Spektator. Jaargang 5

[p. 17]

In Support of D-Deletion
Norval S.H. Smith

Samenvatting

Hieronder verdedigt de auteur van een eerder in Spektator verschenen artikel (Smith 1973) zich tegen de rekatie daarop die onlangs in Spektator werd gepubliceerd (Zonneveld 1975). (Red.)

In a recent article on D-Deletion in Dutch (Smith, 1973) I have suggested that among the rules necessary for accounting for all the phenomena associated with this process were two - D-DELETION and J-INSERTION - which in two steps accounted for the replacement of /d/ by /j/ in many forms in Dutch, cf.

(1)	goede	goeie	/xūdə/	→ /xūjə/	good
	rode	rooie	/rōdə/	→ /rōjə/	red
	wieden	wieën	/wdən/	→ /w jən/	to weed
	brede	breeë	/brēdə/	→ /brejə/	broad

I had suggested that to accomplish this replacement by one rule /d/ → /j/ would be to miss certain generalisations. Firstly it is the case that in precisely the same set of environments where we find /j/ as a ‘descendant’ of /d/ we also find it as a hiatus-filler. In other words I claimed that there was an independently necessary rule of J-INSERTION and that this could also be used in the /d/ to /j/ process. Further, I suggested that J-INSERTION was a ‘different rule from the low-level rule inserting glides between high vowels and following unstressed vowels as in Ria (girl's name), [rjā], Mantua (Mantua) [mantȳā].’ (ibid., p. 435, fn. 9). A second reason for splitting the /d/ to /j/ change into two rules was the existence of a second set of forms where /d/ disappears. These display exactly the same segmental environment as the abone-mentioned cases:

The difference is the forms of (1) above have a morpheme-boundary between the /d/ and the /ə/, which is not present in words of the following type:

(2)	roede	roe	/rūdə/	→ /rū/	rod
	Brederode	Brederode	/-rōdə/	→ /-rō/	personal name
	slede	slee	/slēdə/	→ /slē/	sledge
	lade	la	/lādə/	→ /lā/	drawer

I assumed that a rule of D-DELETION, then, operated in the above-mentioned segmental environment. J-INSERTION, however, only operates when a morpheme-boundary is present - informally in the following context:

[p. 18]

Examples J-INSERTION alone operates are:

(3)	singular	plural
	koe	koeien	/kū/	/kūjən/	cow(s)
	vlo	vlooien	/vlō/	/vlōjən/	flea(s)
	drie	drieën	/dr/	/dr jən/	three(s)
	wee	weeën	/wē/	/wējən/	birth pain(s)

The cases of (2), where J-INSERTION does not operate, will then be subject to a rule of SCHWA-DELETION, which will operate to produce the forms on the righthand side in (2).

This gives us the following derivations:

	kū + ən	(cows)	xūd + ə	(good)	rūdə	(rod)
D-DELETION	......		xū + ə		ru ə
J-INSERTION	kūj+ən		xūj + ə		....
SCHWA-DELETION	......		......		rū
	kūjən		xūjə		rū

In a reaction to my article (Zonneveld, 1975), Wim Zonneveld chooses for a one-step solution involving a rule replacing /d/ by /j/ in the forms in (1). This rule he calls D-WEAKENING. He gives two basic reasons for disagreeing with my solution of a rule of D-DELETION.

Firstly he objects to my association of the forms under (1) and (2). He points out correctly that there is a significant difference in productivity between the two types - the forms in (1) being fairly productive in their formation, while the forms in (2) represent a quite unproductive process and may often involve idiosyncratic meaning differences. Zonneveld says, ‘...this distinction in productivity, which in itself seems sufficient reason to speak of two different processes...’. It seems to me that there are two sets of facts which we must account for, a) the fact that the processes in (1) and (2) involve precisely the same segmental environment, b) the fact that there is a strong difference in productivity between the two processes. In my version I did this by positing a unitary rule of D-DELETION, and looking after the productivity problem by means of a number of redundancy rules. Zonneveld accounts for the difference in productivity by introducing two separate rules, D-WEAKENING for the forms in (1), and DE-DELETION for the forms in (2), and thereby loses the chance of stating the generalisation that the segmental environment is the same in the two cases.

His second reason for opposing my two-step analysis is that he claims that there is in fact no need for a separate rule of J-INSERTION, and therefore no reason for spliting up the process /d/ → /j/ into two processes /d/ → and → /j/. Clearly my analysis becomes much weaker if there is indeed no independent rule of J-INSERTION. It should be clearly stated that J-INSERTION would have different results from the low-level glide insertion rule discussed above, christened by Zonneveld HOMORGANIC GLIDE INSERTION, so that if the former does not exist then the coincidence between what happens in (1) and (3) above does not exist either.

For this reason I will devote the remainder of this paper to attempting to demonstrate that J-INSERTION is indeed necessary as a phonological rule of Dutch. Zonneveld claims that most of the forms of the type illustrated in (3) can be explained with reference to the rule of HOMORGANIC GLIDE INSERTION

[p. 19]

(henceforth HG-INSERTION). As Zonneveld correctly points out this rule operates not only after high vowels, but after non-low vowels in general. Also the following vowel does not require to be unstressed. Zonneveld gives inter alia the following examples:

Lea	/lējā/	girl's name
Samoa	/sāmōwā/	Samoa
idioom	/d m/	idiom
poëet	/pōwēt/	poet

This rule is required therefore to insert a glide which is homorganic with the preceding vowel in two respects - rounding and backness, cf.

a)	preceding vowel is front unrounded
		//	Ria	/r/	→	/rjā/	girl's name
		/ē/	Lea	/l/	→	/lējā/	girl's name
b)	preceding vowel is back rounded
		/ū/	Papoea	/pāp/	→	/pāpūwā/	Papua
		/ō/	Samoa	/sām/	→	/sāmōwā/	Samoa
c)	preceding vowel is from rounded
		/ȳ/	Mantua	/mantȳa/	→	/mantȳā/	Mantua

In other words we have /j/ inserted after front unrounded vowels, /w/ after back rounded vowels, and // (a front rounded glide as in French huit [it], (eight)) after front rounded vowels. We can formulate the following rule (a modification of Zonneveld's formulation):

Now let us examine the cases where glide-insertion takes place before various schwa-initial suffixes.

With front unrounded vowels there is no difference from the HG-INSERTION cases, eg.:

singular	plural
drie	drieën	/dr/	/drjən/	three(s)
wee	weeën	/wē/	/wējən/	birth pain(s)

Clearly both J-INSERTION and HG-INSERTION would give the same result here.

[p. 20]

With back rounded vowels there is an obvious difference, however, eg.:

singular	plural
koe	koeien	/kū/	/kūjə/	cow(s)
vloo	vlooien	/vlō/	/vlōjən/	flea(s)

Here we have a /j/ inserted not a /w/. Zonneveld explains this, however, by suggesting that we have here not an insertion in the plural forms but a deletion in the singular. He says that as far as the native vocabulary is concerned there is a constraint against the occurrence of monosyllabic nouns ending in a vowel. Whatever the origins of the words concerned may be, this constraint is certainly not true for present-day Dutch. No Dutch speaker will feel, for instance, that words like plee [ple:] (water-closet) or po [po:] (potty) are in any sense foreign-sounding or un-Dutch. In any case there are examples of purely Dutch nouns ending in a vowel, eg. ree [re:] (roe-deer), vee [ve:] (cattle), knie [kni] (knee). Clearly Zonneveld's analysis, with the underlying forms /kūj/ and /vlōj/, is still conceivable, whether or not we recognize the above constraint. It does have the advantage that these two words now form there plural regularly. Monosyllabic nouns ending in back vowels normally take the plural suffix -s while those ending in glides take the suffix -en, eg.:

singular	plural
roe	roe's	/rū/	/rūs/	rod(s)
po	po's	/pō/	/pōs/	potty(s)
sjah	sjah's	/ʃa/	/ʃas/	shah(s)
boei	boeien	/būj/	/būjən/	buoy(s)
kooi	kooien	/kōj/	/kōjən/	cage(s)
kraai	kraaien	/krāj/	/krājən/	crow(s)

Thus in my solution the words koe and vloo were exceptional in their plural suffixation, which is no longer the case in Zonneveld's analysis. Note, however, that Zonneveld has introduced a new irregularity in these words. They have now to undergo a rule of final glide deletion, which Zonneveld suggests may be needed in the grammar anyway to take care of relationships like the following:

zee	/zē/	sea	v.	Zeeuw	/zēw/	Zealander
Hebreeër	/hēbrējər/	Hebrew	n.	Hebreeuws	/hēbrēws/	Hebrew a.

The first case is perhaps more obious if we compare the forms Zeeuw (see above) and Zeeland /zēlant/ (Zealand).

However, this process of final glide deletion is clearly highly irregular - there are very few forms illustrating it, and what is perhaps more important, there are no forms (other than such as koe and vloo) where it plays any role in the inflectional morphology of Dutch. Thus instead of marking words like koe as exceptional with regard to a morphological rule (that spelling out the plural suffix) we would now have to mark them as exceptionally undergoing a very minor phonological rule of Dutch.

Zonneveld adds the example strooien /strōjən/ (made from straw) derived from the word stro /strō/ (straw) suggesting that this and the two words discussed above complete the list of such words in Dutch. Another relevant case would seem to be the word moe /mū/ (tired). This also has the archaic form /mūdə/ moede.

[p. 21]

We will, however, ignore this form, since it is probably not synchronically relevant. According to the dictionary, and a number of informants, moe has the following forms:

simple	-e form	comparative	superlative
mu	mūjə	mūjər	mūst

Once again Zonneveld would explain these forms as coming from underlying /mūj/. To do this however he would have to extend his morpheme-structure constraint to adjectives.

As far as Zonneveld's solution for the back rounded vowel cases is concerned, that it is a case of glide-deletion rather than a case of glide insertion, this is only possible because the forms he posits /kūj/, /vlōj/, /strōj/ involve vowel-glide clusters that are permitted by the morpheme-structure rules of Dutch. Permissible clusters are as follows:

front vowels		back vowels
w#	ȳw#	ūj#
ēw#		ōj#
		āj#

Let uw now examine the cases where glide-insertion takes place after front rounded vowels:

a) with /ȳ/
	individu	individuën	/indvdȳ/	/individyən/	individual(s)
	continu	continuë	/kontnȳ/	/kontnȳ ə/	continuous
	cru	cruë	/krȳ/	/kryə/	crude
	(some informants produced /j/-forms for the last two words, in general these formations seemed to be awkward)
b) with //
	reu	reuën	/r/	/rjən/	male dog(s)
	keu	keuën (keu's)	/k/	/kjən/	cue(s)
	bleu	bleuën	/bl/	bljə/	timid

By HG-INSERTION we should expect only forms with //. These seem to be the normal case with /ȳ/, though there is some disagreement among informants. With //, however, insertion of /j/ seems to be regular. Among some informants there was a tendency to insert /j/ in the Dutch words, but not in the French loans so as all the /ȳ/-forms and bleu. Zonneveld, however, would forecast only forms with // here. In addition his explanation for the back vowel cases will not work here either. He would have to derive the forms with /j/ from underlying: /rj/, /blj/. However the sequence /j#/ is not permitted by the morpheme-structure rules of Dutch, which permit only final combinations of front tense vowels with /w/ and back tense vowels with /j/. In other words it is difficult to see how we could deal with the //-forms otherwise than by means of a rule of J-INSERTION.

The /y/-forms would in our analysis would be exempted by a redundancy rule from undergoing J-INSERTION, for those speakers who do not have the /j/-forms.

[p. 22]

A further problem for Zonneveld's analysis arises with forms involving the diphtong ui [ä], eg.:

singular	plural
ui	uien	/ä/	/äjən/	onion(s)
kruid	kruien (or kruiden)	/krät/	/kräjən/	herb(s)

HG-INSERTION will not explain these forms. Zonneveld gives kruien the phonetic transcription [krʌən]. This is incorrect, however. The first portion of the ui-diphthong, which 1 write [ä] (see fn. 4 of Smith (1973)) is unrounded, and usually written [ʌ] by Dutch scholars. The second portion is front and rounded and is omitted in Zonneveld's transcription.

In order to explain forms like [äjən] Zonneveld would have on the one hand to posit a base form /äj/ for ui [ä] (onion) and on the other hand change his D-WEAKENING rule so that it applied after (certain) diphtongs - a position where he claims it does not apply, eg.:

/krädən/ → /kräjən/

There is some apparent support for this analysis. In final position some Dutch speakers pronounce ui as [äj] - in other words the glide changes from being rounded to being unrounded.

This analysis would have unfortunate consequences, however, as far as morpheme-structure is concerned. We would have to allow for sequences of lax vowels and two glides in morpheme-final position, which would clearly complicate the morpheme-structure rules.

To sum up, it seems that there is a body of evidence that J-INSERTION is required as a phonological rule of Dutch - a) Zonneveld's attempt to explain away forms like koeien, where a /j/ appears after back vowels is unconvincing, b) he cannot explain the /j/ appearing after the front rounded vowel //, c) he cannot explain the appearance of /j/ after the diphtong /ä/, d) he cannot explain the identity of the segmental environment in the cases of /d/-deletion and the cases of /də/-deletion.

The first three points here all can be explained if we accept the existence of a rule of J-INSERTION. Once we have accepted the need for this rule, we have beside d) above good reason to accept the reality of the rule of D-DELETION.

Bibliography

Chomsky, N. and M. Halle 1968 The Sound Pattern of English, Harper & Row.

Smith, N.S.H. 1973 ‘The Phenomenon of D-DELETION in Standard Dutch’, Spektator 2:7, pp. 421-37.

Zonneveld, W. 1975 ‘A Reanalysis of D-Deletion in Dutch’, Spektator 4-4, p.p. 231-9.