So it's a statistical artifact, like the rarity of retroflex ejectives, or the correlation between a two-way Pʰ-P contrast and front rounded vowels* or whatever, not a real crosslinguistic tendency like "there won't be more nasal vowels than oral vowels, and it's likely that there will be fewer" or "front voiceless and back voiced plosive gaps are more likely than vice versa".

* I completely pulled this out of my ass. I don't know if it exists.

[Chengjiang]

(I made a database for questions like these! Use it!)

Good to know. By "questions like these", do you mean questions concerning cross-linguistic tendencies/universals?

[Ketumak]

I've seen discussions of word order correlations in verb-initial languages were conclusions drawn were qualified with reference to sample size.

Matthew Dryer has used very large samples to disprove somr thesies about word-order correlations. He also looked at sample composition. IIRC he used 400 languages weighted for family and geographical area.

[Hydroeccentricity]

I don't think it's unlikely that something like this would happen. I mean, in the Elman Service days of anthropology, people were making all kinds of conclusions about the development of human societies that turned out to just be about Polynesia. The part of the world that provided the neatest, simplest laboratory for them to test their conclusions turned out to be the main source of their data. Shocking!

Something similar might happen with linguistics, and it's even more likely to happen to conlangers who all have their own favorite languages (anyone up for making a collaborative Welsh/Basque/Piraha hybrid later?). The main thing is, as Ketumak said, just making sure that your conclusions are qualified based on statistical concerns.

[Nortaneous]

Chengjiang: No, consonant inventory patterns -- more specifically, questions involving languages with fewer consonant phonemes than Finnish. (I wouldn't expect a language with >12 consonants to have fewer than three plosives.)

[din]

And where can I find this database?

[KathTheDragon]

This is the thread.

[Chengjiang]

This is the thread.

Oh, yes, that thread. I remember that. That is an excellent thread.

[din]

Ah yes, I've seen that before. Thanks

[Seirios]

I don't think japanese ever actually lost p though. It still shows up in geminates and after n

True. Before Japanese regained /p/ from Chinese and other loanwords, however, [p] could reasonably be regarded as an allophone of /ɸ/. (I think the re-phonemicization of [p] happened before the [ɸ] > [h] shift; correct me if I'm wrong.) Of course, [ɸ] could also be treated as an allophone of a phoneme /p/.

Allow me to digress. I remember in Modern Japanese there's a minimal "pair" of h-p-b : harahara, parapara, barabara, also identical in accent. They're all onomatopoeias, but in Japanese lots of onomatopoeias also serve as standard adverbs, so are these three. Anybody know when such contrast was re-introduced to/re-formed in Japanese?