Sunday, February 03, 2008
Metathesis everywhere
When two sounds exchange their positions (for example, clip > plik) we call it metathesis. In most languages, this doesn't seem particularly common, neither in historical changes nor in the grammar. Kwarandzie has no grammatically caused metathesis, but nonetheless is absolutely full of historically metathesised words, sometimes even coexisting with non-metathesised variants. Thus for palm spines, some speakers say taqaneft and others tanaqeft; "forget" is dnagh for some speakers, dghan for others; "irrigation channel" is variously qentret or qetrent... I've found tens of examples where either synchronic variation or transparent external comparison demonstrates metathesis (usually of non-adjacent consonants, though there are one or two cases with vowels, not counting standard North African schwa alternations), and hear new ones every couple of days. Does this remind anyone of anything they've seen, or is it just odd?
Subscribe to:
Post Comments (Atom)
19 comments:
One of the criticsms often levelled at the late Sergei Starostin's Sino-Caucasian reconstructions was exactly this: "metathesis everywhere" (in his reconstructions, proto-Sino-Tibetan very often had metathesized consonants in comparison with proto-Caucasian, or vice versa of course). Certainly this is not a "real" phenomenon in your sense, but may be the existence of cases like yours may show that Starostin's theories are not as misplaced as some suggest. But unfortunately I am not aware of any actually observed phenomena like these. Have you tried looking at Elizabeth Hume's metathesis database?
Seems like the examples you use all involve the presence of sonorants. Could this be a factor?
Hassaniya is particularly marked by metathesis, and in my time working there, particularly with Haratine, whom I lived with, I was bombarded with (and often confused by) the proliferation of metathesis. Despite the insistence of Haratine that they speak the same Hassaniya as the Beydhani, I would say the speech of Haratine is particularly marked by metathesis, and that this is probably only one of several unique features for which their speech is disparaged by Beydhan (and why I was made fun of for talking like a Haratine). One of my closest Haratani friends was particularly prone to metathesis in his speech, and this observation makes me wonder if this is somehow influenced by a vestigial influence of his ancestral language, which his family name (Camara) indicates would have been Soninké. Some of the examples which come to mind are “jowz” instead of “zowj” (though I wondered whether this was influenced by “juz’ ” - part), “saghal” instead of “ghasal,” etc... I wonder if there is a connection?
Sporadic metathesis is not uncommon as a diachronic process, as you said, eg. English aks changed to ask though the earlier form continues as a 'sub-standard' variant. Often metathesis plays a role in language play (think 'pig Latin') - could it be that the doublets you have found arose in that way?
Ed: Good point, in that most examples do seem to involve sonorants (eg girem/gimer "come first, precede"), but not all do - contrast for example tikttert vs. tiskkert "palm of hand" (the shift of t to s before k is regular, for younger speakers at least).
Khawaji: that is certainly interesting, especially given that Hassaniyya is after all right next door to Tabelbala. Can you think of any other cases? (joz doesn't really count, though - that particular metathesis covers practically half the Arab world, including most of the Levant.)
Hmm, I find metathesized forms coexisting in Siwu, too, e.g. oturi ~ otrui 'person'. Interestingly, sources from a hundred years ago also have both forms, so it's not just my ears. I'm not sure yet what's going on here, though I have the feeling it may have to do something with register or speed of speaking (there is lots of elision in rapid speech, too).
Prof. Austin: As you say, historical metathesis is certainly attested in other languages, not least English (wasp < waps is another case); but, impressionistically at least, it seems a lot less common than better-known types of sound change. In some cases, it seems plausible that speakers wanted to differentiate the term from a similar Arabic word (eg taqaneft from kernaf), especially in light of the fact that many speakers are acquainted with a secret register of the language designed to keep Arabic monolinguals from picking up the odd word; extrapolating a similar motivation back to the time when Berber rather than Arabic was the regional lingua franca could explain the cases of metathesis in words of Berber origin. Its occasional presence in words of Songhay origin is harder to explain in this manner, but in one case it may have distanced the word from a potential insult - heymu "fast", from *howmi, would have been homophonous with "shut mouth").
Cognates in Passamaquoddy-Maliseet and Mi'kmaw with fellow E. Algonquian language Penobscot exhibit gobs of unpredictable metathesis. There does not seem to be much of a sonorant effect of the type mentioned by ed, but another phonotactic factor might be plausible. Namely, that these languages permit many different kinds of initial consonant clusters, and the cues for these can be easily mistimed. In the case of the languages I work with, marginalization and fragmentation of the speech communities might reduce the opportunities for corrections to learning errors of this kind.
I was originally drawn to Maghreb-area languages in large part because because they seem to have developed along a phonotactic-typological line very similar to that seen in the northeastern-area E. Algonquian languages that I work with. Namely, what might be called a "schwa-plus-strong-vowels" system: a system in which schwa is distinct from the other vowels in its peculiar semipredictable relationship to syllabification, and a system in which iambic reduction (particularly in relation to said schwas) gives rise to complex initial clusters. And again, among the languages I work with, it's the ones that are most extreme in vowel reduction to schwa, and in iambic vowel deletions are also the ones with the most cases of glitchy historical metathesis. So again this makes me inclined to think that this increased tendency to metathesis comes from the fact that these kinds of cluster-producing vocalic/syllabic phonotactics reduce the ready clarity of relative timing of release cues (i.e. it's easy to mishear /kt/ as /tk/, but less so /kata/ as /taka/).
In other words, while we can't make happy predictive statements about specific and exceptionless mechanical application of metathesis---since it's not so---we can still make a more constrained prediction, claiming that the chance of occurrence of metathesis is greater in this kind of phonotactic system. The facts seem to support that, at least.
a secret register of the language
Verlan.
-------------
Proto-Caucasian (and Proto-Dené-Caucasian, too) had a wicked consonant inventory, and clusters of two consonants must have been common, plus, if we can trust the shaky vowel reconstructions, a schwa was present, so maybe that explains something.
That said, I haven't noticed a lot of metathesis so far (a lot of material is in the Articles and Books section of starling.rinet.ru).
On the other hand, can you give me a citation for the criticism? The Wikipedia article currently cites almost no criticism because there are so few, if any, critics that have actually looked at large amounts of material...
Hm. My first initial guess would have been nasalization - note how the the examples you cite in the original post all include a nasal consonant. What is the exact nature of the "n" sound? Then I would have asked about accent, intonation and perhaps tone. Is it possible that prosody could be at least partly responsible? And then I had the same thought Professor Austin had coupled with a vague recollection of reading something on the subject. Finally I remembered: "Parlers secrets d'el-Jadida: Notes préliminaires" by Nasser Berjaoui, EDNA 2/1997, link.
I believe you are familiar with that journal, Lameen :)
Did you ask the speakers whether they had ever employed some sort of secret language?
Also, what about the statistical distribution in relation to word length? I would expect there would be a lot more three- and foursyllabic words with metathesis then there are monosyllabic ones. Am I right?
Redcatblackcat: That sounds like a plausible and testable theory.
Anon: I'm not sure syllable is a very useful measure in north African-type languages, but they do tend to be words with 3 or more non-schwas; rarely does it apply to a simple CVC word.
One day I was feeling particularly grumpy (I drank decaf coffee) and made my own critique of Starostin's site. Be forewarned, it's a brutal attack and it may seem that I'm unfairly picking on a dead man but I couldn't help myself. I dislike Starostin's theories with a passion: How not to reconstruct a protolanguage and see a multitude of antiquated views and outright errors.
As for metathesis, and as the commenters above are mentioning, my first instinct is to look at the phonotactic patterns of the language. Definitely a first step. If you can rule out phonotactics, it may just be spurious somehow or caused by a societal element.
As redcatblackcat reveals, metathesis has an opportunity to arise en masse in languages which are the "most extreme in vowel reduction to schwa". In fact, I suspect this may have happened in Pre-Proto-Indo-European. This particular kind of metathesis is explained by the concept of "sonorancy hierarchy". When a language changes radically by such vowel deletions, there is a universal tendency to "readjust" the phonotactics of the newly created clusters so that they don't violate this hierarchy.
In this case, it doesn't sound like vowel deletion is the cause but it may have something to do with changing rules regarding legal syllabic structure... possibly?
I recently read parts of the translated foreword of the North Caucasian Etymological Dictionary; it finds regular metathesis somewhere.
Thanks for the link to "How not to reconstruct a protolanguage". I'll read it ASAP.
Tiene comes to mind (Larry Hyman's written some papers on it). It's the only language I know of with anything that approaches metathesis as 'regular' sound change. It affects coronal consonants (which seems like a goodish fit for the examples you quote too) and seems to have arisen as a change which maps particular place articulation constraints to syllable positions.
I'd be extremely hesitant about using Sino-Caucasian reconstructions as an example, simply because metathesis is an easy way of increasing the number of potential 'cognates' in the absence of reliable systematic correspondences.
I'd be extremely hesitant about using Sino-Caucasian reconstructions as an example, simply because metathesis is an easy way of increasing the number of potential 'cognates' in the absence of reliable systematic correspondences.
The vast majority of the comparanda don't involve metathesis, and systematic sound correspondences are present. Check out the Articles and Books section of starling.rinet.ru for papers and book chapters in pdf format.
And where metathesis is present, it's almost always within East Caucasian (Nakh-Daghestanian), a grouping that probably nobody doubts anymore...
There are instances of metathesis in Maghrebi Arabic as well, though not as widespread as the one you describe : in MA, their main targets are sonorants. Cf. ritla < 'litre', n3el < la3ana لعن, meṛṛawla < mellawra, mellour (from behind), semsh < shems (although this last example is related to a "confusion" of sibilants in some dialects : for ex. sajra < shajra, zuz < zuj (2), zuz < juz…).
A number of these sound shifts are described in Ph. Marçais's "Esquisse grammaticale de l'arabe maghrébin"
Post a Comment