Lo and behold, it turns out that the Hirara dialect of Japanese located on the island of Miyako shows just such a development according to Masayoshi Shibatani in Languages of Japan (1990), p.409 who offers the example of 人 hito 'person'. In this unique dialect we see the development of [pɨtu] > [pɨ̥tu] > [pˢtu] which is strikingly parallel to my Pre-IE explanation of the development of sibilantization in *h₁ḗdti. That is, Mid IE *éd̰atai ['ʔed̰ətʰəj] 'he eats' > ['ʔe.d̰ᵊ̥tʰi] (via Reduction) > early Late IE *ʔḗd̰ti ['ʔe:d̰ˢtʰi] (via Syncope).
20 Nov 2009
Japanese dialect mirrors suspected PIE development of sibilantization between two dental stops
So I was looking on the internet for something else, and as it often happens, I came across something unrelated to what I was looking for but which nonetheless had value for another problem that I pondered on several moons ago, the origin of the intervening PIE sibilant in a sequence of adjacent dental plosives *-TT- (eg. *h₁ḗdti [ʔé:d̰ˢtʰi] 'he eats')[1]. My instinct has always been to attribute it to the age of the Syncope rule when unstressed schwas were deleted. The theoretical deletion of intervening schwa between two dental stops, I reasoned, might likely have left traces of friction stemming from a devoiced vowel, lost by the latemost Proto-IE stage.
Lo and behold, it turns out that the Hirara dialect of Japanese located on the island of Miyako shows just such a development according to Masayoshi Shibatani in Languages of Japan (1990), p.409 who offers the example of 人 hito 'person'. In this unique dialect we see the development of [pɨtu] > [pɨ̥tu] > [pˢtu] which is strikingly parallel to my Pre-IE explanation of the development of sibilantization in *h₁ḗdti. That is, Mid IE *éd̰atai ['ʔed̰ətʰəj] 'he eats' > ['ʔe.d̰ᵊ̥tʰi] (via Reduction) > early Late IE *ʔḗd̰ti ['ʔe:d̰ˢtʰi] (via Syncope).
Lo and behold, it turns out that the Hirara dialect of Japanese located on the island of Miyako shows just such a development according to Masayoshi Shibatani in Languages of Japan (1990), p.409 who offers the example of 人 hito 'person'. In this unique dialect we see the development of [pɨtu] > [pɨ̥tu] > [pˢtu] which is strikingly parallel to my Pre-IE explanation of the development of sibilantization in *h₁ḗdti. That is, Mid IE *éd̰atai ['ʔed̰ətʰəj] 'he eats' > ['ʔe.d̰ᵊ̥tʰi] (via Reduction) > early Late IE *ʔḗd̰ti ['ʔe:d̰ˢtʰi] (via Syncope).
Labels:
history,
japanese,
language,
linguistics,
pie,
pre-IE,
prehistory
Subscribe to:
Post Comments (Atom)
Interesting example, but it does little to explain why IE inserts *s only between two dental stops specifically. I initially prepared for a comparision involving the much more famous sibilation *t > ts / _u.
ReplyDelete"Dissimilation" sounds like the keyword to me…
First, you overlook a key fact that both *t and *s were articulated in the same region: dental. So naturally, the surfacing of an intervening s-element is most likely to occur between two dental plosives, less likely in non-dental or non-plosive environments. If we actually explore the effects on a schwa sandwiched between two dental plosives using our very own tongue, we should notice that the schwa gains height as we shorten its duration between the stops (ie. the vowel becomes increasingly closed, synonymous with vocalic height). The vowel effectively will 'disappear' in a murky cloud of dental fricative noise (ie. sibilantization).
ReplyDeleteSecond, in the example I use from Hirara, it's important to note that the syncopated vowel is a much more closed phoneme than the Pre-IE schwa in my theory. This additional vocalic height (and thus increased closure) would then be more apt to produce lingering effects of friction in wider environments when syncopated, hence sibilantization appears even when the first stop is decisively non-dental.
It's not enough just to use 'dissimilation' without a clear description of the process firmly grounded in phonetics and acoustics. It may be called dissimilation of a sort but this hardly explains it in itself.
One more thing, in the case of historical Japanese, the change of /tu/ to /tsu/ and /ti/ to /tʃi/ is a more general phenomenon involving the lenition of dental plosives to affricates before high vowels.
ReplyDeleteTo call this process 'dissimilation' requires you to first describe in scientific terms what the originally common feature might have been between dental plosives and high vowels, otherwise it cannot be classified as dissimilatory in nature. I see no relationship at all.
The articulatory connection is clear — what is less so is why should clusters such as *[kt] and *[tk] remain. In the latter in particular, if you try retaining an independant release (so [-tʰk-]), the offglide gains a sibilant-like quality as well.
ReplyDeleteThere's another problem too: the change occurs not only between voiceless dental stops, but also when a voiced one is involved, where shwa devoicing is certainly not expected. (Nor is the change of a voiced vowel to *[z].)
By "dissimilation" I was naturally referring to the IE change, not the Japanese (to clarify, I thought you might have had an argument for the lost vowel to have been a close one). As a mechanism it isn't without problems, I admit: typically (in the examples I've seen, anyway) dissimilation leads to the substitution of another phoneme, not to the insertion of one (or the introduction of a new allophone). But it does grasp the essence of the conditioning.
Most IE languages' representation would actually be explicable starting from a spirantized intermediate *[θt], but Anatolic then has to go and torpedo that…
Tropylium: "The articulatory connection is clear — what is less so is why should clusters such as *[kt] and *[tk] remain."
ReplyDeleteThis is clear as well once we take note of the articulatory position of the schwa before deletion. Between two dental plosives, the friction from the schwa can only rationally be found in the dental region as well. In a devoiced environment, [s]-like friction is produced because [s] is nothing more than dental friction by definition. In a voiced environment, ie. between two fully voiced stops, we'd naturally expect the voiced counterpart [z].
In a word with *-tk- or *-kt-, any former intervening schwa would be caught between dental and velar articulation. In these cases then, we should rather expect alveolar friction (ie. [ʃ]-like).
Considering that *s was a PIE phoneme and that **ʃ was not, it's natural that its speakers would preserve [s]-like friction faithfully while [ʃ]-like friction was thrown away as meaningless noise.
"There's another problem too: the change occurs not only between voiceless dental stops, but also when a voiced one is involved [...]"
Yes, in theory devoicing itself would not be the cause, only the symptom. Yet, truth be told, I have trouble recalling of a PIE word with two consecutive breathy voiced stops, eg. *-dʰdʰ-, *-gʰgʰ-, *-bʰbʰ-. Can you? Keep in mind that in my phonological model of PIE, only traditionally "breathy voiced" stops exhibit uninterrupted voicing throughout the phoneme while "plain stops" are reinterpreted with marked phonation, creaky voice (ie. interrupted voice).
So while devoicing is not required for this change to work, nonetheless I can only think of examples that imply devoicing of the original schwa. Again, despite this, do not confuse the logical underlying cause with a mere symptom.
As a speaker of a language that doesn't contrast [s] and [ʃ] I can tell you that [ʃ] sounds a lot closer to [s] than to zero. That's (basically) the correspondence we see in your proposed Semitic loans in IE too.
ReplyDeleteYes, in theory devoicing itself would not be the cause, only the symptom.
What do you mean? A symptom of what? Aren't you precisely ascribing allophonic shwa devoicing as the cause of sibilant insertion?
(T)ruth be told, I have trouble recalling of a PIE word with two consecutive breathy voiced stops (…).
So have I, but if one adjacent voiceless consonant is sufficient for devoicing, shouldn't we find also patterns such as *nət > *nst, *təl > *tsl ("3" **(t)slejes ?!)
But given your comment on devoicing only being a symptom, I feel I've miss'd your explanation of what you think the actual cause is.
"Aren't you precisely ascribing allophonic shwa devoicing as the cause of sibilant insertion?"
ReplyDeleteNo, you misread my statement of correlation as a statement of causation. Devoicing is related to the friction by pure happenstance but not caused by it.
Friction is caused by the tongue, regardless of voice but it just so happens however that without an example of *-dhdh-, all examples of spirantization in PIE that I'm aware of simultaneously show devoicing.
"So have I, but if one adjacent voiceless consonant is sufficient for devoicing, shouldn't we find also patterns such as *nət > *nst, *təl > *tsl ("3" **(t)slejes ?!)"
No, because you're bounding a vowel (ie. a continuant) next to another continuant. There clearly are no such examples in PIE and I've never said anything of the kind.