Artikelen / over Dialectologie

A Triggered Nasalization [i]

(translation from the orignal article "Afgedwongen nasalering" in Tijdschrift voor Nederlandse taal- en letterkunde, deel 110 (1994), blz.55-67)
 

In several languages - I would venture to say: in most lan­guages - a vowel followed by a tautosyllabic nasal consonant is or becomes nasalized. I will mention, only as an example, nasalization in French:  un bon vin blanc. In Dutch this kind of nasalization, with a complete lost of the nasal consonant, is rather rare. More important, in Standard Dutch as well as in a number of Dutch dialects, especially those of the sou­thern part of the language area, a type of nasalization occurs by which a short vowel is nasalized when it is followed by a tautosyllabic nasal consonant, followed by another consonant.

The map shows the area of the southern Dutch dia­lects and the spread of this nasalization in one single word, brand, which means 'fire'. In the places marked with a dot the vowel is not nasalized, in the places with short dashes it is, and the nasal consonant is more or less shortened. Nasaliza­tion appears in the provinces Zeeland, West- and East-Flan­ders, the western part of North-Brabant and Antwerp. The spread of nasalization, in other words, differs only in de­tails from the one in the word brand..

There are two aspects of this southern Dutch nasaliza­tion which makes it very interesting. First of all, nasaliza­tion only takes place if the nasal consonant is followed by (at least) one consonant in the same syllable. Some of the following consonants are morphological or syntactic, clitics for example, others are part of the non-complex word. See the examples in (2); the examples are arranged according to the place of articulation of the consonants.

      (1) a. V [+ant][+ant]       

             lõ:mp ('stupid'); ã:mbellen ('to ring')

          b. V [+cor][+cor]

             kã:ns ('chance'); hoe kã:n da? ('how is that pos­sible?')

          c. V [+back][+back]

             bã:ngk ('bank'); dat kõ:ngk ook ('I could that also')

A second point is that the vowel which becomes nasa­lized, is lengthened, at least not shortened. This seems to be in conflict with universal principles especially when the second consonant is added by a morphological rule. In general the syllables of a language have some kind of absolute length or weight. This means that there will be compensation or correction somewhere within the syllable when it is enlarged with a phonological element, or a combination of such ele­ments, for example by the adjunction of a morpheme. Usually in such cases the vowel is shortened.

This is more or less similar to what happens when one consonant is joined to another. Nooteboom-Cohen (1984: 131) have worked out that in such cases the length of the cluster is less than the sum of the length of both of them. Cf. also Zwaardemaker-Eykman 1928: 300-301).

Shortening of vowels in a morphological context is very common in the southern Dutch dialects. There are different situations in which such shortening appears. I give here only a brief survey of a number of instances from the dialect of the western part of North-Brabant, as examples. In other dialects the situation is more or less the same. See nr. (3).

(2) Shortening

a)                          by adding a morphological t:

ik [kεik]  ('I look')  <->    hij [kεkt] ('he looks')

ik [pra:t] ('I talk')         hij [prot] ('he talks')

ik [we:t]  ('I know')         hij [wIt] ('he knows')

b)                          by adding a morphological s:

[tεit] ('time')        <->    in [tεts] ('in time')

ik [lo:p] ('I run')           [lops] ('in heat')

hij is een veel-[pra:t]       hij heeft veel [prots]

('he talks too much')         ('he has too much talks')

c) with a suffix of more than one segment ('a diminutive')

[wεif]  ('wife')       <->    [wεfke] ('little wife')

[stra:t] ('street')           [strotje] ('little street')

[be:k] ('brook')              [bIkske] ('little brook')

The left column in (2) contains forms with a long vowel or a diphthong; they are the basic forms of the words. In the right column a suffix is added. In (2a) it concerns the suffix of the third person of the present singular. In the right column of (2b) there are forms with a morphological s, which repre­sents in some cases a genetive-marker, in others it is adver­bial. But here too adding a consonant produces a shor­tened vowel. In (2c) we have three diminutives from words with a long vowel. Besides these forms, which are the results of morpho-phonological rules, there a lot of words, which have there shortened vowels lexicalised.

The phenomenon that I want to discuss in this paper is totally in conflict with the shortening just described. in this case a syllable or a word gets a lengthened vowel when a morphological -s or -t is added, or when in a given structure a second consonant follows the nasal consonant. The latter, which also can be seen as lexicalised, are of course the result of historical changes, as the un-changed forms from Standard Dutch and other dialects illustrate. It is clear that there must be some important factor, which causes this anoma­ly, not only because enlarging on the one hand (by a con­so­nant) and at the same time lenthening on the other ought to exclude each other, but also because the resulting syl­lables seem to be to the principles of the syllable-struc­ture of Dutch. (4a) shows the normal rime structure of the syllable and (4b) a disallowed variant (Trommelen 1984, 152) . Never­theless the latter seems exactly the structure of the nasa­lized words in ques­tion.

(3a) rime: [-cons] [+son][+cons]([-son/+cor])

(3b) rime: *[-cons][-cons][+son]([-son])

In (4) a summary is given of all types of nasalization involving a short vowel before a cluster of nasal consonant and another consonant.

(4) NASALIZATION

(a)   before a cluster [+ant]+[ant]: m + f, v, p, or b

      Dutch      Dialect          Translation

      kamfer     kã:(m)$fer       camphor

      lamp       lã:(m)p          lamp

      lomp       lõ:(m)p          stupid

(b)   before a cluster [+cor]+[+cor]: n + s, z, t or d

      kant       kã:(n)t          side

      bont       bõ:(n)t          fur

      kinds k«:(n)+s          doting

      tanden     tã:(n)$de        teeth

      kan ze?    kã:(n)se         can she?

      kan ie?    kã:(n)tie        can he? 

(c)   before a cluster [+back]+[+back]: ng + Χ, g, k or g

      stangk     stã:(ng)k        stench

      hingkelen  h«:(ng)kelen      to limp

      kan ik?    kã:(ng)k         can I?

(d)   special case:

      langs ('along'): langs -> lanst -> lã:(n)st

In the left column of (4) the Dutch equivalents of the dialect words are given. The tilde across the vowel of the dialect words indicates that the vowel is nasa­lized; the colon behind the vowel indicates that the vowel is lengthened. I have placed the nasal consonants between brack­ets because they have been reduced. I will now discuss the forms in (4) in more detail.

(4a) contains forms with a labial cluster; example kamfer. These latter form also il­lustrates a new fact in that that the syllable boundary plays no role. Apparent nasaliza­tion is triggered by tautosyllabic second consonant as well as by a hetero­syllabic one.

(4b) concerns words with a dental cluster. The vowels involved with nasalization are mostly [+low] vowels, but there are a few words especially with a [+high] vowel before s which are also nasalized: k«:(n)s, for example. This word is deri­ved from kind ('child') by a suffix -s,; then the dental stop is deleted, yielding the cluster ns preceded by a nasa­lized vowel. The under­lined forms have a second consonant which represents the first part of a clitic. Here too the syllable boundary ap­pa­rently has no influence. In words (syllables) with a (homorganic) voiced stop nasalization fails to come, unless the cluster is preceeded by a [+low] vowel: [tã:(n)de] (teeth) versus [bonde] (bound).

(4c) contains velar clusters [+back]; the nasal conso­nant is written as ng. The last form of this group deserves special attention. First, it is also a form with a clitic, but secondly, this form must result from the application of at least two dif­ferent rules. A place-assimilation rule, which changes a velar nasal into a dental, and then the nasaliza­tion-rule under discussion here.

In (4d) we can see another example of this two-fold operation. The original velar nasal of langs assimilates to the dental [+cor] s. The result is an homorganic cluster, and then nasalizing takes place.

The fact that nasalization only takes place before homorganic clusters, no doubt offers the key to the solution of the problem of the unexpected lengthening of the nasalized vowels.

One step towards the solution of the problem I found in a rather old article by the phonetician prof. Grootaers from the University of Leuven (Belgium). In this article from 1946, which is partly a repeat of a much older investigation by Colinet (1896), Grootaers reports about his comparison of the French nasalization and the one in the dialect of the city of Antwerp (Belgium). His conclusion was that in French words like un bon vin blanc no trace of the nasal consonant is left, while in the Antwerpian cases there always remains some degree of consonantal closure. Another observation was that n was more reduced before s than it was before t, and last but not least: the greatest reduction takes place in heterosyl­labic clus­ters, like in dã:(n)se (to dance). Those observations seem to aggree with the ones by James Foley that nasalization is more likely to occur before con­tinuants than before stops (Foley 1977, 61).

Grootaers' investigation provides argu­ments to study this type of nasalization from an autosegmental perspec­tive. Moreover, nasalization has been a favorite topic in autoseg­mental studies. A description of the phenomena studied here has to establish a link between the nasalization of the vowel and the reduction of the consonant. In the traditional his­torical grammar the phenomena of simultaneous vowel leng­then­ing and consonant reduction was called 'Ersatzdehnung', which in my opinion is a very correct name (cfr. a different opini­on in Van Bree 1987, 146). Applied to the process studied here, this means that the nasalized vowel is lengthened to the extent that the consonant is redu­ced, which is exactly what happens.

Let us, therefore, turn to the phonological properties of the nasal stop. In fact the feature nasal is the only element which distinguish a nasal stop from the following consonant in a homorganic cluster. I believe that it is just the homorganicity of the cluster which causes the shortening of the nasal consonant and, at the same time, the lengthening and the nasaliza­tion of the vowel. By homorganic I mean that both consonants, the nasal and the other, have the same place of articula­tion.

I will now illustrate what - phonetically - happens, with the help of the word mans, which is a derivation of man, and means 'brave' or something like that. In (5a) I give the representation of both words, in (6b) the representation of mans in a 'stretched out' way, before and after the nasa­liza­tion.

      (5a)  N                      N

            |                      |

          man  ('man')           man+s ('a man's man','bra­ve')   

      (5b)  'stretched out':

            NNN               NNN                NNN

            |||   ->          |||      ->        |||

       m/aaannn/s        m/aaaann/s         m/aa­aaan/s

The non-complex man, meaning 'man' and the adjective derived from it by a suffix-­s, mans, meaning something like 'brave'. The feature [+nasal] is presented here as an autoseg­ment, symbolized by the capital N. Primarily nasality is only linked to the con­sonant n, in both words, man and mans. The word man has remai­ned the same, in Dutch as well as in the dialects, whereas mans has changed, in the way I have shown in (5b). For the sake of clarity I have as it were stretched out the vowel and the nasal consonant. Three symbols a stand for a short a, three symbols n for a normal a, and three capital N's for the nasal-autosegment of normal proportions, the length of which is identical to that of the original consonant n. In the first phase of the process the con­sonant n is shor­tened a little bit, while the vowel is leng­thened proportionally. The nasal auto­segment remains the same, be it partly attached to the vowel as well as to the consonant. In the second phase both proces­ses are progressing, but not to the extent that the con­sonant would totally disap­pears.

A very important aspect of the change is that the origi­nal short vowel is not only lengthened, which is a matter of quantity, but at the same time has acquired another quality, which is illustrated by a matter of quan­tity: the vowel of mans is extensible without any (reasonable) limit, the same as all [+tense]-vowels in Dutch and its dia­lects: "wat ben je [la:t]", and "hij is veel [ma:ns]. So the lengthened nasalized vowels in fact are become [+tense] vowels.

As is shown in (5b) phoneti­cally speaking the process of nasali­z­ation may be gradual­ly, being the nasal sometimes only partly dele­ted. According to Groot­aers (1946, 136)  (as I said before)  there is another phonetic difference between nasali­za­tion before a cluster -ns  and a cluster -nt, because in the latter case there is less reduction of the nasal con­sonant. The same kind of difference he has observed between tautosyl­labic clusters and heterosyl­labic ones: in an hetero­syllabic cluster the nasal consonant is more reduced.

Now I will present those facts in an autosegmental representation. According to the autosegmental theory I make a distinc­tion between different segmental tiers. In this case I use a sim­plified model, existing of only three tiers: the CV-tier or 'timing-point' tier, representing timing units, the

velum-segment (((tier??)) (+ or - nasal), and the tier which concerns the place of articulation.

      (6a)  'velum'     [-nas]  [+nas]

                          |       |

           'skelet'       V       C

                          |       |

           'place'     [+low]  [+dent]

                    m     a       n

      (6b)  'velum'      [-nas]  [+nas]  [-nas]

                           |       |       |

           'skelet'        V       C       C

                           |       |       |

           'place'      [+low]  [+dent] [+dent]

                     m     a       n       s

                    [-nas]  [+nas]  [-nas]

                       ╪    /  |       |

      ->                   /
                          / 
                       V       C       C

                             /
                       |    /  ╪       |

                           /                   
                    [+low]  [+dent] [+dent]

                 m     ã       ã       s     

In (6a) and the first state of (6b) both types of features ('velum' and 'place') are directly linked with one slot, C or V. The second state of (6b) represents the normal shape of words like mans, the result of a few connected phono­logical processes.  First of all, the nasal stop is deleted: which means that the dental segment is disassociated from the C-unit (symbolized by ╪) and reas­sociated with the follo­wing C; see the dotted line. Then the low segment is associated with the free C-time-unit ('len­gthe­ning') while the V-time-unit is as­sociated with the nasal segment ('nasali­zation'), which lead to disassociation from the [-nas] seg­ment.                

The same kind of change has happened in other words with clusters of homorganic consonants, articu­lated in other pla­ces, for example labials and velars. And it happened only when the consonants involved are homor­ganic. There is no nasaliza­tion (and lengthening) at all when the nasal consonant is followed by a heterorganic consonant.

Therefore, the essential condition to the process is the identity of the seg­ments on the 'place'-tier connected with the two C-time-units: [α]. And be­cause the quality of the vowel involved has no in­fluence, the representation of the Brabantic pro­cess of nasaliza­tion may as in (7).

      (7)   [-nas]     [+nas]    [-nas]

               |         |         |

               V         C         C   ---------------->

               |         |         |

               X        [α]       [α]    'place'

                [-nas][+nas] [-nas]

                   ╪   / |      |

                      /             
                     /            
                   V     C      C

  --->                  / 
                       /    
                      /
                   |     ╪      |

              
                   X    [α]    [α]  

In words with a original long vowel the process is somewhat dif­ferent in some aspects. Compare for example the simplex gaan [ga:n] ('to go') with the derived (genetive) form gaans [gã:s] ('for to go'). This time the autosegmental repre­senta­tion needs the use of the so called root-knot because of the long vowel, which occupies two ti­ming-units.

      (8)       V      V        C            C

                 \   /          |            |

                   r            r            r

                   |           /            / 

                [-nas]    [+nas] \     [-nas] \ 
                                [α]           [α]       'place'             
       
        g       a      a        n            s          

                   V       V   C              C

                    \     / \  ╪              |

--->                 \   =   \ |              |  
                      \ /     \               |
                       r       r              r

                       ╪      / \            / \   

                    [-nas] [+nas]\       [-nas] \

                                 [α]            [α]

           g       ã       ã                  s

The difference between (7) and (8) is that in words with a long vowel, after association of the [+nas] fea­ture to the root of the long vowel, a timing unit is lost. So in this case we don't have compen­satory lengt­hening at all. The length of the vowel is stil the same. And that is the case in most words with vowels which have a complex root.

This brings me to the most important question, the reason why nasalization is related to homorganicity, and that for long times. In earlier stages of the Germanic languages we see that the nasal consonants always had the tendency to assimi­late their place of articulation to the following con­sonants (Krahe 1956, 111). As a result all cluste­rs containing nasals are homor­ganic clusters at the end. There is one sys­tematic exception, that is when the second consonant is a morphologi­cal one, for example a suffix. In that case the nasal sometimes will not assimi­late, but even then there is some kind of accommoda­tion, as (9) shows:

(9)  hij neem+t  -> hij neempt ('he takes')

       hij kom+t -> hij kompt  ('he comes')

In forms like neemt and komt there are two heterorganic consonants. Such clusters seem to be not unproblematic in Dutch, but probably because of paradigmatic pressure m re­mains intact and doesn't assimilate, although the articulatory tran­sition is facilita­ted by an intrusive stop p; see Wetzels (1985).

When both con­sonants are homorganic, the nasal consonant goes fur­ther in as­simila­ting to the fol­lowing con­sonant and in the end or al­most, the nasal comple­tely disap­pears, even when the second consonant is mor­phomatic. One could say that the nasal has been ab­sorbed by the fol­lowing con­sonant, except for the nasal autosegment, because that is the only feature in which the nasal and the homorganic conso­nant differ from each other; ab­sorption re­quires complete identity. The remai­ning na­sality needs space and therefore the precee­ding vowel is lengthened to offer that space. The trig­ger for this kind of nasalization is the gradu­al absorp­tion of the nasal conso­nant by the fol­lowing (homor­ganic) consonant. Without this absorp­tion there is no nasali­zation, at least not in Dutch and the Brabantic (Dutch) dia­lects.

 
But this took place not only in Dutch. In earlier stages of some Germanic languages and in Proto-Germanic the same kind of change has taken place. All cases of nasalization there are in combination with homorganic clusters. Sometimes these clusters are the result of assimilation themselves; see (10).

      (10) a) Got. fimf           Eng. five; Dutch vijf

           b) Got. munths    Eng.  mouth; Dutch (dia­lects) muiden

           c) Ogerm. fangXan Got. fahan

The Gothic number fimf has a m, as the result of as­similation of original n; compare Greek pente and Latin cin­que, for example. The stage [mf], must have pre­ceeded also the English and Dutch forms. Whereas fimf is an ex­ample of a bila­bial ([+ant]) cluster, we find a coronal cluster in Gothic munths ('mouth') and a velar (+back) cluster in Ogerm. fangXan ('to catch') (Krahe 1956, 116).

Later, all these forms lost their nasality, which also happens in the case of more recent nasalized forms. Thus, the pronoun ons, which in West-Brabantic dialects is pro­noun­ced [õ:s], in East-Brabantic dialects has no nasality at all: [os], just as in English, although here the form us is much older, of course (Weijnen??).

Not only in Germanic languages are nasals disappearing. As known, the same process occurred in French but there with a remark­able difference, because not only vowels followed by a nasal + consonant (and here too an homorganic one) are nasa­lized, but also vowels followed by a single nasal.

 
Another example, from a rather different language. In his Foundations of theoretical Phonology James Foley claims that Sanskrit hathá has lost its nasal, whereas the related form hánmi has not, because a difference in stress (Foley 1977, 60);  see (11).

      (11)  Sanskrit  hathá ( <- *hanthá)  opposite to: 

           hánmi

But now I would like to point out that the latter word has an heterorganic cluster, the former an homorganic, thus this could be the reason why as well, and of course I think it is. In any case in Brabantic dialects stress plays no role in nasaliz­ation: cfr. lápswã:s (weak person), ónmε:s (brute) to mã:s (brave), mε:s (man). Another observation which supports this view of the absence of influence of stress on nasaliza­tion is Grootaers statement that especially the schwa is eas­ily­ nasalized in the Antwerpian dialect, being the schwa the only Dutch vowel which never has stress (Grootaers 1946, 136).

On the other hand, not all the southern Dutch dialects prac­tice this nasaliza­tion. In several areas nasalization (and lengthening) is unknown. It is very curious that in those dialects a consonant homorganic to the nasal causes shortening of the preceeding vowel, in stead of lengthening. One example: in the dialect of Hofstade the simp­lex of 'man' is [ma:n] with a long vowel; the plural form is [mans], with a short vowel. Also in a complex form like [mansmens], where­as some West-Brabantic dialects have: [mã:nskεrel], with a long [ã:] (Key­meu­len-Taeldeman 1985: 127). This is exactly opposite to (6b).  

     
Another different situation is that some Brabantic diale­cts do not have nasalization in every case, discussed be­fore. But that doesn't mean that they keep their homorganic nasal clusters intact; on the contrary. The most striking case is found in the dialect of Antwerpen, the capital of the province of Antwerpen in Belgium. In the Antwerpian dialect only the [a] is nasali­zed before a coronal cluster: [dã:se] (to dance), [kã:s] (chan­ce). All other vowels are not (Smout 1905, 25). But in stead of that the origi­nal homorganic clus­ters have changed into heterorganic ones by a process of dissimila­tion. That means that coronal nasals have become velar [+back]: n -> ng. (Smout 1905, 40). This velarization is not restricted to Antwerpen. It also appears in the provinces of Holland. And everywhere we find indications for two pheno­mena: first, that nasalization is a fact also in other regions and second, that nasalization itelf is not really accepted in every case. Vela­rization is in fact some kind of inverted rule which may avoid the non preferred nasalization.

To conclude. Essential in my view is that nasalization in the southern Dutch (Brabantic) dialects is triggered by the ab­sorption of the place feature of the nasal by the fol­lowing homorganic consonant. So the ultimate representation of the process is as in (12), [Fn] meaning all other features of the nasal.

      (12)   [-nas] [+nas]     [-nas]

                ╪   /  |           |

                |  /   |           |
         | /    C           C                                    
          /    /            | 
         V    / ╪           |

                |   /  |           |
                |  /   |           |
                X     [α]  ---->  [α]        'place'

                  
                   [Fn] 

As can be seen: absorption means the complete disappear­ing of the place feature of the nasal which is the only con­sonantal feature of the nasal associated (only) to that spe­cial timing-unit; while assimilation means that one or more fea­tures are beco­ming simi­lar to one or more features of the following or preceeding timing-unit, but still associated to there original timing-unit.

Now we can also give an answer on the question if the Braban­tic syllable violates the Dutch rime. As was pointed out before (4a and 4b) the rime of the Dutch syllable has only two [-cons] or [+son] segments when a [+cons] is fol­lowing, where­as the Brabantic syllable with a nasal con­sonant seems to have three: [-cons][-cons][+son][+­cons], being the [+son] the nasal con­sonant: [lã:mp] (l-V-V-N-C). But as we have seen, the nasal consonant is fully absor­bed by the fol­lowing consonant, while the nasal-autoseg­ment is associated to the timing-unit of the vowel. So there are only two [+son] elements left and the real form of 'lamp' is [lã:p] (l-V-V-C). And this leads to the conclusion that the Brabantic nasa­lized syllables are fully in harmony with the Dutch syllable-template.

Bibliography

Bree, C. van 1987

      Historische Grammatica van het Nederlands, Foris, Dor­drecht.

Colinet, Ph. 1896

      'Mechanische ontleding van (a) n5 en a1, in: Leuvensche Bijdragen 1, 245-248.

Foley, J. 1977

      Foundations of theoretical phonology, Cambridge Univer­sity Press, Cambridge.

Grootaers, L. 1946

      'De neusklinker in het Antwerpsch dialect', in: Lode Baekelmans ter eere, 130-137.

Hooper, J. B. 1976

      An Introduction to Natural Generative Phonology, Acade­mic Press, New York etc..

Keymeulen, L and J. Taeldeman 1985

      'Tussen fonologie en morfologie: de vokaalverkorting in een Brabants dialekt', in: Taal en Tongval 37, 124-164.

Krahe, H. 1956

      Germanische Sprachwissenschaft; I. Einleitung und Laut­lehre, Walter de Gruyter, Berlin.

Nooteboom, S.G. and A. Cohen 1984

      Spreken en verstaan; een nieuwe inleiding tot de ex­eri­mentele fonetiek, Van Gorcum, Assen.

Smout, H. 1905

      Het Antwerpsch dialect met eene schets van de ge­schiedenis van dit dialect in de 17e en de 18e eeuw,  Librairie scientifique E. van Goethem, Gent.

Trommelen, M. 1984

      The Syllable in Dutch, Foris, Dordrecht.

Wetzels, L. and Engin Sezer (eds.) 1986

      Studies in Compensatory Lengthening, Foris, Dordrecht.

Wetzels, W. Leo 1985

      'The Historical Phonology of Intrusive Stops: A Non-Linear Description', in: Canadian Journal of Linguis­tics, 30, pg. 285-333.

Zwaardemaker Cz., H. and L.P.H. Eijkman 1928

      Leerboek der phonetiek, De Erven F. Bohn, Haarlem.