Logo DBNL
'The Vowels of Dutch: Phonetic and Distributional Classes' (1962)

'The Vowels of Dutch: Phonetic and Distributional Classes'

(1962)–W.G. Moulton–rechtenstatus Auteursrechtelijk beschermd

Vorige

The vowels of Dutch: phonetic and distributional classes

Thirty years have passed since the publication of A.W. de Groot's pioneering description of the phonemes of standard Dutch1). In a very real sense we can say that this brilliant presentation has never been surpassed. lts basic findings have remained unshaken, even though subsequent analyses have added useful details or given alternative interpretations2). The following paragraphs are yet another attempt to offer a few footnotes to what De Groot presented to us three decades ago3).

Characteristic not only of De Groot's analysis, but also of the phonetic analyses which preceded it and of the phonemic analyses which have followed it, is a division of the vowels of standard Dutch into five different classes. I give them here in spellings as close to the standard orthography as possible (the use of phonemic symbols at this point would prejudice the case in favor of some particular phonemic analysis), and add a typical example for each:

 

A. A class of vowels traditionally called ‘short’:

i u o bit put bom (bot)4)
e   ɔ bed   bod (bot)4)
  a     bad  

B. A class of vowels traditionally called ‘long’:

ie uu oe biet buut boet
ee eu oo beet beuk boot
  aa     baat  

C. A class of vowels traditionally called ‘diphthongal’:

ei ui ou bijt buit bout

D. The vowel ə, classed by some as a separate phoneme, by others as an allophone of u:

ə de

E. A class of vowels traditionally called ‘foreign’:

εε εu ɔɔ serre freule rose

Since all analysts assume these five classes under one label or another (differing only in their phonemic interpretation of ə), we need to consider the reasons which underlie such a classification. I shall discuss the five classes in reverse order, beginning with E and working back to A.

The ‘foreign’ vowels εε, εu, ɔɔ. These vowels present a familiar problem in phonemic analysis: if certain loanwords show a different phonemic structure from native words, how fully integrated must they be before we need to include them in our analysis? A pertinent answer in the present case is, I believe, the following. We may exclude aberrant phones which occur in relatively small numbers of forms, especially if they are a reasonable imitation of some foreign pronunciation; but we must include aberrant phones which occur in large numbers of forms, especially if they have become conventionalized and do not necessarily constitute genuine imitations of foreign pronunciations. In the case of Dutch, this means that we can agree with Van Haeringen (p. 160) in excluding the long [i:] and [y:] used in intrige, franchise, expertise, prestige, centrifuge, and with Cohen et al. (pp. 23-24) in excluding the long [i:] and [u:] of (English) team and boom and the final nasalized vowel often used in restaurant (and, one might add, the two nasalized vowels of enfin). When Dutch speakers use forms such as these, we may safely say that to this extent they are

not speaking Dutch but rather French or English. Quite different from the occasional use of these aberrant phones is the far more extensive and conventionalized use of εε, εu, ɔɔ. In the first place, two of these vowels (εε and ɔɔ) occur not just in a handful of forms, but in large numbers of them. Secondly, all three of them serve as highly conventionalized markers of ‘foreignness’, and by no means always as genuine imitations of foreign pronunciations. Contrast the ‘pseudo-French’ [ɔ:] of Dutch rose, Rhône, zone, controle, etc. with the [o:] which is actually used in standard French; or the ‘pseudo-German’ [œ:] of Dutch freule with the [ɔi] which is actually used in German Fräulein. Finally, εε and ɔɔ serve also to mark words as unusual in other respects, for example as foreign-learned: [bε:ta] beta; as onomatopoetic: [blε:rə(n)] bleren; as solemn: [hε:r] (beside [he:r]) heir, heer; as interjections: [γɔ:]; etc. For all of these reasons we must surely include εε, εu, ɔɔ in any analysis of the Dutch vowel system, though of course with the realization that their status is somewhat different from that of other vowels.

The vowel ə. Some analysts (e.g., De Groot, Van den Berg) have considered this to be the unstressed allophone of the phoneme which occurs with stress in put; others (e.g., Van Ginneken, Cohen et al.) have considered it to be a separate phoneme in a class by itself. It is significant that the former group of analysts have feit it necessary to devote considerable discussion to the phonemic status of ə, indicating that its membership in the u-phoneme is far from obvious. For reasons which have perhaps been best expressed by Heeroma (p. 300), I agree with the latter group of analysts that ə should be considered a separate phoneme in a structural class by itself. First, ə is phonetically unique: it has no one characteristic vowel color; about all one can say articulatorily is that it is usually more or less mid central, and that it may or may not show lip rounding. Secondly, ə is also distributionally unique: it is typically unstressed, and occurs stressed only under very special circumstances; and, though phonetically short, it differs distributionally from the short vowels of class A in that it occurs freely in word-final position. Finally, the relationship of ə to other vowels is also important in determining its phonemic status. Many other vowels occur unstressed in careful pronunciation, but when the pronunciation becomes less careful they merge with ə; conversely, as less careful pronunciation becomes more careful, ə

merges with any one of a number of different vowels (though of course always with a given vowel in a given form). There therefore seems to be no good reason for identifying ə with one of these vowels (e.g. u) rather than with another (e.g., uu; cf. [mə'zik] beside more careful [my'zik] muziek, etc.).

The ‘diphthongs’ ei, ui, ou. The use of the term ‘diphthongal’ to refer to this class of vowels is phonetically quite justified; on the other hand, since six of the seven ‘long vowels’ may also be diphthongal, the matter needs some discussion5). If we arrange the ‘long vowels’ and ‘diphthongs’ according to the relative degree of diphthongization which they typically show in word-final position, we arrive at the following scale:

monophthongal:   aa  
weakly diphthongal: ie uu oe
moderately diphthongal ee eu oo
strongly diphthongal: ei ui ou

The fact that nine of these ten vowels are diphthongal suggests that all ten of them should be analyzed not as single vowels (or diphthongs) but rather as vowel plus vowel, whereby the second vowel is allophonically non-syllabic. Then, rearranging all ten in the more familiar shape of the vowel triangle, we would get the following system, in phonemic transcription:

ii yy uu biet buut boet
ei øy ou beet beuk boot
εi œy ɔu bijt buit bout
  aa     baat  

Since, as we shall see later, there are strong distributional arguments in favor of analyzing the ‘long vowels’ and ‘diphthongs’ as consisting structurally of two phonemes each, why have Dutch scholars never proposed such a solution? I believe that two answers may be

given to this question. The first answer is phonological. Though a diphthongal glide may be present in the articulation of ie, uu, oe and ee, eu, oo, it may just as well be absent; therefore the presence of the glide is not phonemically relevant; and therefore these six vowels are structurally monophthongs and not diphthongs. In the case of ei, ui, ou, on the other hand, the diphthongal glide is always present (though it may be very slight in ui); therefore the presence of the glide is phonemically relevant; and therefore these three vowels are structurally diphthongs and not monophthongs. The second answer is sociological. If ee, eu, oo are too strongly diphthongized, the resulting pronunciation is considered substandard; and if ei and ui are too weakly diphthongized, this resulting pronunciation is also considered substandard; therefore, in standard Dutch, ee, eu, oo are sociologically monophthongs, and ei, ui, ou are sociologically diphthongs.

The ‘short’ and ‘long’ vowels. In the preceding paragraph we found it possible to justify the traditional label ‘diphthongal’ for the vowels ei, ui, ou. When we now consider the ‘short’ and ‘long’ vowels, we find that these traditional labels can be justified only partially. For one thing, the ‘short’ vowels are not always short, since they can undergo expressive lengthening. The ‘short’ a of dag, for example, is very frequently long when this word is used in the meanings ‘hello’ and ‘goodbye’. This is perhaps a trivial objection; we can keep the traditional label for these vowels if we take it to mean ‘typically short’. A far more serious objection is the fact that the class of ‘long’ vowels contains three which are usually short, namely ie, uu, oe. They are phonetically long only before /r/: bier, buur, boer6); they are short in all other environments, e.g. in biet, buut, boet, in wie, nu, hoe, etc. If we wish to give phonetically honest listings of the short and long vowels, then we must include ie, uu, oe in both lists:

  Short     Long  
ie uu oe ie uu oe
i u o ee eu oo
e   ɔ      
  a     aa  

This suggests that the traditional division of these thirteen vowels into two different structural classes ought to be given up altogether. This conclusion is supported by the behavior of the class A vowels under expressive lengthening: though they then become phonetically long, they do not merge with the corresponding class B vowels but remain qualitatively distinct from them. It would therefore seem that we must place all thirteen of these vowels in a single structural class, and that they are distinguished from one another only through their different positions along the articulatory dimensions of tongue height (high-to-low), tongue position (front-to-back), and lip action (spread-to-rounded)7):

ie   uu   oe
ee   eu   oo
i   u   o
e       ɔ
  aa   a  

It is a striking fact that, even though the structural classes ‘short’ and ‘long’ cannot be justified in terms of these traditional phonetic labels, scholars have nevertheless refused to abandon the classes as such. It is as if they felt the classification to be intuitively correct; and, not being able to justify it on the basis of vowel length, they have tried to find some other theoretical basis for maintaining it. Three different theories have been proposed:

(1) ‘Dof’ vs. ‘helder’. In his 1931 article, De Groot proposed an impressionistic acoustic classification: the vowels of class A are ‘dof’, and those of class B are ‘helder’. Unfortunately this proposal does not advance matters. It merely tells us that the two classes exist (a fact which we all intuitively accept), and it then attaches labels to them. What we need are not labels, but objective reasons to explain why we believe that the two classes exist in the first place.

(2) ‘Scherp gesneden’ vs. ‘zwak gesneden’. In various publications8) N. van Wijk proposed a solution based on a particular theory

of syllable structure. According to this theory the vowels of class A are ‘scherp gesneden’ and the syllables in which they occur are therefore closed even when another vowel follows (latten = lat-ten, with ambi-syllabic t); whereas the vowels of class B are ‘zwak gesneden’ and the syllables in which they occur are therefore open when another vowel follows (laten = la-ten). This is acceptable as a scientific hypothesis, since it can be put to the test of verification by other scholars. Unfortunately, it has not passed this test. As Van Haeringen remarks (p. 160), there is no way of determining that the syllable boundary which follows the ‘zwak gesneden’ vowels of gieten, futen, boeken is in any way different from that which follows the ‘scherp gesneden’ vowels of pitten, putten, pokken; or that the ‘zwak gesneden’ vowels of riet, fuut, boek are less sharply ‘cut off’ than are the ‘scherp gesneden’ vowels of rit, fut, bok.

(3) ‘Lax’ vs. ‘tense’. A third theory was proposed in 1959 by the authors of the Fonologie van het Nederlands en het Fries (chapter 2, § 2). Sounds articulated with the speech organs relatively close to the position of rest are here called ‘lax’ (‘ongespannen’); those articulated with the speech organs relatively far from the position of rest are called ‘tense’ (‘gespannen’)9). Using these definitions, they then pair off the vowels of classes A and B as follows:

Class A, lax: i e a ɔ o u
Class B, tense: ie ee aa oo oe uu, eu

This is by far the most elegant of the three proposed solutions, but it is applicable only after we have already separated these thirteen vowels into two different classes. That is to say, it does not tell us

why we intuitively set up these two classes in the first place; it merely tells us that, after having done so, we can then pair off the members of the two classes with one another as ‘lax’ vs. ‘tense’. If we are simply presented with these thirteen vowel phonemes, and not told which six belong to one class and which seven to the other, we can pair them off as ‘lax’ vs. ‘tense’ in at least one other way:

Lax: ee e eu, u oo ɔ a
Tense: ie i uu oe o aa

Here, as in the preceding diagram, within each pair the vowel labeled ‘lax’ is relatively closer to the position of rest, and the vowel labeled ‘tense’ is relatively farther from the position of rest. This means that we still have not found why we intuitively place the vowels i, e, a, ɔ, o, u in one class, and the vowels ie, ee, aa, oo, oe, uu, eu in another.

The above proposals for a theoretical justification of our intuitive classification are based on phonetics. We can come a good deal closer to understanding the reasons for our intuitive classification if we turn to matters of phonemic distribution, that is to say, if we look upon our classification as primarily distributional rather than phonetic. The following distributional facts appear to be relevant:

(1) Occurrence in word-final position. If we group all twenty of the vowel phonemes of Dutch according to whether or not they occur freely in word-final position, we arrive at the following classification:

Do not occur Occur Examples

i u o ie uu oe zie nu moe
e   ɔ ee eu oo zee reu zo
  a     aa     la  
(εε εu ɔɔ) ei ui ou zei lui zou
        ə     de  

From the eleven vowel phonemes which occur freely in word-final position we may subtract the two structural classes which we have already defined: class D, ə; and class C, the diphthongs ei, ui, ou. The remaining seven vowels are then precisely those of class B, traditionally called ‘long’. From the nine vowels which do not occur freely in word-final position, we may subtract the class of ‘foreign’ vowels: the non-occurrence of these phonemes in word-final position is less a feature of Dutch than of the foreign languages from which most words of this

type are derived. The remaining six vowels are then precisely those of class A, traditionally called ‘short’. (These vowels occur in word-final position only in the interjections hè, bà, jò, plus perhaps a few others.) Since it is a feature of distribution which clearly differentiates the vowels of classes A and B from one another, we may henceforth give them distributional labels: the class A vowels are ‘checked’, the class B vowels are ‘free’10).

(2) Occurrence before /j/ and /w/. If we examine the occurrence of vowels before /j/ and /w/, we find another distributional feature which distinguishes sharply between the checked and free vowels: within words, all of the free vowels occur before one or both of these phonemes, whereas none of the checked vowels do. The following examples illustrate the free vowels (and the diphthongs) in these positions. Since the standard orthography is irregular in many forms of this type, each example is followed by a simplified spelling which is phonemic if we accept the vowel notations ie, ee, aa, etc. as unitary symbols11).

  Before /-jə/ Before /-j/ Before /-wə/ Before /-w/
ie drieën (driejə) - nieuwe (niewə) nieuw (niew)
ee tweeën (tweejə) - leeuwen (leewə) leeuw (leew)
aa taaie (taajə) taai (taaj) - -
oo mooie (moojə) mooi (mooj) - -
oe groeien (groejə) groei (groej) - -
uu - - duwen (duuwə) duw (duuw)
eu reuen (reujə) - - -
ei bijen (beijə) - - -
ui buien (buijə) - - -
ou - - bouwen (bouwə) -

The two distributional features described above distinguish sharply between the checked vowels on the one hand and the free vowels (and

diphthongs) on the other, with no exceptions. The remaining distributional features have a small residue of exceptions, largely cases in which ie, uu, oe pattern primarily like free vowels, but in part also like checked vowels.

(3) Occurrence before final consonants. If we examine the occurrence of stressed vowels before final consonants, we find that the checked vowels may be followed by structures of the types /-C/, /-CC/, /-CCt/, /-CCs/, /-CCst/, and (often unstable) /-CCts/; but that the free vowels (and the diphthongs) may be followed only by structures of the types /-C/, /-Ct/, /-Cs/, /-Cst/, and (often unstable) /-Cts/. (The few exceptions will be noted below.) In these formulae the symbol /C/ stands for any one of the phonemes /p t k f s x m n l r/ and, after free vowels (and diphthongs) only, /j w/, with the proviso throughout that no consonant may follow itself. Because the phoneme /η/ functions before /k/ as /C/, but otherwise as /CC/, such forms as dank, zinkt, links are considered to have the clusters -nk, -nkt, -nks, and such forms as ring, mengt, langs to have the clusters -ng, -ngt, -ngs, etc. If we use the term ‘finals’ for the /C/'s in these formulae, and the term ‘post-finals’ for the /-t -s -st -ts/ which may follow them, we can then make the statement: checked vowels may be followed by two finals, but free vowels by only one.

The following examples illustrate the occurrence of clusters after checked vowels. A dash indicates non-permitted clusters of a consonant following itself. Examples of /-CCts/ are omitted; these are all of the type gescherpts, ongemerkts, geverfts, geschorsts, verzorgds, opgewarmds, gekarnds, in which the /t/ is often omitted. Examples of /-CCs/ and /-CCst/ are also omitted for the last twelve combinations of /CC/; these are all of the type stipts, stiptst, echts, echtst, in which the first /t/ is often omitted.

 

Clusters following checked vowels

  /-CC/ /-CCt/ /-CCs/ /-CCst/
/rp/ terp werpt dorps scherpst
/rt/ kort - arts kortst
/rk/ werk markt kerks sterkst
/rf/ korf verft zwerfs herfst
/rs/ kers worst - -
/rx/ berg burcht ergs ergst
/rm/ berm kermt warms armst

/rn/ kern karnt ? ernst
/lp/ hulp helpt wulps wulpst
/lt/ geld - milds wildst
/lk/ wolk melk zulks schalkst
/lf/ golf helft zelfs ?
/ls/ hals hulst - -
/lx/ balg volgt volgs ?
/lm/ helm walmt kalms kalmst
/mp/ ramp prompt lomps stompst
/mt/ hemd - verlamds verlamdst
/nt/ hand - gekends gekendst
(nk) dank zinkt links flinkst
(ng) ring mengt langs hengst
/mf/ nimf ? ? ?
/ms/ gems komst - -
/ns/ mens kunst - -

  /-CC/ /-CCt/   /-CC/ /-CCt/
/pt/ stipt - /fs/ lafs lafst
/ps/ rups gipst /sp/ wesp raspt
/ts/ rots botst /st/ kast -
/kt/ tact - /sk/ grotesk ?
/ks/ heks tekst /xt/ nacht -
/ft/ gift - /xs/ rech(t)s zach(t)st

The following examples illustrate the occurrence of clusters after free vowels. Examples of /-Cts/ are omitted; with a few exceptions (naakts, vreemds, koorts) they are all of the type gestreepts, beleefds, gevreesds, verhoogds, geleends, vertaalds, begroeids, besneeuwds.

 

Clusters following free vowels

  /-C/ /-Ct/ /-Cs/ /-Cst/
/p/ hoop doopt loops diepst
/t/ boot - fiets poetst
/k/ boek naakt reeks ziekst
/f/ dief hoofd hoofs liefst
/s/ poes haast - -
/x/ leeg maagd hoogs oogst
/m/ riem beemd teems heemst
/n/ been maand zoons dienst
/l/ doel naald deels geelst
/r/ dier woord kaars heerst
/j/ mooi nooit taais mooist
/w/ leeuw duwt nieuws ruwst

Exceptions, with two finals

  /-CC/ /-CCt/ /-CCs/ /-CCst/
/rn/ deern toornt Baarns vertoorn(d)st
/rp/ wierp      
/rf/ wierf, stierf, zwierf, bedierf      
/lp/ hielp      
/lf/ twaalf      

The forms with the two finals /rn/ constitute a clear exception to the general rule that free vowels may be followed by only one final; the exception applies, of course, only to those speakers who distinguish toorn from toren, Baarn from baren, etc. An isolated exception is twaalf, provided this is not phonemically /twaaləf/; however, anaptyctic /ə/ may of course occur in all forms with /r/ or /l/ plus labial or velar. The remaining exceptions all involve the occurrence of ie before /CC/ (phonetically short in this environment) in the past of class III strong verbs; they show that, ever since the period when these analogical forms were created, ie has patterned in part like a checked vowel.

(4) Occurrence before medial consonants. If we examine the occurrence of stressed vowels before medial consonants, we again find that the checked and free vowels pattern quite differently. Though the free vowels occur frequently before /v/ and /z/ (leven, geven, lezen, wezen, etc.), the checked vowels are very rare in this position (grovve, puzzle). Conversely, though the checked vowels occur frequently before /f/ and /s/ (heffen, blaffen, bessen, kussen, etc.), the free vowels are fairly rare in this position (tafel, lafenis, Friese, bloesem). This matter has been well discussed by Heeroma (pp. 300-301), and we need not elaborate on it here. When we turn from these single consonants to consonant clusters, we find a distribution quite similar to that in word-final position. Checked vowels may be followed by structures of the types /-CCə/, /-CCtə/, /-CCdə/, /-CCsə/, /-CCstə/, but free vowels only by structures of the types /-Cə/, /-Ctə/, /-Cdə/, /-Csə/, /-Cstə/ and, rarely, /-Czə/, /-Czdə/. If we use the term ‘medials’ for the /C/'s in these formulae, and the term ‘post-medials’ for the /t d s z st zd/ which may follow them, we can then make the statement: checked vowels may be followed by two medials, but free vowels by only one. Since the distribution (including the exceptions) is so similar to that in word-final position, we may content ourselves with just a few examples:

Clusters following checked vowels

  /-CCə/ /-CCtə/ /-CCsə/ /-CCstə/
    /-CCdə/    
/lm/ helmen kalmte schelmse kalmste
    walmde    
/rf/ amorfe sterfte zwerfse herfsten
/rv/ korven verfde    

Clusters following free vowels

  /-Cə/ /-Ctə/ /-Csə/ /-Cstə/
    /-Cdə/ /-Czə/ /-Czdə/
/r/ deuren kaarten heersen heerste
    woorden laarzen laarsde
/f/ strofe kreeften hoofse liefste
/v/ leven hoofden    

Exceptions, with two medials

  /-CCə/ /-CCde/ /-CCsə/
/rn/ deernen vertoornde Baarnse
/rp/ wierpen    
/rv/ wierven, stierven, zwierven, bedierven    
/lv/ twaalven twaalfde  

The exceptions, with two medials following a free vowel, are exactly the same as in word-final position. The form koortsen also needs special mention; it can be interpreted either as /-Ctsə/, with an unusual post-medial /ts/, or as /-CCsə/, with two medials.

(5) Occurrence before the diminutive suffix. A further distributional feature concerns the shape of the diminutive suffix after stems ending in a stressed vowel folio wed by /l r m n/. If a checked vowel precedes the liquid or nasal, the diminutive suffix has the shape -etje: zinnetje, dennetje, karretje, bolletje, sommetje, dunnetjes. But if a free vowel (or a diphthong) precedes the liquid or nasal, the diminutive suffix has the shape -tje or, after /m/, -pje: kieltje, zeempje, baantje, boontje, zoentje, uurtje, deurtje, rijmpje, tuintje, Paultje. However, Van Haeringen (pp. 162-163) has called our attention to a large and apparently growing number of exceptions to this rule. All of them involve forms in which ie, uu, oe occur before /l m n/, that is to say, precisely in the environments where they are phonetically short, but never before /r/,

where they are phonetically long. In some cases the ‘regular’ and ‘irregular’ forms exist side by side: bloempje and bloemetje, wieltje and wieletje. In other cases, especially after /m/, ‘regular’ forms which have long existed continue to be used, but new formations are always ‘irregular’: roemetje, pseudoniemetje and, as names for the children of Bloem and Priem, de Bloemetjes, de Priemetjes. This is therefore another distributional category where ie, uu, oe pattern usually like free vowels but, when phonetically short, like checked vowels. The category is particularly valuable because it shows that the patterning as checked vowels is clearly on the increase.

The preceding paragraphs have shown us that, if we use distributional rather than phonetic criteria, it is quite possible to find objective data which support the traditional intuitive assumption that the vowels ie, ee, aa, oo, oe, uu, eu belong together as a separate structural class. We have found (1) that all and only these vowels occur freely in word-final position, and (2) that all and only these vowels occur before /j/ and/or /w/12). At the same time, however, we have also found that, in the remaining three positions which we examined, the vowels ie, uu, oe pattern at least partly like the ‘short’ vowels of class A, namely (3) before final consonant clusters, (4) before medial consonant clusters, and especially (5) before the diminutive suffix. Furthermore, we have found that ie, uu, oe pattern as ‘short’ vowels precisely in those environments where they are themselves phonetically short, i.e. where not followed by /r/, or where followed by /r/ plus a non-dental consonant (wierp, wierpen, stierf, stierven, etc.). We therefore seem to face a dilemma. In order to justify grouping ie, uu, oe with the ‘long’ vowels of class B, we have had to abandon phonetic criteria and turn to distributional criteria; but an examination of further distributional criteria has shown us that it is phonetic criteria after all which determine, in part, the class of vowels to which ie, uu, oe belong, since when they are phonetically short they often pattern like the ‘short’ vowels of class A.

There is, I believe, a very simple way out of this apparent dilemma. We can return to the familiar, traditional classification of the vowels of Dutch on the basis of the phonemic opposition ‘short ≠ long’. The vowels i, e, a, ɔ, o, u are phonemically short; and the vowels ee, aa, oo, eu are phonemically long; but the vowels ie, uu, oe are phonemically neither short nor long - they simply do not share in this opposition. Phonetically they are short in some environments, long in others; distributionally they pattern usually like the phonemically long vowels13), but occasionally like the phonemically short vowels. They themselves, however, are phonemically neither short nor long; in this part of the vowel system the opposition ‘short ≠ long’ is suspended.

Since this analysis of the vowels ie, uu, oe does not seem to have been proposed before, it may be helpful to cite analogies from other parts of the Dutch phonemic system. Among the stop consonants, Dutch has a three-way opposition ‘labial ≠ dental ≠ velar’. Within the labials and dentals there is a further opposition ‘fortis ≠ lenis’, giving /p/-/b/, /t/-/d/; but there is no such opposition within the velars, since the phones [k] (fortis) and [g] (lenis) are in complementary distribution. The resulting phonemic structure can by symbolized as follows:



illustratie

Similarly, among the fricatives Dutch has a three-way opposition ‘labial ≠ dental ≠ velar’ (plus glottal /h/). Within the labials and dentals there is (probably for most speakers) a further opposition ‘fortis ≠ lenis’, giving /f/-/v/, /s/-/z/; some speakers also have this opposition within the velars, giving /x/-/γ/, whereas others have

[x] and [γ] in complementary distribution. The resulting phonemic structures are therefore as follows:



illustratie

If we apply this same type of analysis to the phonemically short vowels, to the phonemically long vowels, and to ie, uu, oe (which are phonemically neither short nor long), we can indicate the relationships among them as follows. Two diagrams are given, one with the notations used thus far, and another with a possible phonemic notation14).



illustratie

Where two symbols appear in the same box, separated by a comma, the first represents a vowel which is phonemically short, the second a vowel which is phonemically long.

In the above diagrams we have, somewhat arbitrarily, excluded the ‘diphthongs’ and the ‘foreign’ vowels. If we include these two sets, we can give a diagram of all four classes of stressed vowels (i.e., all vowels except /ə/), together with an indication of the phonetic features which distinguish each vowel from all others. A notation such as ‘short ~ long’ indicates that these two features are phonemically relevant; a notation such as ‘short ˜ long’ indicates that these two features are phonetically present but not phonemically relevant.



illustratie

Such an arrangement indicates the fact that ie, uu, oe show short, long, and diphthongal allophones, in non-contrastive distribution; and that ee, eu, oo show monophthongal and diphthongal allophones, also in non-contrastive distribution. It also suggests the increasing diphthongization as one goes from ie, uu, oe (weakly, allophonically diphthongal) to ee, eu, oo (moderately, allophonically diphthongal) to ei, ui, ou (strongly, phonemically diphthongal).

There are two novel aspects to the above analysis of the stressed vowels of Dutch: first, that the vowels ie, uu, oe do not participate in the phonemic opposition ‘short ≠ long’; and, secondly, that i, u, o are structurally the short correlates of ee, eu, oo. As a supplement to this synchronic analysis we may therefore discuss, very briefly, two diachronic questions. First, at an earlier period the vowels reflected in modern ie, uu, oe seem to have been long in all environments; when and how did they develop short allophones in most positions? Secondly, at an earlier period the short vowels reflected in modern i, u, o seem to have been phonetically high; when and how were they lowered so as to become the short correlates of modern long ee, eu, oo? In attempting to answer these two questions, we shall use as examples the earlier, in part reconstructed forms of the words reflected in modern pit, put, bot, hek, pot, pad, tijd, huid, zuur, liet, voet, heten, lopen, beet, breuk, zoon, eten, open, daad.

The short i, u which we reconstruct for the West Germanic period seem to have been high vowels: *biti, *bruki, *sunu. Later, after the merger of unstressed vowels as ə and the resulting phonemicization of the umlaut of u, these forms became *bitə, *brykə, *sunə, still with stressed short high vowels. Well before the Middle Dutch period, however, the vowels of forms like these seem to have been lowered,

giving something like *betə, *brøkə, *sonə. The reason for believing that such a lowering took place is the fact that, upon the lengthening of short vowels in open syllables, these forms became Middle Dutch bete, brueke, sone. Here the lengthened ē and ō of bete and sone were (1) phonemically identical with the lengthened ē and ō of Middle Dutch eten and open (< WGmc. *etan and *opan), and (2) were apparently phonetically lower than the ē and ō of Middle Dutch heeten and loopen (< WGmc. *haitan and *hlaupan). If such an extreme lowering took place in open syllables, there is every reason to believe that at least a partial lowering also took place in closed syllables. We may therefore surmise that Middle Dutch pit(te), put(te), but(te) and bot(te) were phonetically something like [pet(tə) pøt(tə) bot(tə)], in contrast with hec(ke), pot which were phonetically perhaps [hεk(kə) pɔt].

After this lowering had taken place, the short and long vowels of Middle Dutch must have formed a system something like the following:

  Short     Long  
- - - tijt huut  
pit put but, bot liet   voet
hec   pot heeten   loopen
  pat   bete brueke sone
        daet  

After the diphthongization (except before /r/) of Middle Dutch î and û (tijt, huut) to what eventually became modern ei (ij) and ui (tijd, huid), the remaining long vowels except ae seem to have been raised slightly. The result was the following forerunner of the modern system:

  Short     Long  
- - - liet suur voet
pit put but, bot heeten   loopen
hec   pot bete brueke sone
  pat     daet  

(A still later raising of the row bete, brueke, sone led to the merger of the vowels of bete and heeten, sone and loopen, modern beet and heten, zoon and lopen.)

It is at this post-Middle Dutch stage that we find the answers to the two questions asked above. First, since in this system the long

vowels of liet, suur, voet had no short correlates, the way was open for them to develop short allophones in most environments. We cannot, of course, know just when such a development began; but from this stage onwards it was structurally possible for them to become short without merging with any already existing short vowels. Secondly, though the phonetic lowering of old i, y, u in closed syllables may well have been earlier (as it was surely earlier in open syllables: *bitə, *brykə, *sunə > *betə, *brøkə, *sonə, then lenghtened to *bētə, *brøkə,*sōnə), it is not until this stage that they appear clearly to have become the short correlates of the long vowels of heeten, loopen and, somewhat later, also of the long vowels of beet, breuk, zoon15).

 

William G. Moulton

Princeton University

Princeton, New Jersey


Vorige