Spektator. Jaargang 6
(1976-1977)– [tijdschrift] Spektator. Tijdschrift voor Neerlandistiek–
Auteursrechtelijk beschermd
[pagina 27]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Nope 2: D-Deletion as Peyton Place
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
0. Introduction.Ga naar eind*This article has been written in the vein of a rich tradition. For over half a century now the subject of ‘intervocalic d’ in Dutch has occupied the attention of grammarians both Dutch and Flemish (Schönfeld, Kluyver, Van Haeringen, Mansion, Pauwels, Goemans, Van Loey, and others), and very recently interest has revived through a sequence of papers in Spektator approaching the subject from the generative phonological angle: Smith (1973), Zonneveld (1975), and Smith (1975). The present paper on intervocalic d has a threefold purpose, critical in one respect, and creative in two. Firstly, I would like to present a critical review of Norval Smith's 1975 paper ‘In Support of D-Deletion’ (Spektator 5:1), an answer to my paper ‘A Reanalysis of D-Deletion in Dutch’ (Spektator 4:4). I think Norval Smith's reaction to my paper fails to reach the standard set by Smith (1973), and I hope to be able to explain why. Secondly, this criticism contains as a more positive and purposeful part a first coherent attempt at a generative account of homorganic glide insertion in Dutch, a process I claim to be tied up closely with the subject of intervocalic d. Thirdly, I will briefly show how the recent exchange of papers between Norval Smith and myself has a striking parallel in the earlier history of work on intervocalic d, as a small contribution to, say, the history of Dutch linguistics. The first two aims will be pursued, as is understandable, more or less hand in hand: my criticism of Smith (1975) will be followed more or less in the same breath by my own proposals. The final paragraphs of this paper will be occupied with giving the contents of the pages preceding them a place in history. Throughout this paper the following shorthand labels will be used: Smith (1973) will be referred to as PHD, Zonneveld (1975) as NOPE, and Smith (1975) as YEP. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1. Nope 2I think it is clear that ‘D-Deletion’ in the title of YEP should be understood as an abbrevation of, roughly, ‘part of the analysis of D-Deletion as given in Smith (1973)’. In the first place, YEP reacts to only part of the contents of NOPE (there is no answer to NOPE's reanalysis of derivational constraints and | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 28]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
positive exceptions), and in the second place the text of YEP itself gives support to two, as far as I can see independent claims made in PHD: (i) Dutch phonology contains a rule of D-DELETION; and (ii) Dutch phonology contains a rule of J-INSERTIONGa naar eind1. More precisely, on the one hand Smith defends his assumption that there is one rule of D-DELETION (1)Ga naar eind2 operative in alternations such as those in (2),
while NOPE assumes a rule of D-WEAKENING (3) for the forms of (2a), and a rule of DE-DELETION (4) for those in (2b).
On the other hand, Smith defends an analysis involving a rule of J-INSERTION (5) for the forms of (2a).
while NOPE leans heavily on the interaction between the rule of D-WEAKENING (3) and a rule of HOMORGANIC GLIDE INSERTION (henceforth HGI; I will come back to the formulation of this particular rule below). Thus PHD and YEP derive goeie, deeën, and leer as in (6), while NOPE derives these forms as in (7).
Although the theoretical issues of derivational constraints and positive exceptions are of great interest, it seems unprofitable to go deeply into them here. Because of the lack of comments by Norval Smith in YEP, I would merely repeat myself. Those interested are referred to the relevant passages in PHD and NOPE. Fortunately, the issue of the number of rules affecting the d's in the forms of (2) can also be dealt with rather concisely. YEP repeats one argument from PHD | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 29]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
in favour of the ‘one rule’ analysis, and reacts to two arguments out of four in favour of the ‘two rules’ analysis of NOPE. Since the points concerned were somewhat illtreated in NOPE, let me go into them briefly here. Smith's argument in favour of the ‘one rule’ analysis is that it gives one ‘the chance of stating the generalisation that the segmental environment is the same in the two cases’ (YEP, 18). The point here is, of course, whether a generalisation based on formal similarity is in fact the correct generalisation. Given a set of linguistic data, we have no a priori method of deciding which, out of the infinitely many generalisations possible in principle, are the correct ones. At the present state of our abilities it is only by weighing one argument against another that we can try to arrive at what seem to be the generalisations. In NOPE I argued that in this particular case formal similarity seems to be overruled by other considerations: (i) there is a distinct difference in productivity between the two types of alternations in (2); (ii) there is a difference in style, where the alternations of (2a) do, but those in (2b) do not appear to be governed by casualness of speech; (3) deletion of de as in (2b) is in some cases accompanied by a modification of meaning; and (iv) an ordering anomaly is solved (and a derivational constraint avoided) by positing two different rules. Smith reacts to the first two of these observations: he suggests to account for the differences in productivity and style by means of ‘redundancy rules’ (PHD, 427; YEP, 18). This proposal is somewhat vague in view of the tasks and characteristics of such rules within the standard theoryGa naar eind4. Moreover, the redundancy rule (8) formulated in PHD, and intended to express the observation that the deletion of d before a boundary is ‘productive in a colloquial style’,
almost completely repeats the structural description of the rule of J-INSERTION (5). Does this capture the formal generalisation the ‘two rules’ analysis of NOPE fails to capture? Since it seems unlikely to me that the answer to this question should be ‘yes’, I still conclude that the ‘two rules’ analysis of NOPE of the data in (2) is to be preferred to the ‘one rule’ analysis of PHD. Let us then take up the second issue of YEP, the question of whether the alternations in (2a) should be accounted for by means of D-DELETION and J-INSERTION, as proposed in PHD, or by means of D-WEAKENING and HOMORGANIC GLIDE INSERTION, as suggested in NOPE. Smith's defense of PHD, which concentrates on independent motivation of J-INSERTION, has two sides: (i) the reanalysis of the strange koe/koeien and vlo/vlooien alternations given in NOPE is ‘unconvincing’ (22); and (ii) the rule of HGI, independently present in the phonology of Dutch (cf. PHD, 435, fn. 9; NOPE, 234), and proposed in NOPE to take over, in a manner of speaking, the task of J-INSERTION, fails to make the grade. Let me deal with these two claims in this order. PHD argues that the alternating forms of (9)
show that the insertion of a front glide after back vowels is motivated in Dutch phonology, and in fact these forms provide the basis of the D-DELETION/ | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 30]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
J-INSERTION analysis of the forms in (2a) as exemplified in (6)Ga naar eind5. NOPE adds to this pair the alternation stro/strooien ‘straw, n./adj.’, but argues at the same time that these forms should be analysed as having an underlying rather than an inserted glide. This underlying glide is proposed to disappear in singular through a rule of GLIDE DELETION, which appears to be needed independently. This reanalysis involving GLIDE DELETION has the virtue of regularizing the plurals koeien and vlooien (nouns ending in a back vowel generally take -s, not -en; nouns ending in a glide generally take -en), and of regularizing the underlying forms with respect to a proposed morpheme structure constraint (10), which forbids monosyllabic native formatives to end in a vowel.
YEP in its turn calls this reanalysis into question. It gives five exceptions to the morpheme structure constraint, and observes that ‘the process of final glide deletion is clearly highly irregular... and what is perhaps more important, there are no forms (other than such as koe and vlo) where it plays any role in the inflectional morphology of Dutch’ (20). I think none of these objections against NOPE's proposals on the alternations in (9) has any real substance. Although I do not think that a rule or constraint can be invalidated by citing exceptions, however many, I propose to leave constraint (10) out of discussion here, since it is hard to argue about a morpheme structure constraint which is not backed up by a reasonably deep analysis of Dutch phonology. Secondly, the fact that GLIDE DELETION is ‘highly irregular’ can only be admitted. Finally, I find the argument concerning the shortage of examples from inflectional morphology somewhat awkward, to say the least. I have no idea what it is based on, or what it intends to say. It is precisely the reality of the distinction between, and characteristics of, different types of morphology which is being debated in current work in generative phonology, and arguments based on a priori assumptions seem to me of little validity. I maintain, therefore, that the reanalysis of the alternations in (9) as outlined in NOPE is to be preferred to that of PHD. I will react to Smith's arguments concerning the invalidity of HGI as a replacement of J-INSERTION in a somewhat different manner. A point to be made beforehand is that very little appears to be known on the characteristics of the intervocalic hiatus in Dutch, and the rule of HGI presented as ‘a first approximation’ (234) in NOPE is highly impressionistic as a consequenceGa naar eind6. This time, let me be as precise as possible in my answer. Smith claims that NOPE's rule of HGI fails in two respects: (i) it does not explain the [j] appearing after the mid round front vowel [φ̅]; and (ii) it does not explain the appearance of [j] after the diphthong [äẅ]. Furthermore, Smith claims that a rule of J-INSERTION is required to explain these two j's.
It seems unavoidable to relate the first of these objections to the question of which glide is inserted after the high front rounded vowel [ȳ] in hiatus position. For me the facts are as follows: [w] appears after [ȳ] in all cases, although it alternates freely with [j] after stressed [ȳ]; [φ̅] appears only stressed and is obligatorily followed by [j]. Examples are displayed in (11). | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 31]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
These data can be accounted for by the pair of rules HOMORGANIC GLIDE INSERTION (HGI) (12), and GLIDE ASSIMILATION (13).
Sample derivations involving these two rules are given in (14).
The set of facts Smith wants to account for appears to be slightly different, as far as I can make out. The data he presents are scant and confusing, all we have is in fact the following (I correct printing errors):
I abstract from these data the following three groups of speakers: (i) ‘normal’ speakers who have [j] after [φ̅], and [ẅ] after [ȳ]; (ii) those who are like (i), but have [j] after stressed [ȳ]; and (iii) those who have [j] after Dutch [φ̅], and [ẅ] after French [ȳ] and [φ̅]. Notice how these observations can be incorporated into the present analysis. First of all, none of these speakers have a rule of GLASS. Secondly, they vary only as to a subportion, actually one feature, of the lefthand environment of HGI. In order to see this, consider the representations of these respective environments in (16a), and the sample derivations in (16b) (under the assumption of a structural change [α back, β round]). | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 32]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
This appears all very straightforward. The analysis outlined here in fact obviates the need for another redundancy rule proposed but left unformulated by Smith, exempting /ȳ/-forms from his rule of J-INSERTION. Given the rule of HGI as formulated in (12) it becomes easy to account for the homorganic transitional glides appearing after the diphthongs [ϵj], [äẅ], and [aw], which I admit were left lying in the dark in NOPE. All that is needed is to include the feature [+syll] within the segmental brackets already containing [-stress, α round] in (12), and within the corresponding brackets in (16a). This way, glide insertion after diphthongs will be governed by backness, as required: [j] will be inserted after [ϵj] and [äẅ], and [w] after [aw]. The inclusion of the feature [+syll] in the position suggested obviates the need of Smith's rule of W-ASSIMILATION (17) (cf. PHD, 423).
which was designed to account for the pair of back glides in ouwe, and similar forms with [aw]. (18a) gives the derivation of this form under Smith's analysis, (18b) shows the derivation presently proposed.
As will be clear, the over-all analysis outlined here not only shows that the glide [j] after [φ̅] and [äẅ] can be derived with the help of the rule of HOMORGANIC GLIDE INSERTION when formulated properly, but also, and again, it establishes the rule of J-INSERTION as proposed by Smith in PHD and YEP as a non-rule of Dutch phonology. Moreover, I finally conclude that the interacting rules of D-WEAKENING and HOMORGANIC GLIDE INSERTION are in fact able to account for the alternations of the type displayed in (2a). | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
2. Historical PerspectiveOutlined above is a generative account of the alternations deden/deeën, | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 33]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
broeder/broer, and so on, which makes use of rules of D-WEAKENING and HOMORGANIC GLIDE INSERTION for the former, and a rule of DE-DELETION for the latter type. It is an alternative (in my mind a preferable alternative) to an account involving rules of D-DELETION and J-INSERTION. Let me conclude this paper by pointing out an interesting historical parallel to this pair of alternatives. I would not wish one to look upon this section as presenting in any way more arguments for or against either alternative, I merely intend to show that in a broad sense the two analyses discussed correspond in a surprising manner to accounts given by some of our predecessors, albeit, of course, against completely different theoretical backgrounds, and with different purposes. With this in mind, let me give two quotes from C.B. van Haeringen's important 1926 paper ‘Intervocaliese D in het Nederlands’, where we find the first sketch of two schools of thought on the alternations in (2) above. Van Haeringen statesGa naar eind8: There is some uncertainty as regards the nature of j or w that appear in [the alternations of (2a)]. Some tend to think that the hiatus, created by the disappearance of d between the vowel preceding d and the one following d, was filled up by j or w. (3) He attributes the former interpretation to Te Winkel (1901, 53/4), Schrijnen (1920, 57ff.), and possibly Schönfeld (1920). The latter interpretation is attributed to Van Wijk (1907, 20), and Fijn van Draat (1923, 252ff.). Van Haeringen ranks himself among the advocates of the weakening theory, and comments: Can j be considered a transitional sound, a ‘glide’ developed into an independent sound? This can be imagined after a palatal vowel. Earlier mentioned forms such as zeeën and drieën illustrate this. Thus one could call the j in bieje ‘bieden’ very well, and that in gereje and beneje also, a transitional sound. After velar vowels, however, such a transitional sound is awkward... When one tries to imagine the words [geboden and hoeden] without d, and wants to speak without a ‘hiatus’, then the only glide phonetically imaginable is a w-sound. (7) Later he states that ‘d between vowels has (either been maintained), either been lost, or has developed into j’ (9), and finally Van Haeringen arrives at the following tripartite division (12/3).
This division is repeated in a second 1927 article by Van Haeringen, which concentrates on the historic geographical origins of the various changes, and is of less importance here. In the years following these two papers, the hiatus filling character of j and the d-deletion interpretation find a supporter in the Leuvenian scholar Goemans (1931), who backs up his position with data from the interesting dialect of Mechelen. He views j-insertion as governed by the palatality of the vowel following deleted d. This work seems to have gone largely unnoticed in the north, | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 34]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
and is pointed out only by Pauwels (1949), and Van Loey (1952). In a somewhat curious paper, Barnouw (1942) also rejects j as resulting from weakening. He seeks the source of the appearance of j in the influence of suffixal n, and in the pressure of the diphthongs oei, ooi, and aai. In a 1963 lecture, concerned principally with subregularities within the over-all data pertaining to intervocalic d, Van Haeringen reiterates his position of 1926, and he does the same in a contribution to the recent Van Loey Festschrift (1975), in a discussion of Barnouw's paper. Reviewing these works on intervocalic d, one will agree, I think, that there exists a surprising parallel between the two synchronic analyses sketched in section 1 above, and the two mainly diachronic accounts offered by earlier grammarians. Thus, the weakening hypothesis put forward by Van Haeringen since 1926, although nowhere an explicit account of the changes, or even meant as such, has strong overtones of the analysis defended in section 1, and especially the tripartite division by Van Haeringen displayed in (19) bears a striking resemblance to this analysis, with its rules of DE-DELETION, HOMORGANIC GLIDE INSERTION, and D-WEAKENING. On the other hand, the analysis proposed by Norval Smith in PHD and YEP, with rules of D-DELETION and J-INSERTION, has predecessors in scholars such as Te Winkel and Schrijnen, somewhat less so in Goemans and Barnouw. One wonders whether there is anything new under the sun. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 35]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Bibliography
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
