Spektator. Jaargang 11
(1981-1982)– [tijdschrift] Spektator. Tijdschrift voor Neerlandistiek–
Auteursrechtelijk beschermd
[pagina 342]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The descriptive power of the Dutch theme-vowel
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 343]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
For Schane these data motivate an analysis in which in the end subjunctive j'arrive must be derived from an extended underlying source as follows:
 in (a)) in the case of j'arri[v] in (c); this cannot be the person marker as liaison shows, and the assumption is therefore that arriver differs from vivre in a theme-vowel immediately after the stem underlyingly, deleted finally and before vowels.
The flow of this type of argument is expressed best in Schane's own words: [W]e have set up for each morpheme a unique underlying representation. We have shown how phonological information from outside the paradigm is often relevant for determining the underlying representation for morphemes which occur only within the paradigm. An underlying representation further more exhibits structural properties which are not overtly marked at the phonetic level. Such a representation of a sequence of morphemes is then subjected to the ordered set of morphophonemic rules. The end result is the phonetic representation of the sequence in question. (1966: 758). Schane's analysis will perhaps strike some as ‘abstract’ in the sense of the controversy of the same name, at least because of the convenient deletion of underlying schwas necessary to block phonological rules. On the other hand, it has survived the abstractness debate scratchless; it was not mentioned among objectionable analyses in Kiparsky (1968), or on any other list since. Selkirk (1972) still attaches to it epithet ‘correct’, and its essential features are maintained in the most recent elaborate overview of French phonology, Dell (1980). This seems to be a proper state of affairs, yet slightly paradoxical. Perhaps the very simple reason is that Schane (1966, 1972) admits that final schwa deletion is ‘optional’ depending on the ‘variety’ and ‘style’ of French (‘southern dialects’ and ‘nursery rhymes’ have final schwa), which may have influenced the verdict on the degree of abstractness of this analysis. Furthermore, Dell (1980) adds that the theme-vowel in fact shows up in some restricted morphophonological environments (before the future marker after obstruent-liquid clusters: ‘manoeuvrera’ (he will manoeuver)). It would be odd, however, to sustain the view that precisely the existence of these almost accidental cases save Schane's analysis from the abstractness axe, thereby claiming that if they did not exist Schane's analysis would be ruled out. The | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 344]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
much more important point is of course that the effects of the underlying schwas show up in all varieties and dialects, that the generalizations are the same throughout, and that the differences are only superficial. If this is accepted there appears to be no reason not to accept the view that a system comparable to French but accidentally without surface schwas should be able to stand on its own without the support of related dialects or styles, given an amount of morphophonemic evidence at least as varied and rich. Precisely this appears to be the case in Dutch, that is, an analysis where schwa is abundantly supported morphophonemically, but superficially completely absent, will be developed in section 7. of this paper on the basis of the data presented in 1. -6. This describes the two major aims of this paper, but not its full contents. In some recent publications, Booij (1979, 1980), and Kooij (1981), both known as adherents of more ‘concrete’ versions of phonological thought, have advanced a number of objections to earlier expositions on some of the material presented here, and have given alternatives for some subcases. In an ‘Appendix’ to this paper I will discuss their views, and show why they are very likely to be incorrect. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1. Open syllable lengthening.The phenomenon that illustrates best the way in which verbs distinguish themselves from nouns and adjectives in the way intended here is the so-called lengthening in open syllables. The rule may be formulated as:
It will be clear that for nouns in...VC- this rule will apply if plural -an is added. But a further generalization holds which may be captured as in (1):
The following examples illustrate this generalization:
It appears that verbs can undergo lengthening even in the absence of contextual -ə. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
2. Intervocalic D-weakening.In Western dialects of Dutch the stop d weakens in intervocalic surroundings. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 345]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In an informal notation this phenomenon allows itself to be expressed as follows:
The application of this rule appears to be constrained in many strange ways, some phonological, some extra-phonological, i.e. grammatical and sociolinguistic. Howeverl the following generlization seems to hold whatever these further conditions:
This can be illustrated as follows:
Thus, verbs can undergo weakening even in the absence of contextual schwa. Given this, one is not surprised to find that the verb baden - see above - can be used as an illustration of both generalizations: baai-t/-en/-de. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
3. Prevocalic schwa-deletion.In Dutch, schwa is deleted exceptionlessly before another vowel:
Some of the many examples are as follows:
Given this, the following generalization holds:
Examples of this phenomenon run as follows: | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 346]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Forms such as aarden can be hypothesized to derive from aarda-an, etc. The same could be expected to hold for the plurals of the nouns in the leftmost column were it not for the fact that these do not naturally adopt the -an plural: they prefer -s. Another expectation not borne out is the lack of d-weakening among the several candidates offering themselves here. It may or may not turn out that counterfeeding rule ordering is the mechanism to capture this. At any rate, the examples show that stems schwa-final elsewhere delete schwa both regularly (before V) and unexpectedly (finally and before C) in verb-forms. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
4. Post-schwa n-deletion.Dutch has a very productive phonological rule deleting n from the configurations ən# and əns. Thus for instance ope /opening (id.), Leuve ‘Louvain’/Leuve s (adj.)/Leuvenaar (inh.), but ochtend ‘morning’, arend ‘eagle’, etc. This rule also applies freely to all instances of inflectional -ən discussed so far in this paper. It may be formulated as follows:
Concerning these sequences the following generalization seems to hold:
Consider in this respect the following examples:
Of course, nodeletion of n follows from the formulation of the rule as soon as a nonsibilant suffix follows. To be explained, however, is the perseverance of n in the two unextended verb-forms open (e.g.), that is, 1 sg.pres.indic. ik open | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 347]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
and imp.sg. open! Perhaps most impressive in this respect is the fact that the following cases are ambiguous only by virtue of poor orthography; in normal speech, if n is deleted they cannot have a verbal interpretation:
Note that in sections 1.-3. we wanted to explain why in particular in verbal inflection phonological rules apparently applied to strings that did not strictly speaking meet their structural descriptions. This time we must explain the opposite: why is a phonological rule not applied to a form that in all relevant respects meets its structural description perfectly? And again the forms involved are verbal ones. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
5. Word final j-deletion.It is not uncommon to find in the literature the singular-plural pairs below described via a rule inserting j before the plural suffix -ən:
If this were to count as the correct view on these pairs, however, one would be at a loss to square it with the following two observations: first, backvowel-final nouns never take the ən-plural, and second, back vowels before another vowel always go via w. See the following examples:
The much more obvious account is therefore to suppose that the lexical stems koei and vlooi will be predestined to lose their i word-finally. This immediately explains their plurals since all nominal stems of this type take -ən: boei-en ‘buoys’, kooi-en ‘cages’, fooi-en ‘tips’, etc. As a bonus these underlying representations can be related rather naturally to the popular ‘analogical’ singulars koei and vlooi if the existence of the latter is seen as the result of a lexical marking purposefully deleted. This same mechanism may apply to a form moei underlying the adjectival alternation moe ‘tired’ - moeie (infl.) in those instances where ‘analogical’ moei is in fact heard. We therefore propose a rule of word-final j-deletion as a part of Dutch phonology. Stems will be lexically marked to undergo it. Then, consider the following observation to be explained:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 348]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
This is so because of the following data:
Just as under 4. an explanation is required here for a nonevent: vlo, moe, and vloe do exist as nouns, not, however, als verbal forms: there we find ik vloei af, ik vlooi, vermoei (me niet)!(‘don't tire me’), etc. Note in the meantime that the argument presented here for a j-deletion analysis of these forms is not crucial for the view that again the verbs are the odd ones out. If these are cases of i-insertion before -ən one would still be led to wonder about the source of the same i-full verbforms without schwa as a trigger. The puzzle is now of type 1.-3., but the pieces still don't fully it. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
6. Assimilation of d.There is an extensive literature on the assimilatory behaviour of the Dutch voiced plosive d. I will not begin to discuss it here. On the other hand, a number of the observations cited there must be briefly mentioned as a prerequisite for a discussion of the last peculiar property of verbs to be discussed below. The first observation is so obvious and well-known that perhaps its relevance tends to come through as insignificance: the past tense suffix has the shape -de after voiced segments (sonorants and voiced obstruents), and -te after voiceless ones (voiceless obstruents). Furthermore, this suffix is unique in this respect, as the following data show:
Add to this the observation that in Dutch plosives as a rule show the behaviour of the (a)-examples (they cause regressive assimilation), while fricatives as a rule show the behaviour of the past tense suffix (they themselves are the victims):
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 349]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 (although I am convinced that it follows from the data presented here that the past tense consonant in fact is a fricative, this is irrelevant to this exposition; nothing is clarified and much confused by taking this independent step here). Given the language-particular rule of progressive fricative devoicing, and the usual Indo-European rule of regressive voicing assimilation for obstruents ordered after it, we now obtain the following sample derivations:
This much in mind we proceed to a further series of slightly less preliminary observations. Let us take that variant of Western Dutch that allows vacillation of voice in obstruent-clusters in for instance the following cases:
 -. This gives derivations of the following nature:
The curious fact now occurs that in the current dialect all clusters may be either voiced or voiceless, as we have just seen, but this does not include the plosive dental clusters. Thus, as a supplement, the following data exist:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 350]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
This point is particularly spectacular in those cases where an expected ambiguity between [Adj...d -e] and [Adj...  - de...] (where  -d is a simple case of degemination) is not borne out:
 - in a dental plosive environment, sending both d- and t- to t , as follows:
It will be worthwhile to have a look now at the behaviour of the 3 sg. inverted pronoun ie in this dialect, the mate of sentence-initial ‘hij’. For many speakers the following data obtain (gloss: ‘did he Verb?’):
These data derive their interest from the top row middle column three. The lack of final devoicing characteristic of Dutch elsewhere has led some to believe that -ie triggers a rule of fricative voicing in intervocalic surroundings. Since the fricatives of the three above examples are underlyingly voiced (1 sg. ik ga[f]/1 pl. wij ga[v]-en), the true test for this proposal lies in cases of underlying voiceless fricatives followed by -ie. Unfortunately, these are very scarce, since in nearly all cases suffixes of weak verb inflection intervene: slof-t-ie ‘does he shuffle’ and slof-te-die ‘did he shuffle’ (1 sg. ik slof/1 pl. wij sloffen), etc., do not of course decide the issue either way. Eliminating these, we have in fact at our disposal only the following two exemplary verbs:
The lack of intervocalic voicing in precisely these cases can be taken as the knock-out blow for the intervocalic voicing analysis. We thus proceed to say that 3 sg. (d)ie in this dialect attaches to underlying verbforms immediately, the | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 351]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
consonant surfacing after schwa. On this basis we fully expect the following in this dialect to be possible:
 ie-rule has its proper place in the set of rules developed so far:
Notice that past tense -
 ie-rule. There, representative data of the following kind can be found:
A pronoun -
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 352]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
but, of course, is forced to allow them in cases such as:
In order to see this point best these sentences should be read with unstressed die, that is, with neutral final emphasis, or even strong final emphasis of incredulity. Emphatic die automatically triggers demonstrative interpretation which allows vacillation in (a-c). This holds two ways: either die qualifies the object NP, or it refers back to earlier hij:
Put together these observations indicate that vacillation is a prerogative of a syntactico-semantic class excluding, at least, past tense and 3 sg., and including on the other hand, as far as one can tell, demonstratives and a number of spinoffs such as both dan's and the definite article de. It is worth pointing out, in our final set of data, that the latter is involved in a most intriguing phenomenon, which seems too fascinating to go unnoticed here. Beside this intrinsic interest it provides a striking piece of confirmation for the existence of what will below turn out to be the central rule of this section, the rule of t-zero. Notice first, in this respect, that Dutch allows semi-idiomatic PP's as below either with or without an intervening modifier:
Some prepositions allow case-marking to appear under these circumstances. Sometimes the result is written as a single word:
The curious fact now occurs that those prepositions in -t which allow this construction, delete the initial d- of the following article:
Of course, our t-zero rule immediately explains this. Even more curious is the behaviour of the preposition te in these constructions. First, there is not a single | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 353]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
case where te is followed by the article de. Otherwise, the following occurs:
The difference between ter and ten is partly idiosyncratic. Het-words always take ten, de-words take one or the other. This time the behaviour of te follows under the minimal assumption of a rule of allomorphy which deletes its schwa from the context te-  - as a telltale symbol for this type of ‘behaves as’ phenomenon. It holds optionally for demonstratives etc., where we found additionally that d/t-plus- give obligatory t. This point was then sustained with the observation that in the dialect where 3 sg. attaches immediately to the underlying verb-form, the expected...d-ie forms are out, and only...t-ie occurs. Another illustration of the same phenomenon was found in the behaviour of the preposition te in idiomatic PP's. Given this account, and if our several morals drawn from it are accurate, we are now in a position to formulate our sixth generalization-im-Bunde on Dutch verb-forms:
In other words, we must explain in our account why the facts about past tense formation for weak verbs ending in...d are as follows:
That is, why do we have at the same time bood- | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 354]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
7. Why Weak Verbs don't exist.In one way the above array of observations about Dutch is but a haphazard pick out of the infinite number of observations that can in principle be made about this language. In another, much more significant sense, it is of course not. In fact, I propose that the facts presented in sections 1.-6., and more especially generalizations (1)-(6), should form a coherent system in any phonological description of Dutch worth that title. We have not hesitated, in this respect, to stress time and again the conspicuous role of the lexical category Verb in all six generalizations enumerated. Furthermore, if we assume that the highest degree of coherence is gained by the system that finds one single source underlying these generalizations about Verbs, it follows that there must be a single phonological property, hitherto unspecified, from which they can be predicted. It will be the purpose of this section to show that this type of reasoning fits exeptionally well this particular case. In Zonneveld (1978a), where sections 1.-3. were presented in a more elaborate form, the behaviour of inflectional verb-forms towards the phenomenon of d-weakening was taken as the point of departure. This was accidental in that a description of this behaviour was not the prime object of that work, but fortunate in that it provided an immediate clue to the source of the lengthening and schwa-deletion phenomena. The line of reasoning applied was simply this: if the rule of d-weakening applies before schwa, and if inflectional verb-forms are subject to the rule both in the presence and in the absence of schwa, this shows that inflectional verb-forms must always have schwa. Derivations were postulated as follows:
where e is an extra vowel to be discussed further below. Note that it disappears before an obstruent and word-finally. Without further machinery this analysis accounts for the lengthening and schwa-deletion cases:
This result was presented more tightly in Zonneveld (1978b). Notice that all of these cases can be described as motivation for a covert trigger. In Trommelen and Zonneveld (1979) was added a case of a covert umbrella, the open-examples of section 4. Without explicitly saying so, the authors of course expected such a case to exist on the basis of the properties of the first three. Again, as was | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 355]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
pointed out, not a single additional assumption is required in order to bring this phenomenon under the scope of the already existing machinery:
A novelty in this work, however, was the addition of the insight that there is no reason to constrain the extra schwa to those instances of verbal inflection where it will be deleted, that is, to the pre-obstruent and word-final environments. In fact, the most general hypothesis is that each inflectional verb-form is followed by the e, the theme-vowel from then on, which is deleted after having done its job both as a trigger and as an umbrella by the simple instruction reading ‘delete theme-vowel’. The relevant new derivations are now as follows:
 is bled by prevocalic schwa deletion in all these cases, but this state of affairs is independently motivated: it is the only way to derive ik weid from ik weide-e.
This paper brings two new groups of data under the scope of this analysis, and again without charge. The procedure will now be familiar, and relevant sample derivations will speak for themselves:
Again, the theme vowel is used as an umbrella in order to protect the end of the stem from being affected. This is precisely the situation of the open-phenomena. This, and the reverse situation found in cases of d-weakening, lengthening, and schwa-deletion are handled by the single mechanism of the extra theme vowel. The reduction is complete. Be this as it may, the ‘allegro’ (if not ‘fortissimo’) character and wide scope of this exposition should not dull our critical attitude towards its details. At least one of these, in fact, warrants serious discussion. The motivation of the illfated theme-vowels seems straightforward for the cases discussed under 1.-5. In 6., however, there is a source of inconsistency which becomes apparent upon reinspection of our last three sample derivations. Earlier we have made the literal claim that each inflectional verb-form will be followed by the theme-vowel. Why is it, then, that in the last set of sample derivations, only the second and third have such a vowel, and not so the first? Notice that this is not an accidental oversight, but a necessary situation; on the one hand the theme- | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 356]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
vowel must be present in two and three in order to block t-zero, on the other it must be absent from the first in order to allow the same rule to apply. This seems strange, since the first is just as inflectional as the other two. There is a difference, however, and a crucial one, I should like to suggest. It is a simple fact of the above sections that the motivation for the existence of an underlying theme-vowel comes from weak verbs. Only these, and not the strong (irregular, ablaut, etc.) ones have been shown to be involved in the phenomena leading to the extra schwa. Vond, however, featuring immediately above, is the past tense of vinden and is strong. We have here a first indication, then, that strong verbs do not take the theme-vowel. The reader may well recall further indications from the text in that also stond(t-ie), bood(t-ie), and had(t-ie) undergo t-zero, and therefore cannot have a theme-vowel at that point of their existence. Since all are strong, our hypothesis is sustained. But there is more. The only examples of strong verbs motivating the theme-vowel among the phenomena of sections 1.-5. are those given by Zonneveld (1978a) for some cases of d-weakening. Especially, the verbs showing ij/ee-ablaut are especially prone to the process, although not exclusively these:
In Zonneveld (1978a) the position was taken that the ij/ee-ablaut is among the morphological circumstances favouring the application of the rule, and that some verbs may be individually marked to the same end. But this requires a theme-vowel to be present on strong verbs, a position we are now very much willing to vacate. The alternative is as follows. The observation that these verbs undergo so extremely easily the otherwise laborious rule of d-weakening must be seen as an indication that they are not subject to that rule. As has been pointed out before, pronunciations such as ik glij[t], snij[t], hou[t] are not even avoided, they are completely out. Wij glijden, sneden, houden, and deden, on the other hand, do conform to the usual conditions on d-weakening. We therefore mark these verbs as subject to insertion of d before schwa. If ordered before d-weakening itself, this rule gives precisely the results required:
This analysisGa naar eind1, which is motivated by internal observations regarding d-weakening, of course requires that there be no theme-vowel underlyingly in forms such as ik glij, and this is of course the strongest type of evidence one can hope to find. The evidence available goes even further than this if we note that imperatives behave as expected: glij!, rij!, and so on, and that the underlying stem surfaces in a number of seemingly ad hoc compounds:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 357]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The next step to take on the basis of this converging evidence is obviously to dispose ourselves of the notion ‘weak verb’ of traditional grammar. There is no reason, for instance, to suppose that the statement ‘all weak verbs have a theme vowel (and all strong verbs have not)’ is of any relevance to a proper description of Dutch. The notion ‘weak verb’ is simply a highly informal nomenclature for those verbs of Dutch taking a theme-vowel underlyingly. Thus, verbs that are completely neutral from the point of view of this paper and weak, take the theme-vowel in a completely harmless manner. Talmen is an arbitrary example:
Dutch morphology is therefore enriched with the statement that ‘ablaut is nonthematic’, and everything else is automatic, i.e. unmarked. At least one further extension of this position is possible. There is a set of ablaut verbs in Dutch which show, besides, alternation between a-aa in past tense:
To say that these verbs undergo ablaut in their past tense paradigm, however, may well be another informal nomenclature from traditional grammar. Recall that we have a means at our disposal which can take this phenomenon under its wings in the rule of lengthening discussed in section 1. This rule, and not the ablaut description, explains why length surfaces in plural and lack of it in singular, and not vice versa. But if these are cases of lengthening, it is again necessary to deny them the underlying theme vowel, so as not to derive *ik spraak and so on. Since these verbs are strong by virtue of their present-past ablaut, this is again precisely the result expected in the analysis defended here. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
8. Conclusion.In this paper we have tried to explain why Dutch verbal phonology is so uneventful as compared to that of the other syntactic classes. We named the theme-vowel, a covert morphological element that hides its existence well and is betrayed only by the effect it has on its phonological environment, as the culprit. It explains why verb-stems tend to maintain their integrity so strongly, on the one hand by forcing entire verbal paradigms to undergo phonological rules triggered by schwa, and on the other by protecting them from the disturbing effects of phonological rules conditioned by the word-edge. We investigated six phenomena that lead one to suspect the influence of the theme-vowel: open syllable lengthening, d-weakening, prevocalic schwa-deletion, | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 358]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
post-schwa word-final n-deletion, word-final j-deletion, and t-zero. The lastmentioned rule gave us a crucial clue as to one further important property of the theme-vowel: it may well turn out to be the sole distinguishing property for strong and weak verbs in Dutch. This hypothesis is most desirable in that it gives one the opportunity to say that all Dutch verbs have a theme-vowel and hence show uniform paradigms, and that strong verbs are simply the ones specified as exceptions to this unmarked situation. Thus only the latter are a burden upon the grammar, which is precisely the state of affairs one should wish to obtain a priori. In the introduction to this paper we referred back suggestively to the analysis of French verbal phonology in Schane (1966). This appeal was neither insignificant nor accidental, in the sense that French verbal phonology is traditionally a famous ‘testing-ground’ for morpho-phonological theories (see Trager 1944, 1955; Bloomfield 1945; Hall 1948, etc.). In fact, as is well-known, also ‘modern’ linguistics owes a great deal to the language of de Saussure and Schane, in particular to its ‘liaison’-phenomenon. Thus, in Selkirk (1972: 246 ff.), the difference between the two sentences below, where liaison is to be blocked in (b), was explained as the result of a trace-effect of the moved constituent:
Furthermore, we have seen how Schane (1966, 1968) explains the properties of some phonological phenomena, partly involving liaison, by assuming a much more elaborate morphological structure than a superficial glance at the phonetics would lead one to suspect. To do so gives one information not only about the morphophonology of French, but of course also about phonological theory, as this is now given the authority to sanction such an analysis and its characteristics. Thus, if the analysis is accepted, one gets with it much that has been called into question in some recent phonological discussions: a phonological representation can be relatively far removed from its corresponding phonetic output; the link-up is accomplished by a set of at least partially ordered phonological rules; there will be a considerable amount of interaction between phonology on the one hand, and morphology and syntax on the other; syntactic surface structure will have to be prearranged so as to serve as an input for morphophonology; and so on. Much of this is discussed in Schane (1967), Selkirk (1972), Rothenberg (1978), Dell (1980), and so on. The arguments developed in this paper may well be viewed as claiming that morphophonologically Dutch is in many ways an extreme variety of French, and hence as providing support for the type of phonological theory that underlies the analyses of both languages. Upon comparison we find that our proposals are based upon phenomena of one type out of two, that is, we have surveyed the area between phonology and morphology rather than that between phonology and syntax. As the evidence develops, however, there is no reason a priori to think that Dutch will not exhibit phenomena of the latter type as well. We find an indication of this in one type of case involved in the phenomenon of fricative assimilation illustrated in section 6. by examples such as diep[s]ee and | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 359]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
mat[x]las. Notice in this respect the contrast between the initial fricatives of the following verb-first sentences:
If this contrast is correct observationally (although subtle it seemed to hold at least for a number of people I have consulted), this may well be an example of unexpected interaction between phonology and syntax in Dutch. To search this area as well can be seen as a fruitful course to take in future research. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
9. Appendix.In the eight sections of the main body of this paper, I have presented an analysis of Dutch verbal phonology in the belief that the six independent phonological phenomena discussed there all point towards a single generalization: the existence of a morphological property called the theme-vowel peculiar to Dutch verbs. As pointed out in the introduction, some portions of the analysis, specifically the first four phenomena, have been presented earlier in Zonneveld (1978a, 1978b), and Trommelen and Zonneveld (1979), although with different aims. These expositions in their turn have been commented upon twice in the recent literature - where two reviews by Booij (1979, 1980) count as one, and Kooij (1981) as the other -, with the common conclusion that the postulation of a theme-vowel in the morpho-phonology of Dutch is unwarranted, and that the generalizations towards it can or should be explained otherwise. It would obviously be unfortunate for our view if their objections were to qualify as valid. I strongly believe, however, that there is no reason to hold this pessimistic view and, since some of the objections of B. and K. may have come to mind to others in reading this paper, it may be instructive to show here why. As the authors approach the subject from separate angles, the exposition will take place in two subsections. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
9.1. Appendix a.The main theme in B.'s two critical notes on the analysis of Dutch verbphonology in Zonneveld (1978a) and Trommelen and Zonneveld (1979) is that it appears to be too abstract in the sense of the phonological controversy of the same name initiated by Kiparsky (1968). As is by now common knowledge, the latter paper argues that analyses involving the following art-work: should be prohibited by phonological theory. The arrow is sometimes called ‘absolute neutralization’, and ‘B’ an ‘imaginary segment’. Consider on this the following two passages from Booij (1979): | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 360]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[the theme-vowel analysis] implies the diacritic use of phonological features, i.e. absolute neutralization. In my opinion Kiparsky's [1968] objections against absolute neutralization are well-motivated...(257) There are two possible reactions to this. First, it is not clear to me that B.'s appeal to the ‘abstractness controversy’ is at all relevant here. As B. admits, the rule of theme vowel deletion assumed in the analysis proposed here (see for instance the sample derivations in section 7.) is not context-free in Kiparsky's sense, but deletes schwa in the morphological context theme. Second, I do not see that the rule maps the theme-schwa onto some other, underlying, phonological segment, as required in B.'s argument. And finally, my proposal implies a diacritically marked vowel, and not an imaginary, otherwise nonexistent vowel called in for precisely the situation at hand; of course, a diacritic marking is precisely Kiparsky's cure for abstract illnesses. The other avenue focuses on the issue from a more general angle. First, as Kiparsky has shown to be well aware in later publications, one can speak only of ‘the diacritic use of phonological features’ if only a single phonological rule motivates one to do so. But, we may ‘assume that wholly abstract segments are to be allowed when more than one rule refers to them crucially’ (1971: 585). In Kiparsky (1973) one finds a list of analyses of this latter type, judged to be ‘internally well-motivated’. Among them is the celebrated example from Yawelmani originally due to Kisseberth. Since the analysis defended here is also of this type, the reference to Kiparsky's work to support this objection, even if applicable, is in fact immaterial. Second, the references to Crothers and Hyman are odd in that neither author in fact has an axe to grind in this issue: they discuss it at various levels of lucidity, but cannot be said to reach a verdict either way. Hyman is of course known as the proponent of a very abstract analysis of part of the phonology of the Nupe language, and after a defense in his 1975 introduction, he winds up with the careful contention that ‘[j]ust how abstract phonology is remains a question that has yet to be answered in a manner satisfactory to all’ (90). Crothers' attitude is not very much different. He briefly discusses, e.g., a case from Turkish where a consonant that never surfaces figures in two phonological processes. Comparing descriptions with or without such a consonant, he concludes that: ‘What is needed in a case like this is not more discussion of which solution is the simplest or the best, but more facts about how the speaker of Turkish actually handles the situation, or how speakers of other languages handle similar ones’ (5). More facts is precisely what is available in our case, since the analysis is based on six independent phenomena. The similar situations in other languages exist in Yawelmani, Nupe, and so on. As a conclusion of this first brief appendix, then, we may say that our original idea expressed in the introduction that the questions of Dutch verbal phonology raised here can be most fruitfully discussed without being overly worried about matters of abstractness, is well-supported by closer examination of this area. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
9.2. Appendix b.In an interesting recent paper, Kooij (1981) discusses the properties of Dutch | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 361]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
verbal phonology as they appear from the phenomena discussed in Zonneveld (1978a), against the background of more general characteristics of the phenomenon of analogy. That is to say, in none of the three cases of sections 1.- 3. can the theme-vowel be held responsible for the quiet phonological life of Dutch verbs, but in two (open syllable lengthening, and prevocalic schwa deletion) there is a better (‘lexical’) analysis without the theme-schwa, and in the third he puts forward the idea that (e.g. in historical change) verbs qua lexical category strive for uniformity of their stems. His aim is to show that these suggestions not only explain the relevant facts, but actually explain them in a better way. From many points of view K.'s proposals are interesting enough to justify more elaborate treatment. The conclusion, however, will again be that there is no reason to be convinced by them. The order of discussion will be that of sections 1. and 3. together, followed by 2. K. discusses the facts of open syllable lengthening and prevocalic schwa deletion in almost precisely the same way, with objections to our analysis followed by suggestions for a better one. In both cases the outcome is a ‘lexical’ rule dealing separately with the phenomenon at hand. Let us see how this point is reached. In section 1. I have called the facts of open syllable lengthening those illustrating best the type of phenomenon to be explained here. K. appears not to share this view, and his first objection is as follows: ‘there is no good reason to derive the verbal forms ik baad, ik daal from the nominal ones, bad and dal, resp.’ (97) [my transl. W.Z.]. It is simple to agree to this, since the point of section 1. is that native speakers will tend to search for a common source underlying the data ik baad, wij baden, bad, baden, and so on. It is only by accident that the nominal singulars resemble most closely this common source at the surface. A similar objection to the proposed role of the theme-vowel in cases of prevocalic schwa deletion can be dismissed in a similar manner. K. attributes to Zonneveld (1978a) the ‘stubborn’ view that ‘in the lexicon there is a relation between the noun aarde and the verb aarden such that aarden is the derived element and aarde the dominant element.’ (98). Under this slightly odd terminology seems to go hidden some confusion as regards my actual argument, which was that, given the abundant alternations of the ambassad For K. the crucial step towards a different approach seems to rely on some recent claims of Hooper (1979), who, as a proponent of so-called ‘natural’ phonology rejects the common-source approach in favour of lexical relations between different words. K. notes: ‘Hooper (1979: 119-22) illustrates with several examples from language acquisition that children do not see the need to derive related, but paradigmatically different, words from a common source’ (97-8). The plural is presumably a figure of speech, since Hooper provides only one example. But apart from this, K. appears to be attacking here directly a position close to the one we have committed ourselves to just above. Therefore, it is worth pointing out that the volume in which Hooper (1979) appears also contains a number of very critical comments on her views by Harris. His point is, among others, that there can be no direct leap from language acquisition data to the organization of grammars of more mature stages. To quote the relevant passage: | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 362]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The considerable developmental literature accumulated over the last decades suggests that the acquisition process involves a series of successive approximations to adult grammars, and that some adjacent stages involve discontinuities that might well be characterized as ‘radical’. Be this as it may, to proceed as Hooper seems to suggest would be heuristically disastrous...(288) If Harris is right - and the available evidence compiled by e.g. Menn seems to point in this direction - K. can clearly not derive an argument pro his position from this particular reference. In the meantime, however, in order to deal with the facts of Dutch within a Hooperesque framework, K. makes the following proposals. For cases of open syllable lengthening there is to be a lexical redundancy rule which ‘expresses the fact that in a limited and countable number of cases the nouns have a short vowel in singular and the verbal paradigms have a long vowel throughout. There is no rule of open syllable lengthening in Dutch’ (97). But this is partly terminological and partly more serious. It is the former because the ‘name of’ open syllable lengthening could be given to K.'s ‘lexical’ rule without further consequences. But also notice that the redundancy rule gives no information on the length of the vowel in the nominal plurals. This may be an oversight, but it is one which happens to reveal a serious drawback of the redundancy ruleapproach, one I am not the first to point out. Let us say that the correct rule reads, partly: ‘there are long vowels in nominal plurals and throughout verbal paradigms’. The point is that this conjunction is just as good as: there are long vowels in nominal singulars and throughout verbal paradigms. But this one is incorrect, and excluded in principle in our account because nominal singulars go uninflected. Given the principle that nonoccurring situations should be excluded to the benefit of occurring ones, the observationally correct version of K.'s rule is descriptively inadequate. If we rephrase this slightly, we see that K. seems to confuse two things. The number of cases marked for lengthening is quite ‘limited’ [beperkt] which indicates that it is a minor rule (as Hans Gilijamse has pointed out, a proposal adopted in Zonneveld 1978a). At the same time the real issue is that there is an apparently unlimited generalization to be expressed: if a noun undergoes lengthening in plural, the related verb undergoes lengthening throughout. This is a restatement of generalization (1), and this is what is left unexpressed in K.'s alternative. By and large, the same comments are applicable to K.'s solution for the verbal cases of prevocalic schwa deletion. In the analysis proposed here the phonological content of the theme-vowel is derived independently from some of the rules that motivate it: prevocalic schwa deletion and d-weakening both show that the theme-vowel must be schwa and no other phonological segment. K.'s rule of verb-final schwa-deletion, on the other hand, might have been a rule of p-deletion, or a-deletion, or any insertion, or what have you, and no relation is made with independent mechanisms. In itself it strikes one as commendable that K. takes the risk of drawing on the awkward notion of ‘analogy’ to deal with the final cases, those of d-weakening. This time, K. offers not a rule, but a set of principles. Since they are, as neither uncommon nor very reproachful in this field, both implicit and unformalized, the following is only a brief paraphrase of a sequence of paragraphs from K.'s text:Ga naar eind2 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 363]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
These are proposed to explain why d-weakening, which originally applies in a two-sided vocalic environment, may overapply in a single-sided one: braden → braaien and braad → braai; the second clause is intended to block goede → goeie spreading to goed → *goei. While the intentions are clear, then, it seems at the same time that a proper evaluation of this proposal cannot be given under its current level of development. K. seems to anticipate this when he writes that this is only a ‘first modest step’ towards an approach of these problems under his own views. This is fair enough but should not keep one from enquiring at least about the following. First, there is the question of the independent motivation of the two principles as to their operation in Dutch phonology. One would very much like to see another phenomenon discussed in the same terms, since its support would add to the credibility of K.'s suggestion. This is closely related to a second point, which brings us to the question of the domain of the two principles: are they diachronic or synchronic, that is, do they describe or predict a historical development, or do they make predictions on a single stage where the presence of data A will entail the presence of data B? This is not clear from K.'s description, so we will have to look at either possibility. If the principles are diachronic, K.'s attempt to explain the phenomenon this way is ill-chosen, since we are fairly sure that at the stage of Dutch where d-weakening was most productive (let us say Middle Dutch for want of more detail), there was an overt theme-vowel in many environments. This seriously weakens the case for the role of analogy in this case, and leaves us as yet without an explanation for the Modern Dutch facts as well. Suppose then that the principles are synchronic. From there we may ask further questions, for instance the following third one. Recall from above that Dutch has a (completely automatic) process of homorganic glide insertion, giving for instance:
This process may well be viewed as affecting the stem in the infinitive, since the quality of the inserted segment is predictable from the stem-vowel. If this can be maintained, the fact that this segment does not show up analogically in the singulars runs counter to the principles. At any rate the notion of stem seems to require further definition, since it is highly unlikely that the inserted segment is part of the suffix. This again brings us to another, fourth, question. There is at least one point in K.'s article where the author equates the notions of ‘infinitive’ and ‘citation-form’ for verbs. This adds vagueness in that there appear to be many verbs for which this does not hold, especially the ones with prepositions ‘as prefixes’ such as opnemen ‘to accommodate’ (citation form, and infinitive after auxiliaries) vs. op te nemen (infinitive elsewhere). This is not a minor point, since there is a relevant difference in the domain of application | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 364]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
of the two principles vs. the theme-vowel. The latter is relevant only in relation to stem-final segments, since only these are affected or protected by the presence of the theme-vowel after the stem. Not so for K.'s proposal, however. Thus, in the verb opvallen ‘to attract attention’/op te vallen the former serves as the citation form and as an infinitive after auxiliaries, and the latter as the infinitive elsewhere. Stem-initial v is affected by the progressive assimilation of fricatives mentioned earlier: op[f]allen. Thus we must know more about K.'s view on such cases in order to evaluate his ideas. This is a productive stemaffecting phonological process, which again does not spread analogically, see op te [v]allen, and so on. These four questions relate fairly directly to the case at hand, that of d-weakening. There are also two more general ones, which may be worded as follows. First notice that one weakness of K.'s system has gone unobserved in our discussion of the three separate phenomena: it treats verb-forms as special in all three cases, but does not derive the oddities in their behaviour from a single source. This in itself would not be a vice if the different sources would be independently motivated. But, as we have seen, they are not, and thus his system fails in an important way in terms of our discussion. Secondly, as noted his discussion captures only the cases of the earlier work on this subject, where the theme-vowel was motivated as a trigger only. One will be very interested to see how Kooij would capture its umbrella-behaviour as well, in a general and independently motivated way. His treatment of the earlier phenomena does not offer clues to this, and until then there is also an empirical difference between the two accounts. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
References
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
[pagina 365]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
en France
/V V
/_____ - V
/ ə _____{
/ ə _____{
s
wegen
deuren
van die mensen
voor de jongen
e
an
e
an
e
an
ie
ie
e-
ie
ie
ie-rule
e does not take part in the optional alternation with plosive d by the vacillation-rule. This is because again some otherwise expected ambiguities do not hold:
ie appears to attach itself here to the verb-form, preceded by a #-boundary judging from the final devoicing of the gaf-tie column (notice that this cannot be regressive assimilation, since -tie cannot be underlying; vide zoudie where voice cannot be obtained independently). This dialect does not allow ambiguities in cases such as the following:
...Its output is an immediate input to the t-zero rule, which gives ter and ten, and explains the absence of te de...cases by neutralization. Thus, these idiomatic expressions are slightly peculiar semantically (being idioms), syntactically (being case-marked), and morphologically (see the allomorphy), but their phonological regularity goes unquestioned.
...give t as a rule on the surface; this rule fails, however, for past tense formation of verbs.
ie → boodt-ie ‘did he offer’, and dood-
e(-
ie) → doodde(-die) ‘did he kill’. Again, the puzzle is in the verbforms.
ie
e-
ie
e-
ie
d d
en
en
en
en
en
en
en
en
-en
-de
-end (pres. part.)
-t
-de
-t
-de
-d (past part.)
-en
-den
-en
-den
-en
-den
-eur type, native speakers will use these to reconstruct a common source aarde underlying the set ik aard, wij aarden,
aarde, aardes, and so on. Notions such as ‘dominant’ and ‘derived’ do not really come into play here.