Final Devoicing in Russian-Americans

Hello, I’ve just seen it yet again (“it” = the phenomenon I’m about to describe) and I am intrigued enough to use the Phonoloblog to solicit other phonologists’ views.

Russian is claimed in the research literature to have Final Devoicing; i.e. all obstruents are realized as voiceless in word-final position, irrespective of whether they are underlyingly voiced or voiceless. Yet, when I elicit these forms from Russian speakers I’ve met (usually, students in my classes), I get either partially devoiced or even fully voiced forms – certainly not neutralization of /b/ with /p/, /d/ with /t/, etc.


(a) Student A, asked to make a recording of Russian for a phonetics term paper, produces clear, measurably voiced tokens of [god] ‘year’, [ljog] ‘lay down’, and [‘gorəd] ‘city’.

(b) Student B is asked to demonstrate Russian palatalization contrasts in my class. The example words, by chance, happen to end in underlying voiced obstruents. She starts out by applying Final Devoicing, but on later repetitions produces voiced outputs instead.

I believe that the failure of the published descriptions to carry over to these speakers does not mean that the published descriptions are wrong; rather, I think it’s because the Russian speakers I’ve met are all Americans, who speak fluent English and use English continually in everyday life. English, of course, freely allows final voiced obstruents. Do you agree that this is the most reasonable hypothesis? If true, we might expect that there are no monolingual Russians, living in Russia, who fail to devoice reliably. Is this so?

If the hypothesis of L2 influence is true, I think it would count as a rather intriguing form of cross-language phonological interference. Usually we think of interference like this: “Speaker cannot say [X] in L2, because [X] is unsayable in L1”. But here, it’s “Speaker fails to say the unfaithful surface form [X] in L1, because the faithful candidate *is* sayable in (newly dominant) L2”. The part I find intriguing is that the speaker could perfectly well keep on applying final devoicing in Russian, and thus, in a sense, speak “better” Russian. Nothing in English phonology prevents her from doing so.

Here are two conjectures for what is going on. (a) It’s just an effort to realize the orthography more faithfully (Russian spells the underlying values); this predicts we would not find the effect in speakers who can’t read Russian. (b) More wildly/interestingly, we’re seeing the OO-Correspondence constraints (i.e. for voicing), which have been hypothesized to be innately highly ranked, rising up to dominate a markedness constraint ( *[-son,+vce]#) which has been weakened by exposure to L2. (For the proposed high innate ranking of OO-Correspondence, see and

Lastly: is there a literature on this topic?

Thanks in advance for your input/advice.

16 thoughts on “Final Devoicing in Russian-Americans

  1. Pavel Iosad

    Well, I can’t think of any literature offhand (I’ll pass on the link to people who know better).

    Still, (offhand again) I think there is one more possibility, that is, that when the words are carefully said in isolation, the speakers may produce a released stop (Russian stops tend to have quite strong releases, and the dental ones are quite affricated), and add a vowel segment at the end, which bleeds FD. I think, as a Russian Russian speaker, I can reproduce that.

    What do they do in word-internal clusters?

  2. Sergey Kniazev

    Hi, Bruce – it’s really nice to meet you here

    “Do you agree that this is the most reasonable hypothesis?”
    I do (although the possibility suggested by Pavel seems to be probable as well) – I notice from time to time that my Russian examples sound somewhat strange when I use them in English presentations

    “If true, we might expect that there are no monolingual Russians, living in Russia,
    who fail to devoice reliably. Is this so?”
    It’s definately so

    “is there a literature on this topic?”
    no, I guess :(

  3. Rick Wojcik

    Bruce, the study of syllable-final devoicing in Russian goes back to the beginning of phonology, with Baudouin de Courtenay’s theory of phonetic alternations. Baudouin originally looked at stem-final alternations between voiced and voiceless obstruents, and he concluded that there was a single voiced “phoneme” represented by the voiced alternant. Later phonologists, beginning with his student Shcherba, revised Baudouin’s theory of phonemics to eliminate all traces of so-called phonemic neutralization, so they called the alternants two separate phonemes. But, as you may have noticed in native speakers, there is still some detectable trace of voicing (which may manifest as a shorter duration) in the pronunciation of final voiced consonants. They are not always completely neutralized from a purely phonetic perspective.

    That obstruents tend to devoice across all human speakers is a fact of articulation. Any constriction in the mouth tends to reduce the pressure drop across the glottis, shutting down vocalization. English-speaking children have to learn to suppress this tendency in order to pronounce their language, but Russian children don’t need to in syllable-final position. So it manifests as a phonological rule because it happens to produce noticeable phonetic alternations. What you are noticing in the bilingual Russian students is their attempt to do what English-speaking children do in their earliest years, i.e. suppress the devoicing at the ends of syllables. (We often call it word-final devoicing, but it is technically syllable-final.)

  4. Rob Hagiwara

    Not sure about the phonology of L2 on L1 but there’s a small and growing phonetics literature on changes in L1 that seem to be induced by exposure/experience/dominance of L2. I’m aware of fairly recent stuff by Yeni-Komshian (with and without Flege), but I’ve only just begun to explore this area.

  5. Lisa Davidson

    There is some phonetic work on devoicing in Russian, which has shown that devoicing is not complete, though I think all of the subjects in this study spoke English to some degree (and the L2 issue was not part of their focus).

    Burton, M. B. & Robblee, K. E. A phonetic analysis of voicing assimilation in Russian.
    Journal of Phonetics 25, 97-114. 1997.

    There’s also the work on incomplete voicing in other languages, like Catalan (Charles-Luce, J. & Dinnsen, D.A. (1987) A reanalysis of Catalan devoicing, JPhon, 15, 187-190) and German (Charles-Luce, J. (1985) Word final devoicing in German: effects of phonetic and sentential contexts, JPhon, 13, 309–324)

    In a related matter, I recently finished a study examining certain acoustic characteristics that distinguish different kinds of obstruent-obstruent and obstruent-nasal consonant sequences in Russian, including #CC, C#C, and #C@C (@=schwa). We recorded both English/Russian bilinguals now in NYC (Russian L1, learned English in their teens) and Russian monolinguals in Moscow and showed that there was absolutely no difference in their production of these sequences for consonant duration, burst duration, and presence or absence of stop burst (we were interested in seeing whether bilingual Russian speakers ever get unreleased stops especially in C#C). In fact, all Russian speakers always release C1, and there are no durational differences for any of the measurements we took.

    So I don’t know if final devoicing is the same or not, but if you’re right, it would be interesting to see which phenomena are subject to L2 interference and which are not.

  6. Rick Wojcik

    Interesting subject, Lisa, although I think that we need to keep final-devoicing very distinct from regressive voice assimilation. You are probably aware of this reference, but have you looked at Barinova’s articles in Razvitie fonetiki russkogo iazyka? I own this book, but it would be hard for me to dig it out of my haystack of stored materials. If necessary, I can find the articles, scan them, and send them to you. Just let me know.

    One interesting, but connected issue, has to do with progressive voice assimilation in English–something of the reverse of the Russian and Polish phenomenon. In English borrowings of Russian name pronunciations, there are a lot of initial voiced obstruent clusters that are unpronounceable in English. So English speakers try to resolve the problem by altering the phonetic targets. Take an intial /zd/ cluster, as in Russian “zdravstvuj”. English speakers resolve the articulatory problem in one of two ways. One strategy is to insert an epenthetic /@/ between the consonents: [z@drastvuy]. The other is to devoice the initial /z/. (This strategy of devoicing initial fricatives before other obstruents is quite natural in many languages, but there is no clear basis for it in English speakers–unless you are a Stampean Natural Phonologist, of course. :)) What happens then is that English progressive voice assimilation takes over, thus rendering the pronunciation [strastvuy]. You’ll see both pronunciations quite frequently in Russian 101. Just an interesting observation.

  7. Michael Becker

    This is only marginally relevant:

    I am a teaching a course in Israel this semester, and I have two Russian speakers in my class. They speak Hebrew and English non-natively, two languages that have final voiced obstruents. When we talked about final devoicing, they both agreed that [xljep] ‘bread’ was pronounced with a “p”, not a “b”, i.e. they perceive the contrast as neutralized when asked about it.

    When a third student asked what happens when a vowel-initial word follows, we tried “bread and butter”. One speaker said that it can only be [xljep i masla], and the other speaker said that in casual speech, [xljeb i masla] is okay too.

    I wonder whether mono-lingual Russians would accept the voiced version in that context.

  8. Lisa Davidson

    Yes, it’s true that article is on assimilation. But it still may have some relevant information for Bruce, depending on how extensive the phenomenon he’s interested in is (that is, what he observed may be conditioned by assimilation too). In any case, the studies on other languages have also found that final devoicing isn’t always complete.

    Rick–your example of devoicing by the English speakers could also be a result of L2 interference. [str] is possible in English, whereas [zdr] obviously is not, so I’m not sure I think it’s accurate to say “there is no clear basis for it in English speakers”. I guess in this case I wouldn’t necessarily call it a case of assimilation.

  9. Rick Wojcik

    Lisa, I don’t think that one can argue an L2 effect here. My impression is that the devoicing of initial obstruents also produces initial clusters that are illegal in English, e.g. /vnuk/ is alternately pronounced [v@nuk] or [fnuk] by English-speaking learners of Russian. I still devoice those clusters all the time, even though I know better. BTW, this would be a good subject for a serious study–the pronunciation of initial voiced obstruent clusters in L2 acquisition of Slavic languages. My conjecture is that learners must always go through two stages to acquire many of those clusters: 1) Suppression of the epenthetic vowel, and 2) suppression of initial obstruent devoicing. Vowel epenthesis is well-attested in L2 learners. (I think that the seminal work on that may have been done by Elaine Tarone, working with ESL learners back in the 1970s.) I’m not sure about initial devoicing, but it is a common enough practice in many languages, e.g. German dialects. Of course, you’ll observe both processes frequently in L1 acquisition of Slavic languages and in motor aphasia.

  10. Pavel Iosad

    Re Michael’s comment, yes, you can say that. Not as a matter of course, but especially in fast speech the boundary gets deleted. It is clear, however, that the final consonant is resyllabified into the the onset, because in normal pronunciation, if the final consonant is voiced, the ‘i’ is pronounced as the mid high vowel, not as the front high one, which is exactly what happens if /i/ follows an onset non-palatalized consonant (with FD, the /i/ is high, as is normal in onsetless syllables)

  11. Pavel Iosad

    In the last sentence of the previous comment, “high” means “front” of course. Sorry.

  12. Jie Zhang

    There was an MA thesis in 2005 at the University of Kansas by Olga Dmitrieva (now in the Ph.D. program at Stanford) that studied near neutralization of Russian final devoicing. She recorded two groups of speakers — one in Russia, who have no knowledge of English, one in Kansas, who have different degrees of English proficiency. She found a strong correlation between the degree of neutralization and the degree of English proficiency — monolingual speakers generally exhibited complete neutralization; advance English speakers have more voicing residue. I don’t know if the thesis has been submitted for publication.

  13. Emanuel Quadros

    “What you are noticing in the bilingual Russian students is their attempt to do what English-speaking children do in their earliest years, i.e. suppress the devoicing at the ends of syllables.”

    Rick’s quote reminded me of a recent study from researchers in the Memory Control Lab, University of Oregon. They suggest that our tendency to temporarily forget L1 words during early stages of L2 learning reflects an adaptive strategy of inhibiting the access to lexical items of the first language, which are easily accessible and could distract the L2’s learner.

    In the case at stake, the process of learning the L2 phonology involves the supression of a process that is active in the L1 (devoicing). My guess is that there could be a similar strategy of inhibition here. The learner temporarily “forgets” final devoicing, even when speaking Russian, inhibiting a process that is not desired in the L2, where the voiced/voiceless contrast is crucial in final position.

  14. John Archibald

    There is starting to be more of a literature on the effects of the L2 on the L1 in a number of domains. Some, probably, beyond the interests of this group. Quickly,
    Literacy: Bournot-Trites, M. and U. Tellowitz. (2002). Report of Current Research on the Effects of Second Language Learning on First Language Literacy Skills. Commissioned by the Atlantic Provinces Education Foundation.
    Discourse: Kecskes, I. (1998). “The State of L1 Knowledge in Foreign Language Learners.” Word 49, 3, pp. 321–340.
    Pavlenko, A. and S. Jarvis. (2002). “Bidirectional Transfer.” Applied Linguistics 23, 2,pp. 190–214.
    Syntax: Jarvis, S. (2003). “Probing the Effects of the L2 on the L1: A Case Study.” In V. Cook (ed.), Effects of the Second Language on the First (Clevedon, England: Multilingual Matters), pp. 81–102.

    But back to the Phon side of things:

    Ventureyra, V., Pallier, C., & Yoo, H.-Y. (2004). “The loss of first language phonetic perception in adopted Koreans.” Journal of Neurolinguistics, 17, pp. 79-84.

    Flege, J. E. (1987). “The Production of ‘New’ and ‘Similar’ Phones in a Foreign Language: Evidence from the Effect of Equivalence Classification.” Journal of Phonetics 15, 1, pp. 47–65.

    MacKay, I. and J. Flege. (2004). “Effects of the Age of Second Language Learning on the Duration of First and Second Language Sentences: The Role of Suppression.” Applied Psycholinguistics 25, 3, pp. 373–396.

    And there is a study on the effects of orthography on the pronunciation of Italian children living in German-speaking Switzerland in a collection called “La fonologia dell’interlingua” which I can pass on if anyone is interested.

    So, knowing an L2 can affect your L1 and there are definite influences of spelling.

    My 2 cents.

  15. Andrew Nevins

    again this is not phonology, but as far as L2-effects-on-L1, speakers of many languages (ok, i have the most examples from Greek and Brazilian Portuguese) tell me that once they learn preposition-stranding in English, they start stranding prepositions in their L1 (which disallows p-stranding), as if a sudden realization dawns on them that “gee, i don’t have to pied-pipe this thing along, there’s no good reason for it!”.

Leave a Reply

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