OT and functionalism

Joan Bresnan bresnan at CSLI.STANFORD.EDU
Mon Dec 20 18:46:24 UTC 1999


Dear Harry Perridion,

Thank you very much for your remarks on Dutch phonology, which are
obviously intended for the FUNKNET audience, in view of the third
person form of address to me.

You make the point that it is not literally true that Dutch devoicing
occurs in coda position, because in that position but before b or d
voicing assimilation occurs (if there is no pause).  You comment:

> It is remarkable that OT  is capable of explaining non-existent facts.

I think you are correct that the generalization about the distribution
of voiced obstruents in Dutch is overridden by anticipatory voicing
assimilation in the particular context you note.  But you are mistaken
in concluding that this fact somehow undermines OT phonology or the
point I was using the Dutch example to make.  Phonological systems
actually have more than the three constraints I used in my example.
In particular, there are very well motivated constraints that favor
voicing assimilation.  In Dutch, such a constraint overrides the
positional markedness constraint against voicing contrasts in codas.

The reference I gave for the Dutch devoicing example in my original
message is Rene' Kager's 1999 CUP book _Optimality Theory_.  There and
in the references you will find an excellent discussion, with many
illustrations, of how the interactions of various phonological
processes can be explained more insightfully in OT than in previous
rule-based terms.

Best wishes,

Joan Bresnan


xxxxxxxxxx
> In one of her contributions to the interesting discussion on OT and
> functionalism on this list Joan Brennan wrote the following on final
> devoicing in Dutch:
>
> >>For example, Dutch and German have a voicing contrast in obstruents,
> which appears in syllable onsets--a very salient position-- but voiced
> obstruents are devoiced in syllable codas. <<
>
> This is not correct. The facts are as follows. Voiced obstruents are not
> possible before a pause, or a voiceless obstruent. The 'f' in 'leefde'
> (lived) e.g. is both syllable- and morpheme-final, but voiced, as is the
> 'b' in 'ebde weg' (ebbed away). In fact, syllable-, morpheme- and even
> wordfinal obstruents are voiced before /d/ and /b/, if there is no
> intervening pause, e.g. 'op' [p] (on) + 'doen' (do) ->  'opdoen'  (to put
> on) with [-bd-].
> What we have here is a rather simple case of voice-assimilation, that is,
> if we assume that a pause functions like a voiceless obstruent.
> CC-sequences in Dutch (C is either a stop or a fricative) are either voiced
> or voiceless. The assimilation is regressive, except when the second
> obstruent is a voiced fricative (/z/ or /v/), in which case it is progressive.
>
> >>OT can explain the
> positional neutralization of obstruent voicing in syllable codas and
> relate it to the typological of segment inventories across languages.
> The reason is simply that the *same* constraints reflecting ease of
> perception and articulatory effort are present in the grammars of all
> languages (as the universal constraint set).  No other of grammatical
> theory that I know of does this.<<
>
> It is remarkable that OT  is capable of explaining non-existent facts.
>
> >>In contrast, there appear to be few or no languages whose
> language-internal phonologies have a voicing contrast in syllable
> codas and no voicing contrast in syllable onsets.<<
>
> To a certain extent, American English is such a language: there is a
> contrast in syllable codas between voiced and voiceless stops ('bed' vs.
> 'bet') but not in onsets of syllables that follow a syllable ending in a
> vowel ('latter' vs. 'ladder').
> Such voicing of previously voiceless stops in intervocalic (but
> 'syllable-initial') position is well-attested in the history of a number of
> languages.
>
> Another fact that might disturb the serene picture of voice in the
> languages of the world sketched by some proponents of OT and/or
> typologists, comes from Scandinavian:
>
> In Southern Scandinavian (Danish, southern Swedish and southern Norwegian)
> there is no opposition between voiced and voiceless stops in final position
> after a long vowel, but in S-Swedish and S-Norwegian the merger is a voiced
> stop. Danish developed in most cases a voiced fricative (Old Danish 'ut'
> (out) -> 'ud' -> 'udh' ('dh' stands for the fricative).
>
> It is possible that these Germanic languages are all typologically weird,
> but it is more probable that the assumption that:
>
> >>Across languages, voiced obstruents
> are typologically marked: there are languages having only voiceless
> obstruents (e.g. Polynesian), and languages having both voiced and
> voiceless obstruents (e.g. English), but no (or hardly any) languages
> having only voiced obstruents.  This typological asymmetry might be
> explained in terms of constraints on articulatory effort and
> perception that shaped the evolution of languages to explain why
> voiced obstruents are more restricted crosslinguistically.  But within
> the phonological components of the grammars of particular languages we
> see the same thing: voiced obstruents are restricted in their
> *language-internal* distributions. <<
>
> is simply wrong. In fact it surprises me that respected linguists like Joan
> Bresnan (see the quotations above) and Martin Haspelmath feel obliged to
> make sweeping statements like
>
> >>Voiced obstruents are harder to pronounce than
> voiceless ones, especially in (syllable- or word-)final position.<<
>
> What evidence is there for such a statement? Is it a fact that voiced
> consonants are harder to pronounce?, in what way? and why?
>
> Let us get the facts straight before arguing about their interpretation.
>
>
> Harry Peeridon
> University of Amsterdam
>
>



More information about the Funknet mailing list