Friday, April 14, 2017

Languages in 2117

Charlie Stross, a Scottish science fiction writer, recently posted some speculations on predictions for 2117 that touch rather heavily on the domain of linguistics. Linguists who like science fiction may want to consider commenting over there; he's got some good ideas, but some elements are clearly off. The basic conclusions are:
[B]y 2117, [t]here's [g]oing to be a decline in the number of languages spoken: the main world languages will be down to English, Mandarin, Spanish, and some dialect of Arabic (Arabic is highly fragmented), plus surviving secondary languages with large bodies of adherents (over a hundred million each: for example German, Russian, Japanese).

We're also going to see the widespread deployment of deep learning driven machine translation and, most importantly, near-real-time interpretation. There'll be less reason for a native speaker of an apex language to learn other tongues [...]

And the apex languages will have changed considerably [...]

I suspect that over the next century (assuming we don't lose our technological infrastructure) current mechanisms for writing will be supplanted by newer ones--e.g. the replacement of discrete mechanical keys on keyboards with multitouch keyboards and then with gestural/swipe interfaces, where each dictionary word is replaced by a directional ideogram swiped across a QWERTY keymap, until eventually the ideogram replaces the alphabetic word or is auto-replaced by a corresponding emoji.

So: gradual obsolescence of some grammatical forms, appearance of entire new writing systems, unforseen changes due to the vagaries of machine translation, assimilation of loan words from other cultures, and the 2117 equivalent of "don't drone me, bro" (new shorthand to describe stuff that has become the new normal).

What am I overlooking?

My immediate thoughts would be:
  • Actually, a lot of languages with less than 100 million speakers each will still be around 100 years from now. Even if the Netherlands decided overnight to stop teaching, broadcasting, or providing government services in Dutch - and it won't, quite the opposite - it would take more than 100 years for the language to die out. If anything, the fragmentation of mass media into social media already makes it easier to maintain small languages, and to the extent that e-learning becomes a thing, it will have similar effects. On the other hand, only a handful of Native North American or Australian Aboriginal languages seem likely to make it as far as 2117: right now most of them are already down to elderly speakers only, and revitalization efforts are not likely to succeed without a really drastic rethinking of the school system. This is because of grossly coercive educational policies inflicted on them decades earlier. Chinese educational policy has become significantly less tolerant of minority languages over the past few years, and if that trend continues, I suspect many currently viable languages of China are likely to be in a similar situation by 2117: not yet extinct, but reduced to the point that they seem doomed. More broadly, what to predict about language survival worldwide 100 years from now depends fundamentally on two factors: how compulsory education changes, and how much of the population ends up in big cities. The former, at least, is more than anything else about political decisions.
  • Adequate machine translation does seem likely - not good enough for contexts where precision counts, but easily sufficient for casual conversation or listening to speeches. I wouldn't expect this to have any really major effects on languages, but it might allow literal translations of new idiomatic expressions to spread faster between languages.
  • Emoji are basically discourse markers: they won't become ideograms, they'll become punctuation. If they really catch on, our descendants may be as puzzled by how we get by with just half a dozen punctuation marks as we are by how people used to read with no punctuation at all.
Finally, a line that's calculated to get a lot of linguists up in arms: "[L]anguages are vanishing, and to the extent that we can only reason about things we have words for, this may be a subtle but far-reaching loss." Obviously we can reason about things we don't have words for, and equally obviously not having words for them makes it more cumbersome to talk about them. But more to the point, even where languages are in rude health, words for certain things are vanishing at a rapid pace in them. Algerian Arabic isn't going anywhere, but the vocabulary it used to have for wild plants, for traditional farming technologies, for family relationships that are only relevant in a three-generation household? I don't even think most people my age know them, much less their grandchildren in 2117. Large written languages with sufficiently developed institutions can maintain such vocabulary precariously at the margins by having specialists use it - botanists, agricultural experts, historians, etc. Most languages can't.

Sunday, April 09, 2017

Code-switching as a teaching method?

I haven't done much language teaching in my life, but as a person who likes learning new languages, I've seen a fair range of different teaching methods applied, from only speaking the target language to saying almost everything in English. But the approach used in Simon Bird's "#LilMoshom" series of Cree-teaching videos was new to me, and very interesting. Take a moment to watch some of them before reading further (especially "Respecting pisiskowak" and "Survival tips on the Rez"):

There are a lot of strong points one could comment on - the CGI, the subtitles, and the humour, for instance - but what particularly draws my attention is the way he combines the two languages. To introduce the words he's teaching, he usually speaks in English - but he doesn't just gloss, much less lecture (contrast, say, the more conventional approach used in this Ojibwe video series). When speaking in English, he throws in Cree discourse particles and sometimes even content words, gives the sentence a distinctly non-mainstream English intonation pattern which I assume reflects Cree, and even pronounces the English with a Cree accent. In different contexts, the maker of these videos speaks English like any other Canadian academic, so this appears to be a deliberate teaching strategy. The beauty of this is that, before the learner can even formulate a full sentence, they're already getting a chance to acquire some aspects of language - discourse structure and intonation - that are super-important for actually making yourself understood, yet play a minor role or get left out entirely in many traditional curriculums and textbooks (not to mention grammars!).

Have you ever encountered such a teaching method? If so, did you find it effective?

Sunday, March 26, 2017

Why it's Siwi, not Tasiwit

In English- and French-language discussions of the languages of Egypt, the Berber language of Siwa Oasis in the Western Desert is more and more often called "Tasiwit". Please, don't do this.

In Moroccan and Algerian Berber, as in the Sahel, language names are feminine, and are formed with the feminine circumfix t-...-t: Taqbaylit, Tarifit, Tamazight... In Siwi, however, languages are masculine, as in Egyptian Arabic. Ordinarily, Siwis simply call their language Siwi. When they want to specify the language as opposed to anything else from the oasis, they call it Jlan n Isiwan, "speech of Siwa/Siwis".

If you're writing in a more westerly Berber language, it's quite appropriate to nativise this term into Tasiwit. But if you do so when writing in a Western language, you're just imposing a Moroccan/Algerian convention on a language whose speakers are even less familiar with it than your readers are. On top of that, the feminine of Siwi in Siwi is Tsiwett, not Tasiwit as it would be further west. So just stick with Siwi, OK?

Wednesday, March 15, 2017

Getting from "Hey you!" to "If only"

A well-known Algerian proverb has it that:
لي عندهٌ مية يقول يا ميتين
li `andu mya yqul ya mitin
who has hundred says oh two.hundred
He who has a hundred says "If only it were two hundred!" (literally: "Oh two hundred!")

The ya here is not a general-purpose interjection. Unlike English "oh", it's normally used as a vocative, followed by the name of the person you're addressing. That's its primary function in Classical Arabic too. But in Classical Arabic, you can't use it on its own to mean "if only..."; in fact, that usage isn't very common in Algerian Arabic either. Yet the same extension of function from vocative to wish-marker is found in Algerian Berber. In an 18th century Kabyle poem recorded by Mouloud Mammeri in his Poèmes kabyles anciens (p. 132), an aspiring poet, Muh At Lemsaawd, begs the better-established Yusef u Qasi to accept him as an apprentice:

Ul-iw fellak d amaalal
A wi k-isâan d ccix is

My heart is sick for you
If only I had you as my teacher (literally: "Oh he who has you as his teacher!")

You can't do this in Classical Arabic, nor in English: a vocative followed by a noun phrase is going to be interpreted as an act of addressing, not of wishing. But in Arabic you do find an otherwise unexpected vocative particle showing up in some wish constructions, notably يا ليت yaa layta "if only". And in (slightly archaic) English you have a very similar construction with an infinitive in "to" or a prepositional phrase in "for", instead of with a noun phrase: "Oh to be young again!", "Oh for a thousand tongues to sing!" That suggests that the connection between vocative and wishing reflects some general feature of human cognition, or at least of a rather large culture area.

The obvious connection would be through requests. One reason to address someone is to ask them to bring you something. It's not such a big step from "Hey kid, get me a glass of water" to "Hey, a glass of water!", with the addressee and the verb erased, and the vocative particle effectively serving as much to mark the wish as to get the addressee's attention. But that doesn't really predict forms like the Kabyle one, where the state wished for takes the form of a relative clause, nor even the old-fashioned English constructions discussed, so I'm not really happy with this explanation. Any ideas? And can you think of any parallels in other languages?

Sunday, February 26, 2017

On Olathe

A few days ago, two unarmed young engineers from India were shot in a bar in Olathe, Kansas by a man yelling "Get out of my country!", as was a heroic bystander who tried to stop the shooter. As this contemptible crime put a normally quiet suburb of Kansas City into the international news, journalists and readers worldwide must have been wondering, as I wondered the first time I heard of it a couple of years ago: "How do you pronounce Olathe, and what sort of a name is that anyway?"

The way the locals pronounce it is /ou'leɪθʌ/, as you can hear early in the Mayor's speech. This is remarkably irregular: I can't think offhand of any other word in the English language in which a final e is pronounced /ʌ/, except occasionally "the". You might expect the etymology to provide an explanation, but it turns out to complicate the story further.

The town of Olathe was founded in 1857 by one John Barton, a doctor from Virginia, who - by his own account - got it into his head that "beautiful" would be a good name for the town he envisaged, and:

... meeting Capt. Joseph Parks, head chief of the Shawnees, he said: 'Captain, what in the Shawnee language would you call two quarters of land, all covered with wild flowers? In English we would say it was beautiful." Parks replied: "We would say it was 'Olathe,' "giving it the Indian pronunciation Olaythe, with an explosive accent on the last syllable. Barton made the same inquiry of the official interpreter, an educated Indian, who made the same reply, adding that for English use it would be best to pronounce it "Olathe," with the accent on the second syllable. So it came to pass that the new town was named "Olathe," the city beautiful. (History of Johnson County, Kansas)

In Shawnee, an Algonquian language, (h)oleθí is indeed documented as meaning "pretty" (Gatschet II:2, II:6, III:5); the root also seems to mean "good", judging from its occurrences (spelled <lafi>) in Alford's Shawnee New Testament translation, eg in Matthew 5:45, 19:16, 20:15. One might assume the Shawnees had their own name for the place, but that is not necessarily true, considering they had gotten there barely a generation earlier. Originally from Ohio, they were induced to sign a treaty to move to Kansas in 1831, onto land originally belonging to the Kaws (Kanzas). A few years after the foundation of Olathe, they were pushed out again, to Oklahoma.

It thus seems pretty clear that the original pronunciation of the town's name was /ou'leɪθi/, corresponding better with the spelling (cp. "synecdoche"). How did that turn into /ou'leɪθʌ/? I think the answer lies in English sociolinguistic variation. In the 19th century, standard English word-final /ʌ/ was often pronounced dialectally as /i/, yielding forms like "Americkee" for America or "Canadee" for Canada. In more recent times this pronuciation seems to show up mainly in caricatures of rural or Appalachian speech. The current pronunciation of Olathe as if it were Olatha can thus best be understood as a hypercorrection by people who didn't want to sound uneducated.

Update: A very helpful article linked by Y below, The Pronunciation of Missouri, reveals that the phenomenon is more systematic in the area than I had realised: it extends not only to placenames like Missouri, but even to words like spaghetti, macaroni, or prairie. This makes hypercorrection seem a less likely explanation. Instead, it looks as though final /ɪ/, which becomes /i/ in standard American English, was instead reduced to schwa in parts of the Midwest, including the area surrounding Kansas City. Andrews' (1994) Shawnee Grammar indicates that Shawnee /i/ was often realised as [ɪ], so this fits together nicely.

Friday, February 24, 2017

The Origin of Mid Vowels in Siwi

How does a language with a relatively small vowel system react to pressure from a language with a larger one?

Most northern Berber varieties have a simple four-vowel system: tense /a/, /i/, /u/, vs. lax schwa (/ə/, written e in the official orthography), the latter being mostly predictable and limited to closed syllables. In the eastern and southern Sahara, however, we tend to find slightly larger vowel systems, and it looks very much as though proto-Berber had a rather asymmetrical six-vowel system, close to modern Tuareg but missing /o/: it had tense /a/, /e/, /i/, /u/ vs. lax /ɐ/, /ə/.

Siwi Berber, in western Egypt, has a more symmetrical six-vowel system: tense /a/, /e/, /i/, /o/, /u/ vs. lax /ə/. All of these vowels occur in inherited vocabulary as well as in Arabic loanwords. It is obvious by inspection that, in almost all contexts, *ɐ merged into /ə/. But the distribution of /e/ shows little connection with that of *e: in fact, most instances of proto-Berber *e correspond to Siwi /i/. And the origin of /o/ is not immediately clear at all. How did this happen?

My latest article - written together with Marijn van Putten - proposes some answers. It turns out that proto-Berber */e/ was retained in Siwi only before word-final /n/. Most instances of /e/ and /o/ are found in Arabic loanwords. Within inherited vocabulary, almost all instances of /e/ - and all instances of /o/ - are phonetically conditioned innovations, arising from at least three distinct regular sound changes and one sporadic one. The net effect of this "conspiracy" of sound changes is to extend phonemes otherwise almost entirely restricted to Arabic loans into inherited Berber vocabulary.

If you want the full story, go read our article: The Origin of Mid Vowels in Siwi (published in Studies in African Linguistics 45:1-2 (2016), pp. 189-208).

Sunday, February 19, 2017

A real-life subjacency problem sentence

There are some kinds of questions and relative clauses that you just can't form without resorting to a resumptive pronoun, even in languages - like English - that otherwise don't allow resumptive pronouns to begin with. Ever since Ross (1967) came up with a typology of "island constraints", syntacticians have hotly debated both which ones these are and how to account for them.

Unfortunately, real-life examples of people trying to say such things are very scarce on the ground. As a result discussion of this phenomenon tends to be dominated by artificial examples. Much of the literature on subjacency inadvertently demonstrates how unsatisfactory the result can be (as discussed here: 1, 2). Every once in a long while, however, you find a completely spontaneous case of someone running up against such constraints - and here's today's, courtesy of some person on Reddit:

Step zero: find a couple million complete and utter morons, who it's a miracle they can breathe in and out without f***ing it up, to support you.

Normally, a relative clause starting in "who" would have no overt subject within the clause itself apart from "who", as in:

Step zero: find a couple million complete and utter morons, who in all honesty Ø can barely breathe in and out without f***ing it up, to support you.

But that's impossible here: note the ungrammaticality of:

*Step zero: find a couple million complete and utter morons, who it's a miracle Ø can breathe in and out without f***ing it up, to support you.

Instead, you end up having to fill the subject position to which "who" refers with a resumptive pronoun "they".