Loading ...
Sorry, an error occurred while loading the content.

T from [?] ?

Expand Messages
  • TOms Deimonds Barvidis
    I have seen that [?] occures as an allophone of [t]. I was just wondering, is it possable that [t] comes from [?], not the other way oround? It just that
    Message 1 of 11 , Oct 1, 2009
    • 0 Attachment
      I have seen that [?] occures as an allophone of [t]. I was just wondering, is it possable that [t] comes from [?], not
      the other way oround?
      It just that Longrimol inflectional system has a lot of suffixes that have to forms, one that starts with a vowel and is
      placed after a consonant), and the other that have a /d/ before the vowel (and is attached to words ending with a
      vowel). I was wondering, how to explain the /d/ that appears. (again... it's typical for me to do something, and only
      then think - why, is that possible, and how?)
      I came to an idea that in the beginning it was a glottal not /d/, that later became /t/ (I was inspired by the fact that
      sometimes /?/ appears as an allophone of /t/ in English) and, when lenition occurred, became /d/.
      For example, the Classical Longrimol word "avaden" might be explained lie this:

      Past Adjective from Early Longrimol "aba" (fast) : aba + en>> aba?en >>abaten
      b/v/V_V
      t/d/V_V (both of these changes occur due the lenition)
      >> avaden

      Oh, by the way, I apologize if that is not correct way of noting changes, but I hope you got the point :D

      So, is this possible?
    • Roger Mills
      ... Unlikely. I think the Engl. change is due originally to the non-release of final /t/, which then spread to other (mostly pre-stop) environments (ult. _all_
      Message 2 of 11 , Oct 1, 2009
      • 0 Attachment
        --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...> wrote:

        > I have seen that [?] occures as an
        > allophone of [t]. I was just wondering, is it possable that
        > [t] comes from [?], not
        > the other way oround?

        Unlikely. I think the Engl. change is due originally to the non-release of final /t/, which then spread to other (mostly pre-stop) environments (ult. _all_ non-initial environments in some dialects--Cockney?). It's just accidental that other un-released final stops in Engl. don't do the same (probably due to the need to keep words more or less recognizable, perhaps something to do with "functional load". [?] can replace other stops in Engl. fast speech, e.g. in things like "top position", "hip boots" or "black car", "black goo" where its a case of identical stops or at least same POA.

        There are languages where all (historic) final stops reduced to [?], again most likely due to non-release. That ? then became phonemic, in that ...V# now contrasted with ...V?#.

        > It just that Longrimol inflectional system has a lot of
        > suffixes that have to forms, one that starts with a vowel
        > and is
        > placed after a consonant), and the other that have a /d/
        > before the vowel (and is attached to words ending with a
        > vowel). I was wondering, how to explain the /d/ that
        > appears. (again... it's typical for me to do something, and
        > only
        > then think - why, is that possible, and how?)
        > I came to an idea that in the beginning it was a glottal
        > not /d/, that later became /t/ (I was inspired by the fact
        > that
        > sometimes /?/ appears as an allophone of /t/ in English)
        > and, when lenition occurred, became /d/.
        > For example, the Classical Longrimol word "avaden" might be
        > explained lie this:
        >
        > Past Adjective from Early Longrimol "aba" (fast) : aba +
        > en>> aba?en >>abaten
        > b/v/V_V
        > t/d/V_V (both of these changes occur due the lenition)
        > >> avaden
        >
        > Oh, by the way, I apologize if that is not  correct
        > way of noting changes, but I hope you got the point :D

        Perhaps clearer to use the more conventional notation X > Y / B_C, or use an "arrow" X --> Y etc.
        >
        > So, is this possible?

        I would doubt it. I don't know of any case where ? > t. In the languages I know of (mainly Malayo-Polynesian), /t/ ends up as /?/ due to a series of changes--

        First k > ?, then /t/ can move back to fill the velar gap, t > k

        At a later stage, ? (from *k) > 0, then k (from *t) > ? (and, rarely, may further > 0,thus merging original *k and *t)

        Perhaps your situation could have developed in this way: [?] was originally simply a way of avoiding vowels in hiatus; then, for totally arbitrary reasons, the hiatus-avoidance feature was changed to [t], which could then undergo voicing.

        Basque IINM uses /r/ for hiatus avoidance (at least at morpheme boundaries IIRC). Buginese (Indonesia), oddly IMO, uses /w/ (between identical vowels) in those cases--

        melli 'to buy' + -i '3d sing' > melliwi 'he buys'
        bola+e 'house+def.' + -e 'this' > bolaewe 'this house' IIRC (cf. bola/e/ro 'that house')
        mita 'to see' + -a? '1st sing. > (dial.) mitawa? 'I see' (standard mitaka?)
      • TOms Deimonds Barvidis
        ... It could, I suppose. [?] was a way of avoiding hiatus in previous version also. Well... why not? Then, It just simply happened to switch [?] with [t],
        Message 3 of 11 , Oct 1, 2009
        • 0 Attachment
          Quoting "Roger Mills" <romiltz@...>:

          > Perhaps your situation could have developed in this way: [?] was
          > originally simply a way of avoiding vowels in hiatus; then, for totally
          > arbitrary reasons, the hiatus-avoidance feature was changed to [t], which
          > could then undergo voicing.

          It could, I suppose. [?] was a way of avoiding hiatus in previous version also.
          Well... why not? Then, It just simply happened to switch [?] with [t], without any hidden allophonic relationships :)

          > Basque IINM uses /r/ for hiatus avoidance (at least at morpheme
          > boundaries IIRC). Buginese (Indonesia), oddly IMO, uses /w/ (between
          > identical vowels) in those cases--
          >
          > melli 'to buy' + -i '3d sing' > melliwi 'he buys'
          > bola+e 'house+def.' + -e 'this' > bolaewe 'this house' IIRC (cf.
          > bola/e/ro 'that house')
          > mita 'to see' + -a? '1st sing. > (dial.) mitawa? 'I see' (standard
          > mitaka?)
        • Daniel Demski
          ... I had never heard IINM before! IMO, IMNSHO, IMHO, IIRC, IINM, AIUI etc. are all qualifications, expressing uncertainty. What is up with that?
          Message 4 of 11 , Oct 1, 2009
          • 0 Attachment
            > Basque IINM uses /r/ for hiatus avoidance (at least at morpheme boundaries IIRC). Buginese (Indonesia), oddly IMO, uses /w/ (between identical vowels) in those cases--
            >

            I had never heard IINM before!

            IMO, IMNSHO, IMHO, IIRC, IINM, AIUI etc. are all qualifications,
            expressing uncertainty. What is up with that?
          • Roger Mills
            ... If I m not mistaken . I guess IIRC would have been better.........
            Message 5 of 11 , Oct 1, 2009
            • 0 Attachment
              --- On Thu, 10/1/09, Daniel Demski <dranorter@...> wrote:

              > > Basque IINM uses /r/ for hiatus
              > avoidance (at least at morpheme boundaries IIRC). Buginese
              > (Indonesia), oddly IMO, uses /w/ (between identical vowels)
              > in those cases--
              > >
              >
              > I had never heard IINM before!
              >
              "If I'm not mistaken". I guess IIRC would have been better.........
            • Eric Christopherson
              Some miscellaneous thoughts. Sorry, this post has run long again. I try to condense my thoughts, but there s only so much I can do :) One of the particulars of
              Message 6 of 11 , Oct 1, 2009
              • 0 Attachment
                Some miscellaneous thoughts. Sorry, this post has run long again. I
                try to condense my thoughts, but there's only so much I can do :)

                One of the particulars of language change I am interested in is
                whether glottal stops can change into anything else.

                On Oct 1, 2009, at 11:19 AM, Roger Mills wrote:

                > --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...>
                > wrote:
                >
                >> I have seen that [?] occures as an
                >> allophone of [t]. I was just wondering, is it possable that
                >> [t] comes from [?], not
                >> the other way oround?
                >
                > Unlikely. I think the Engl. change is due originally to the non-
                > release of final /t/, which then spread to other (mostly pre-stop)
                > environments (ult. _all_ non-initial environments in some dialects--
                > Cockney?).

                How would an allophone spread from one environment to many like that?
                (I'm not doubting that it happens; I just would like to know what is
                known about the steps and motivations involved.)

                > It's just accidental that other un-released final stops in Engl.
                > don't do the same (probably due to the need to keep words more or
                > less recognizable, perhaps something to do with "functional load".
                > [?] can replace other stops in Engl. fast speech, e.g. in things
                > like "top position", "hip boots" or "black car", "black goo" where
                > its a case of identical stops or at least same POA.

                To me, it seems easier to pronounce a geminate than [?]+stop, but I'm
                open-minded about it. I would be interested in whether one were more
                likely to develop into the other.

                >
                > There are languages where all (historic) final stops reduced to [?],
                > again most likely due to non-release. That ? then became phonemic,
                > in that ...V# now contrasted with ...V?#.

                Is the following scenario plausible?
                1. Two phonemes merge into a phoneme that has both of the old sounds
                as allophones, e.g.:
                /?/ [?]
                /t/ [t]
                /?/ and /t/ > /t/ [t ~ ?].
                2. After the merger, one of the allophones goes away. Thus:
                /t/ [t]

                After that, old [?] would become [t]. It stands to reason* that #2
                could even happen before the merger is complete, leaving the
                distribution chaotic. Now, I strongly suspect that this sort of merger
                can actually happen, but I'd like to know if anyone has any evidence
                for or against it.

                (At the very least, I'm pretty sure I've read about sound changes
                which were later undone; and it doesn't seem unexpected that
                hypercorrection would sometimes drag originally distinct sounds along.)

                Conlang examples of replacement of /?/ with another consonant:
                1. A friend of mine uses a justification like the one I just described
                to explain the change /?/ > /k/ in one of his conlangs.
                2. I just found out that Takuña, one of the Akana conlangs,
                fortifies /?/ to /p/ preceding rounded vowels. <http://www.superlush.co.uk/conlanging/takunagrammar.pdf
                >
                3. In one of my languages I would like to have /?j/ become /tS)/, but
                I'm not sure how to justify that.

                I've noticed spelling of [?] as <t> in a few English interjections,
                and sometimes people even substitute [t] for [?] when pronouncing them:
                <ut uh> ["V~?(?)V~] (CXS, using " for primary stress)
                <nut uh> ["nV~?(?)V~]
                <ut oh> ["V?,(?)ou] (using , for secondary stress)

                Also, I've read that the final /p/ in <yup/yep/nope/nup> originates in
                a glottal stop, presumably becoming a [p] in <nope> due to the
                roundness of the vowel, and spreading from there to the other forms.
                Again, though, I don't know where I saw that.

                (An aside: Spanish speakers use <sip> and <nop> online. Where did
                these come from? Influence of English, or a similar process to their
                development in English?)

                I should add that it's possible for stops to develop word-finally _ex
                nihilo_ -- see <http://www.pitt.edu/~drm31/EmergenceOfObstruentsAfterHighVowels.pdf
                >. I know that doesn't help Toms, but I thought someone might find it
                interesting.

                * Regarding the phrasal abbreviations mentioned elsewhere in this
                thread, ISTR is interesting because it can stand for either "it stands
                to reason" or "I seem to remember"; context helps to disambiguate.

                >
                >> It just that Longrimol inflectional system has a lot of
                >> suffixes that have to forms, one that starts with a vowel
                >> and is
                >> placed after a consonant), and the other that have a /d/
                >> before the vowel (and is attached to words ending with a
                >> vowel). I was wondering, how to explain the /d/ that
                >> appears. (again... it's typical for me to do something, and
                >> only
                >> then think - why, is that possible, and how?)
                >> I came to an idea that in the beginning it was a glottal
                >> not /d/, that later became /t/ (I was inspired by the fact
                >> that
                >> sometimes /?/ appears as an allophone of /t/ in English)
                >> and, when lenition occurred, became /d/.
                >> For example, the Classical Longrimol word "avaden" might be
                >> explained lie this:
                >>
                >> Past Adjective from Early Longrimol "aba" (fast) : aba +
                >> en>> aba?en >>abaten
                >> b/v/V_V
                >> t/d/V_V (both of these changes occur due the lenition)
                >>>> avaden
                >>
                >> Oh, by the way, I apologize if that is not correct
                >> way of noting changes, but I hope you got the point :D
                >
                > Perhaps clearer to use the more conventional notation X > Y / B_C,
                > or use an "arrow" X --> Y etc.
                >>
                >> So, is this possible?
                >
                > I would doubt it. I don't know of any case where ? > t. In the
                > languages I know of (mainly Malayo-Polynesian), /t/ ends up as /?/
                > due to a series of changes--
                >
                > First k > ?, then /t/ can move back to fill the velar gap, t > k
                >
                > At a later stage, ? (from *k) > 0, then k (from *t) > ? (and,
                > rarely, may further > 0,thus merging original *k and *t)
                >
                > Perhaps your situation could have developed in this way: [?] was
                > originally simply a way of avoiding vowels in hiatus; then, for
                > totally arbitrary reasons, the hiatus-avoidance feature was changed
                > to [t], which could then undergo voicing.

                This brings us to a question I've been wondering lately: how do
                epenthetic/antihiatal sounds develop, anyway? To me it would seem
                unremarkable for [?] or [h] or [h\] to develop to break hiatus, or a
                glide when the vowels in contact are similar to the glide, but the
                development of [t] or [4] for that purpose, to me, begs for
                explanation. I've read that /t/ tends to be phonologically the least
                marked stop (I'm not sure if that's counting glottals or not), but I
                have a hard time imagining it popping up spontaneously. (It is quite
                common though, I think.)

                Maybe it comes from situations like this:
                1. Stems can end in vowels or a variety of consonants.
                2. Final [t] drops, leaving an alternation between suffixed forms with
                [t] and unsuffixed forms without.
                3. [t] in suffixed forms is reanalyzed as being epenthetic instead of
                as part of the stem.
                I guess what I described is basically like the (recently mentioned on
                CONLANG) -t- in French interrogatives, and the "intrusive R" in non-
                rhotic English.

                >
                > Basque IINM uses /r/ for hiatus avoidance (at least at morpheme
                > boundaries IIRC).

                Labrune hypothesizes that for Japanese /r/ too: <http://erssab.u-bordeaux3.fr/IMG/pdf/labrune_article_final_r.pdf
                >

                > Buginese (Indonesia), oddly IMO, uses /w/

                According to Wikipedia <http://en.wikipedia.org/wiki/Epenthesis>, also
                in some Japanese dialects. /w/ also occurs as the middle consonant in
                Arabic "hollow" verbs (ones with only two inherent consonants), so
                they can be used in triconsonantal templates; some cases of that might
                be etymological, but others not. (I think some other sounds do the
                same in Arabic too.)

                > (between identical vowels) in those cases--
                >
                > melli 'to buy' + -i '3d sing' > melliwi 'he buys'
                > bola+e 'house+def.' + -e 'this' > bolaewe 'this house' IIRC (cf.
                > bola/e/ro 'that house')
                > mita 'to see' + -a? '1st sing. > (dial.) mitawa? 'I see' (standard
                > mitaka?)
              • Peter Bleackley
                ... It occurred to me that such a shift could occur with the [N] vs [h] allophony that I mentioned last week. Suppose we originally have p t k ? and then k x
                Message 7 of 11 , Oct 2, 2009
                • 0 Attachment
                  staving Roger Mills:
                  > --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...>
                  > wrote:
                  >
                  >> I have seen that [?] occures as an allophone of [t]. I was just
                  >> wondering, is it possable that [t] comes from [?], not the other
                  >> way oround?
                  >


                  >
                  > I would doubt it. I don't know of any case where ? > t. In the
                  > languages I know of (mainly Malayo-Polynesian), /t/ ends up as /?/
                  > due to a series of changes--
                  >
                  > First k > ?, then /t/ can move back to fill the velar gap, t > k
                  >
                  > At a later stage, ? (from *k) > 0, then k (from *t) > ? (and, rarely,
                  > may further > 0,thus merging original *k and *t)
                  >
                  > Perhaps your situation could have developed in this way: [?] was
                  > originally simply a way of avoiding vowels in hiatus; then, for
                  > totally arbitrary reasons, the hiatus-avoidance feature was changed
                  > to [t], which could then undergo voicing.
                  >

                  It occurred to me that such a shift could occur with the [N] vs [h]
                  allophony that I mentioned last week.

                  Suppose we originally have

                  p t k ?

                  and then

                  k > x > h / V_V
                  k > g > N / otherwise

                  leaving
                  p t ?
                  We could then have
                  t > k
                  giving us
                  p k ?
                  and then
                  ? > t
                  so that we then have
                  p t k

                  Pete
                • Roger Mills
                  ... That is certainly reasonable; in fact it may have happened in an Austronesian language. ... You need a reason for /?/ /t/. It s quite a leap
                  Message 8 of 11 , Oct 2, 2009
                  • 0 Attachment
                    --- On Fri, 10/2/09, Peter Bleackley <Peter.Bleackley@...> wrote:


                    > staving Roger Mills:
                    > > --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...>
                    > > wrote:
                    > >
                    > >> I have seen that [?] occures as an allophone of
                    > [t]. I was just
                    > >> wondering, is it possable that [t] comes from [?],
                    > not the other
                    > >> way oround?
                    > >
                    > > I would doubt it. I don't know of any case where ?
                    > > t. In the
                    > > languages I know of (mainly Malayo-Polynesian), /t/
                    > ends up as /?/
                    > > due to a series of changes--

                    > >
                    >
                    > It occurred to me that such a shift could occur with the
                    > [N] vs [h] allophony that I mentioned last week.
                    >
                    > Suppose we originally have
                    >
                    > p t k ?
                    >
                    > and then
                    >
                    > k > x > h / V_V
                    > k > g > N / otherwise

                    That is certainly reasonable; in fact it may have happened in an Austronesian language.
                    >
                    > leaving
                    > p t   ?
                    > We could then have
                    > t > k
                    > giving us
                    > p   k ?
                    > and then
                    > ? > t
                    > so that we then have
                    > p t k
                    >
                    You need a reason for /?/ > /t/. It's quite a leap (phonologically) IMO, and may violate a universal/tendency for sounds to move toward the back, usually step by step.

                    Encountering a lang. such as you've proposed, I'd be inclined to look for influence from other dialects that didn't undergo the t > k shift, but did merge k and ?. Then somehow that dialect became prestigious and influenced > replaced yours. (This might be the case in a Future Hawaiian, although *here* the t-retaining dialect has little status.

                    In Conlang World, of course, All Things Are Possible :-))))
                  • Basilius
                    Hey guys, why make things so complex? %) The original task is to explain the following: On Thu, 1 Oct 2009 17:45:41 +0300, TOms Deimonds Barvidis wrote: [...]
                    Message 9 of 11 , Oct 2, 2009
                    • 0 Attachment
                      Hey guys, why make things so complex? %)

                      The original task is to explain the following:

                      On Thu, 1 Oct 2009 17:45:41 +0300, TOms Deimonds Barvidis wrote:
                      [...]
                      >It just that Longrimol inflectional system has a lot of suffixes that have
                      >to forms, one that starts with a vowel and is
                      >placed after a consonant), and the other that have a /d/ before the
                      >vowel (and is attached to words ending with a
                      >vowel). I was wondering, how to explain the /d/ that appears.

                      The situation looks quite natural(istic), and the standard (I think)
                      diachronic explanation for it would be something like this:

                      (1) there used to be many stems ending in /d/ (or some other consonant that
                      later becomes /d/ before a vowel);
                      (2) that /d/ (or whatever it was back then) was deleted word-finally and
                      before consonants;
                      (3) stems that didn't originally end in /d/ followed the same pattern by
                      analogy (e. g. since it produced more transparent forms; and it would be
                      nice to retain a few frozen exceptions);
                      (4) suffixes that begin in a vowel but don't trigger the emergence of /d/
                      just didn't begin in a vowel, originally: they had a consonant (e. g.
                      glottal stop or a glide) that later was deleted in appropriate environments;
                      (5) if there are any stems ending in /d/ and not deleting it word-finally
                      etc.: either *this* /d/ used to be a phoneme different from *that* /d/, or
                      such stems ended originally in e. g. /@/ which was later deleted in most
                      contexts.

                      And no, /?/ > /t/ doesn't look natural, to me either :)

                      --
                      Basilius

                      On Fri, 2 Oct 2009 09:36:20 -0700, Roger Mills <romiltz@...> wrote:

                      >--- On Fri, 10/2/09, Peter Bleackley <Peter.Bleackley@...> wrote:
                      >
                      >
                      >> staving Roger Mills:
                      >> > --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...>
                      >> > wrote:
                      >> >
                      >> >> I have seen that [?] occures as an allophone of
                      >> [t]. I was just
                      >> >> wondering, is it possable that [t] comes from [?],
                      >> not the other
                      >> >> way oround?
                      >> >
                      >> > I would doubt it. I don't know of any case where ?
                      >> > t. In the
                      >> > languages I know of (mainly Malayo-Polynesian), /t/
                      >> ends up as /?/
                      >> > due to a series of changes--
                      >
                      >> >
                      >>
                      >> It occurred to me that such a shift could occur with the
                      >> [N] vs [h] allophony that I mentioned last week.
                      >>
                      >> Suppose we originally have
                      >>
                      >> p t k ?
                      >>
                      >> and then
                      >>
                      >> k > x > h / V_V
                      >> k > g > N / otherwise
                      >
                      >That is certainly reasonable; in fact it may have happened in an
                      Austronesian language.
                      >>
                      >> leaving
                      >> p t ?
                      >> We could then have
                      >> t > k
                      >> giving us
                      >> p k ?
                      >> and then
                      >> ? > t
                      >> so that we then have
                      >> p t k
                      >>
                      >You need a reason for /?/ > /t/. It's quite a leap (phonologically) IMO,
                      and may violate a universal/tendency for sounds to move toward the back,
                      usually step by step.
                      >
                      >Encountering a lang. such as you've proposed, I'd be inclined to look for
                      influence from other dialects that didn't undergo the t > k shift, but did
                      merge k and ?. Then somehow that dialect became prestigious and influenced
                      > replaced yours. (This might be the case in a Future Hawaiian, although
                      *here* the t-retaining dialect has little status.
                      >
                      >In Conlang World, of course, All Things Are Possible :-))))
                    • Roger Mills
                      ... A quick reply to just this point. At least in generative phonology terms, it would be a _generalization_ or _extension_ of a rule s environment. For ex.
                      Message 10 of 11 , Oct 5, 2009
                      • 0 Attachment
                        --- On Fri, 10/2/09, Eric Christopherson <rakko@...> wrote:


                        > On Oct 1, 2009, at 11:19 AM, Roger Mills wrote:
                        >
                        > > --- On Thu, 10/1/09, TOms Deimonds Barvidis <emopunk14@...>
                        > wrote:
                        > >
                        > >> I have seen that [?] occures as an
                        > >> allophone of [t]. I was just wondering, is it
                        > possable that
                        > >> [t] comes from [?], not
                        > >> the other way oround?
                        > >
                        > > Unlikely. I think the Engl. change is due originally
                        > to the non-release of final /t/, which then spread to other
                        > (mostly pre-stop) environments (ult. _all_ non-initial
                        > environments in some dialects--Cockney?).
                        >
                        > How would an allophone spread from one environment to many
                        > like that? (I'm not doubting that it happens; I just would
                        > like to know what is known about the steps and motivations
                        > involved.)

                        A quick reply to just this point. At least in "generative phonology" terms, it would be a _generalization_ or _extension_ of a rule's environment. For ex. you have t > ? /V__# (word final), then generalized to t > ? /V__{1.C, 2.#} (any coda position), generalized even further in UK dialects to t > ? /V__{1.V, 2.C, 3.#) to include intervocalic position. This changes the whole rule for t-allophony to its most general form--

                        1a. t > t /#(s)__ (post-s/initial position)
                        --probably more specificity is needed, I believe there are word-internal/post cons. positions where t is t)

                        1b. t > ? elsewhere (all other positions)

                        Such generalizations are held to be the result of a new generation's reanalysis or imperfect learning of the rule..........I think generally you could lump it in with "analogy".

                        I'll try to address more of your interesting points later, when I can take time off from LoCoWriMo :-(((
                      • John Vertical
                        ... Austronesian language. I m not so sure about the unconditional k g N part that leaves /p t/ be. I kno of two attested mechanisms for loss of /k/ while
                        Message 11 of 11 , Oct 6, 2009
                        • 0 Attachment
                          >> It occurred to me that such a shift could occur with the
                          >> [N] vs [h] allophony that I mentioned last week.
                          >>
                          >> Suppose we originally have
                          >>
                          >> p t k ?
                          >>
                          >> and then
                          >>
                          >> k > x > h / V_V
                          >> k > g > N / otherwise
                          >
                          >That is certainly reasonable; in fact it may have happened in an
                          Austronesian language.

                          I'm not so sure about the unconditional k > g > N part that leaves /p t/ be.
                          I kno of two attested mechanisms for loss of /k/ while leaving /p t/, that
                          is > ? (as in Polynesian) and > x (as in Mongolian) - spontaneous voicing, no.

                          Technical gripe, however, as flipping the conditions should be sufficient to
                          fix that (or x > G > N / V_V).

                          For that matter, did anyone mention that time around that *h > N is
                          apparently an attested sound change? Check out
                          http://en.wikipedia.org/wiki/Nyole_language,
                          http://en.wikipedia.org/wiki/Rhinoglottophilia

                          (And yeah, since apparently *mp > p, we also have //n// + //N// > /p/! I
                          dare anyone to come up with a morphophonemic rule that makes even less sense
                          while using only single-glyph X-SAMPA symbols.)

                          ---

                          I am however still curious about the question posed by Eric C: can /?/
                          change to anything else?

                          Now yes, there are plenty of things hiatus/null onset/null coda can change
                          into (including glides, /h/, /r\/ (the English Special, tho arguably a
                          glide), and /N/ (the Samoyed Special)), and /?/ can easily become a
                          suprasegmental feature like tone or glottalization or gemination, but /?/
                          changing to another segment while remaining distinct from null would be
                          interesting. Furthermore, I should specify that I mean diachronical change
                          specifically, not synchronical (for a non-English example, there are quite
                          probably Estonian dialects where //?// + //p// > /mp/ etc, but that's
                          because /?/ comes from word-final *n.)

                          The Basque /r/-for-hiatus things was new for me too, BTW. Is this actually
                          hiatus resolution and not just something that avoided elision? (Otherwise we
                          might as well argue that French fills hiatus by /z/ and a variety of other
                          consonants.)

                          John Vertical
                        Your message has been successfully submitted and would be delivered to recipients shortly.