Discussion: Examples of loanword nativization

From Shigeto Kawahara

Dear colleagues,

I am looking for examples of the following type. In loanword adaptation, sometimes the borrower start using a structure that was not allowed in their native language. For example, Japanese voiced geminates were not allowed in native words, but nevertheless Japanese speakers started using voiced geminates in loanwords. I am looking for examples in which words with these structures are “nativized” and consequently are not allowed to have that structure any longer. For the Japanese case, when a loanword containing a voiced geminate becomes familiar enough (“nativized”), that voiced geminate is eliminated.

If anybody knows other examples of this kind, I’d love to know.

Thanks in advance,

Shigeto

Facebooktwittergoogle_plusredditpinterestlinkedinmail
Posted in Discussion topic
15 comments on “Discussion: Examples of loanword nativization
  1. Joe Pater says:

    You probably already know this paper on Russian loanwords, but just in case (it even contains a proto-OT notion of constraint strength):

    Holden, K. 1976. Assimilation Rates of Borrowings and Phonological Productivity. Language.
    https://www.jstor.org/stable/413213?seq=1#page_scan_tab_contents

  2. Brent de Chene says:

    I don’t have other examples at the moment, but I do have a rather open-ended question for Shigeto (and everyone else) about the example he cites. It is: is there any way to get a handle on the question of why repair strategies for illicit configurations are sometimes systematically different in different strata of the vocabulary?

    In the case at hand, voiced geminates in loanwords are almost invariably repaired by devoicing, as opposed to the alternative strategy of nasalizing the first half of the geminate. Thus, beddo ‘bed’ becomes betto, and *bendo, as far as I can tell, is only marginally conceivable. But in the native vocabulary, there is a well-attested alternation in mimetic words according to which predicted voiced geminates are repaired by nasalization: kosokoso alternates with kossori, but nobinobi alternates with nonbiri (*nobbiri). And lest one think either that this phenomenon is confined to mimetics or that there’s no active repair process because all the items are simply memorized (as they clearly are), there is at least one example showing productive application of the nasalization strategy in a non-mimetic sequence: sore de ‘and then’ varies with sonde, presumably through a stage *sodde (to my non-native ear, devoicing to *sotte is even more outlandish than *bendo in place of betto would be). So there do seem to be two distinct strategies, both of them to some degree productive, but segregated by stratum.

    Illicit semivowel-vowel sequences such as *ye/*wi/*we also show two repair strategies, deletion of the semivowel in the native vocabulary and vocalization of the semivowel (usually) in loanwords.

    I write as an interested observer, but would be curious to learn what is known about this issue. A quick net search suggests some attention to the coexistence of multiple repair strategies for a given sequence in a given language, but not, as far as I can see, stratally differentiated in the manner of the above examples.

  3. Joe Pater says:

    This is an interesting issue. The following paper, in which loanwords show post-nasal voicing to get rid of NT sequences, but native words have fusion / nasal substitution, is relevant I think.

    Steinbergs, Aleksandra. 1985. The Role of MSC’s in OshiKwanyama Loan Phonology. Studies in African Linguistics 16: 89–101.

  4. Shigeto Kawahara says:

    Hi Brent,

    Thanks for the comment! I think Jen Smith, who comes here from time to time, argued that Japanese native phonology favors deletion given a consonant cluster, but prefers epenthesis for loanword adaptation. Her proposal was, I think, that orthography provides explicit cues to the presence of a consonant.

    Within OT, you can posit different rankings of faithfulness constraints for different strata, which would “explain” why a particular structure is resolved in different ways. I am not sure if that is really explanatory, though.

    Back when I was a graduate student, I think Donca (Steriade) was making an argument that the strategy that is exploited in loanwords reflects “pure” phonology, driven by phonetic considerations (*minimum perceptual change*), whereas the strategy that is “morphologized” is nothing more than a historical residue. What is interesting is that it does seem to be the case that phonetically natural patterns emerge in loanword phonology (which I termed “The emergence of the phonetic naturalness”), despite the fact that the native phonology sometimes seems to provide conflicting evidence.

    For the case that you discuss, I tried to suggest in my dissertation that nasalization occurs in native words, because Japanese voiced consonants used to be pre-nasalized consonants. So at that time, pre-nasalization was a natural strategy, which became unnatural when we lost pre-nasalization. Given that it is a pure voicing contrast, devoicing is now the most natural repair. With that said, that was more than 10 years ago, I am not sure if I am happy with this explanation.

    • Joe Pater says:

      Donca actually used the OshiKwanyama case as an example of this too. I cited her paper as: Steriade, Donca. 2001. What to expect from a phonological analysis. Paper presented at the annual meeting of the Linguistic Society of America, Washington, DC in http://people.umass.edu/pater/pater-balantak.pdf.

      • Brent de Chene says:

        Many thanks, Joe. This seems eminently plausible, given that a case like /oN+pote/ [omote] involves telescoping of multiple adjustments.

    • Brent de Chene says:

      Hi, Shigeto

      Many thanks for your detailed reply.

      I suspect that, as you (and Donca) say, the older strategy in cases where there are two is typically unnatural to some extent, for example because further changes have rendered it opaque with respect to its original motivation, exactly as you suggest for the nasalization strategy in Japanese.

      Can I ask you, though, about the example /sore de/ [sonde]? Is this evidence that the nasalization strategy (which does have the advantage of preserving [+voice]) is still alive? Or is it too isolated an example? (After all, we probably wouldn’t want to say on the basis of /sumimasen/ [suimasen] ‘sorry’ that Japanese has a process of intervocalic m-deletion.)

      • Shigeto says:

        I go back and forth between whether there is a true distinction between native phonology and loanword phonology. Your example seems to suggest that there is an active component of native phonology, which yields nasalization. I’ve once thought about the vowel truncation pattern in casual speech, and your example may not be isolated; /yaru+zo/ => /yanzo/, */yasso/.

    • Brent de Chene says:

      Hi again.

      With respect to your original question, how is the following example?

      My native pronunciation of ‘garage’ is [gəɹɑ́ʒ]. This has two semi-exotic (i.e. French) features, final stress and the voiced palato-alveolar fricative. As you know, the normal British pronunciation is [gǽɹədʒ], in which both of those features have been adjusted to more English-typical values. I suspect that many other originally French words have gone through the two-step process illustrated, namely borrowing in a relatively exotic form followed by gradual nativization.

    • Paul Boersma says:

      > I think Jen Smith, who comes here from time to time, argued that Japanese native phonology favors deletion given a consonant cluster, but prefers epenthesis for loanword adaptation.

      As for an explanation of this fact, you cited Jen Smith for an “loanword=orthography” hypothesis and Donca Steriade for a “loanword=naturalness” hypothesis. However, Boersma & Hamann (2009) showed in an explicit OT analysis of a very similar case in Korean that the same single native constraint ranking that the speaker uses to produce and comprehend their native language, leads to deletion in production and epenthesis in loanword adaptation. The explanation requires positing no loanword-specific mechanisms. This extends to the Japanese case.

      • Brent de Chene says:

        Let me note that there’s reason to doubt that Japanese ever actually deals with consonant clusters by deletion. The evidence Smith adduces (as in JK Linguistics 14:65) is limited to a small number of verbal inflectional suffixes (-(r)u, -(r)eba, -(r)are-, -(y)oo, -(s)ase-). I have argued, however (JEAL 25(2016):37-80; lingbuzz/002852), that that the shorter V-initial alternants of these suffixes are the default or underlying forms; if so, clusters never arise in derivations involving them. In cases where morpheme concatenation does produce clusters, the first C assimilates to the second, but the cluster does not simplify (for examples, see note 8 of the same paper).

  5. Joe Pater says:

    Ben Gelbert has some experimental work on garage-related things, and a review of the literature (e-mail me if you want a copy):
    http://scholarworks.umass.edu/dissertations/AAI3200606/
    Also, this Simpson’s clip is directly relevant:
    https://www.youtube.com/watch?v=HXJkLBj6G1k

  6. Brent de Chene says:

    Concerning the word in question, I read today in a Japanese dictionary that “British speakers dislike the American pronunciation.” I thought it was a bit beyond the remit of a dictionary to be making that kind of observation, but after watching the Simpsons clip, I see what they were getting at.

Leave a Reply

Your email address will not be published. Required fields are marked *

*