De macht van het woord

Een selectie uit het taalkundig werk van D.M. Bakker

[p. 109]

On ‘Ellipsis’ in Dutch Nominal Phrases^*

In the 1st century A.D., the famous Roman rhetorician Quintilian defined a certain figure of speech as follows: ‘There is a third figure which one can form by deletion, namely, in which several sentences are connected to one and the same verb, whereas this verb would be required in each of these sentences, if they were standing on their own.’ One of his examples is the following sentence: Vicit pudorem libido, timorem audacia, rationem amentia ‘lust conquered shame, boldness fear, insanity reason’.^1.

In this definition, two currents of thought on the topic of the present paper have come together. On the one hand, Quintilian subsumes this sentence under the deletional figures of speech, which implies that he considers it a kind of ellipsis. On the other hand, he reminds us that in this sentence, three subjects and three objects are connected to one and the same verb.

These two view-points, which we may call the ellipsis thesis and the multiple combination thesis, respectively, are mutually exclusive, as has already been seen by Quintilian himself. If, to use Quintilian's expression, one verb is related to several sentences, nothing is lacking; if, on the contrary, all sentences except for one are elliptical, it is impossible for the verb of the complete sentence to combine with the elliptical ones, because the notion of syntactic ellipsis involves the lacking of an element with which other elements might combine. Now if such an element is lacking, no mention has to be made of elements which are present.

In spite of this evident self-contradiction, one may have the feeling that Quintilian was right in describing his figure of speech in such a paradoxical way. The fact is that these two view-points are indeed correlated with two aspects of the sentences we have been considering. On one level, verbs, or rather occurrences of one and the same verb, are lacking; on another, there is a relation of some kind between the verb and the seemingly elliptical clauses.

The scientific conscience, however, will not be content with a self-contradictory definition, and one can hardly be surprised that there have been many grammarians who have cut the Gordian knot by accepting either the multiple-combination or the ellipsis-thesis. This is exemplified by the following 17th century definitions: ‘Somtyds wordt een woord gevoeugt op

* The present article is a slightly revised version of a paper read at the Societas Linguistica Europaea Annual Meeting, Prague, 8th-11th October 1970. I am indebted to Mr. Orrin W. Robinson, Leiden, for seeing through the English text.

1. ‘Tertia est per detractionem figura, quae dicitur έπ∊ς∊υγμένον in qua unum ad verbum plures sententiae referuntur, quarum unaquaeque desideraret illud, si sola poneretur’ (Quintilian, Institutio Oratoria IX: iii, 62-64).

[p. 110]

verscheidene dingen/dat ik syn geld, of hy myn herte hadde’ ‘sometimes a verb is combined with several substantives: if I only had his money, or he my heart’;^2. ‘Zeugma defectus est verbi, seu dictionis in aliqua oratione expressi, quum in aliis desideratur’ ‘zeugma is the lacking of a word which is present in some sentence and needed in others’.^3.

Moreover, it is quite understandable that in our time, linguistic theory being stricter and more explicit than in former days, the Gordian knot cutting type of definition is embraced more than ever. Fig. 1 shows a lemma of Tesnière's,^4. which exemplifies the multiple-combination thesis; several transformational generative approaches are expressions of the ellipsis thesis.^5.

Fig. 1

In fact, however, there is no Gordian knot to cut. If only we take a level distinction into account, we can avoid both self-contradiction and the one-sidedness which is inherent in Gordian knot cutting. This I will try to show with a Dutch sentence:

(1)	rode bloemen wijzen op een hartstochtelijk temperament en witte ‘red flowers indicate a passionate temper and white op onschuld innocence’

We can state the following facts about this sentence.

(1) Whoever reads witte, knows of course that he must think of white ‘flowers’. To be able to do so, he must have ascertained a relation of some kind between witte and the preceding bloemen.

(2) This relation is one between words, or, more precisely, between occurrences of words. Now one is easily inclined to identify all sorts of relations between words and to call them all syntactic relations. But there is no reason to assume that all relations between words are of the same type.

(3) Suppose, however, that the relation between witte and bloemen is

2. Petrus Leupenius, Aanmerkingen op de Nederduitsche Taale [Remarks on the Dutch Language], ed. by W.J.H. Caron (Groningen, 1958), p. 58.

3. Fr. Sanctius, Verae brevesque grammatices Latinae Institutiones (Genevae, 1764) (iuxta exemplar Salmanticae excusum anno 1595), p. 99.

4. L. Tesnière, Éléments de syntaxe structurale (Paris, 1959), p. 350.

5. See, for instance, N. Chomsky, and George A. Miller, ‘Introduction to the Formal Analysis of Natural Languages’, in: Luce, Bush, Galanter (eds.), Handbook of Mathematical Psychology II (New York-London, 1963), 298 (cp. S.C. Dik, Coordination. Its Implications for the Theory of General Linguistics [Amsterdam, 1968], p. 77); Rod. A. Jacobs and Peter S. Rosenbaum, English Transformational Grammar (Waltham, Mass.-Toronto-London, 1968), 204ff.; for an earlier treatment see Z.S. Harris, ‘Co-occurrence and Transformation in Linguistic Structure’, Lg. 33 (1957), in: Fodor and Katz (eds.), The Structure of Language; Readings in the Philosophy of Language (Englewood Cliffs, N.J., 1964, 1965), p. 185. See also J.D. McCawley, ‘The Role of Semantics in a Grammar’, in: Bach and Harms (eds.), Universals in Linguistic Theory (London, etc., 1968), 140ff.

[p. 111]

indeed a syntactic relation. Then there remains a difference between the combination of the adjective witte with bloemen and the combination of another adjective, rode, with the same word. The word rode is immediately juxtaposed to bloemen and precedes it, the adjective witte is separated from bloemen by a couple of words and comes afterwards. Thus, if both relations are syntactic, they are formally different relations. And consequently, one must accept two types of adjective-substantive combination, which is not a very economical solution.

(4) The two adjectives, witte and rode, are combined with one and the same substantive. This seems only possible if both adjectives are co-ordinated. Now our example is a specimen of co-ordination, it is true; but it is a co-ordination of clauses, not of adjectives.

All these facts point to the hypothesis, that witte and bloemen are not syntactically related. Nevertheless, that there is some relation between the two words seems equally true. Therefore, we must answer two questions: (1) How can we confirm our hypothesis that the mentioned relation is not a syntactic one? and (2) If it turns out that this hypothesis is true, what, then, is the nature of the relationship between witte and bloemen?

To answer the first question, let us consider the (isolated) sentences (2) and (3):

(2)	Those are red roses
(3)	I bought red roses

It is clear that the occurrence of the word red in (2) is syntactically related only to that of the word roses in the same sentence, just as the occurrence of red in (3) is related only to the occurrence of roses in (3). Strictly speaking, there only exist syntactic relations between occurrences of words; so when we speak of syntactic relations between words, we mean in fact ‘generalizations of relations between word occurrences’.

Now there is some reason for such a generalization, in so far as both occurrences of red roses denote the same concept, namely, an undetermined number of red-coloured flowers of a certain kind. Each individual occurrence, however, refers either once to the same bunch, so that in these two sentences mention is made twice of one and the same bunch, or the bunch referred to in (2) is another bunch than that referred to in (3). To refer to one bunch once, one needs only one occurrence of red roses; to refer once to one bunch and once to another, one needs two occurrences of red roses minimally. And if one wants to refer twice to the same bunch, one equally needs two occurrences minimally.

Let us return to (1). Here two different sets of flowers are meant: one of red flowers, one of white flowers. It is impossible for the single occurrence of bloemen to refer to two different sets of flowers; this follows from our consideration of (2) and (3). We can assume therefore that the occurrence of bloemen has been used up, so to speak, by helping to denote the set of red flowers. It cannot serve again to the denotation of the set

[p. 112]

of white flowers. So we may conclude that the element witte is not syntactically related to the element bloemen.

But now we must answer the second question: we know what this relation is not, but not what it is. In the preceding, no mention has been made of yet a third current of thought on this topic. It comes down to the assumption that in relation to witte, the concept of a set of flowers has been understood tacitly. This means that the relation of witte to bloemen is not syntactic but contextual. We can understand that witte means ‘white flowers’ only because bloemen has been realized in the context. This was, roughly speaking, the view of Charles Bally,^6. but the same idea arose as long as more than twenty centuries ago.^7. I think this view is correct, but I would like to stress that, the element witte being realized within a context which contains an occurrence of bloemen, one can only think of white flowers if it is possible to form a syntactic combination witte bloemen. So we have a contextual relation which is founded upon the possibility of a syntactic one, and therefore I will call this relation a syntactic-contextual one.

There is a further consideration. In Dutch, the following is a perfectly acceptable sentence:

(4)	Ik had twee rode bloemen, maar hij één witte ‘I had two red flowers, but he one white’

The element bloem of the potential combination één witte bloem is not present as such in the text, which contains the plural bloemen. The relation between bloemen and één witte, being a syntactic-contextual one, supposes that the reader or listener abstracts, so to speak, from the plural feature of bloemen and takes this word into account only as a denotation of the concept which is common to both the singular bloem and the plural bloemen.

Apart from the syntactic-contextual, there exist also syntactic-situational relations. In the market, at a cheese stand, I can ask

(5)	één pond oude ‘one pound (of) old’

the situation providing no word, no concept, but only the stuff called ‘cheese’.

Furthermore, in Dutch (as in German) even a compound member can combine with an adjective in the way just described, as in

(6)	Uw openbare en gezinsleven ‘Your public and family life’

The element -leven can only be a compound member, for otherwise it could not be preceded immediately by gezins-. Therefore, the relation between -leven and the adjective openbare can only be a contextual one. If we compare (4)-(6) with (1), we can perhaps even say that the occurrence in

6. Ch. Bally, Linguistique générale et linguistique française (Bern, ²1944), p. 159 (cp. p. 86).

7. For instance Caecilius of Calacte, see E. Ofenloch, Caecilius Calactensis (Lipsiae, 1907), p. 61 (79).

[p. 113]

(1) of bloemen is used in relation to witte, not as a plural, not as a substantive, not even as a word, but only as a consituational datum functioning as an approximation of the lingual concept which is present in the ‘underlying’ potential combination witte bloemen, and that this approximation, in the case of (1), is exact only ‘by chance’.

Returning once more to Quintilian's definition, we may state that there is a multiple relationship in such cases, but that two elements are related to one third element in entirely different ways. If we compare the potential underlying combination witte bloemen with the element witte in (1), we can express the formal difference between the two by means of the term ellipsis or by a deletion transformation. But neither the multiple combination thesis, nor the ellipsis thesis, are capable of describing sufficiently what is going on. They are only capable of describing part of the informational system at work.

While one has a certain motivation, as we have seen, for applying a deletion transformation in the description of sentences like (1), there are also applications of this descriptive technique which are senseless or at least redundant. Consider

(7)	Grote en heilige waarheden ‘Great and holy truths’

This type has sometimes been described by means of deletion transformations applied to conjunctions of sentences like truths are great and truths are holy. Even if we make use of referential indexes so that (7) is derived from

(8)	waarhedenx1 zijn groot en waarhedenx1 zijn heilig

the solution is not correct since it overlooks the difference between mentioning one and the same set twice, as in (8), or once, as in (7). Sentence (7) denotes just one set of truths once, this set being one of which each element is characterized by two qualities, so that each such element can be called

(9)	een grote en heilige waarheid ‘a great and holy truth’

Sometimes, linguists have been led to accept another solution for comparable instances, namely, by introducing a ‘rule schema’ which formalizes the possibility of combining a substantive not only with a single but also with two or more co-ordinated adjectives. Now the description by means of a deletion transformation is correct in so far as it accounts for a relation of each separate adjective with the substantive. In terms of the above-mentioned rule schema, however, one can only express the relation of the co-ordination of adjectives as a whole to the substantive.^8.

I think we need neither the deletion transformation nor the rule schema

8. See Dik, Coordination, pp. 92-115, 185-90.

[p. 114]

in the description of (7). In my opinion, the dilemma results from the use of a merely serial grammar. If the adjectives are related to each other (by co-ordination), and if simultaneously they are each separately connected to the substantive, this can only be expressed by means of a triangle as shown in Fig. 2.

Fig. 2

Now this triangle can only be mapped on to a line by reducing at least one line to a point. So in the deletion approach the lines A₁S and A₂S are noted, subsequently considered as two points, and connected by means of a co-ordination. In the rule-schema approach, the line A₁A₂ is noted, by means of the rule schema, subsequently considered as a point and connected with the remaining point, S. If we accept the possibility of two-dimensional structures, however, the choice between the two approaches becomes urgent only if we insist on mapping this two-dimensional structure on to a one-dimensional series of symbols, so that the latter suggests triangularity. But there is no theoretical consideration which forces us to do so. It is, indeed, necessary to map the triangle on to a line, that representing ‘surface structure’, which is always linear, as regards word sequence. But this line need not suggest triangularity in itself, because surface structure does not do so either. Thus, we get Fig. 3.

Fig. 3

But now consider

[p. 115]

(10)	Franse en Amerikaanse afgevaardigden ‘French and American deputies’

This group is equivalent to (7) serially, but not semantically. In (10), two sets of deputies are denoted, and as the occurrence of afgevaardigden can help to denote only one of them, at least one of the adjectives must be connected with afgevaardigden not by means of a syntactic relation, but by means of a syntactic-contextual one. If afgevaardigden denotes the set of all deputies, either French or American, both adjectives are syntactic-contextually related to the substantive. If the substantive helps to denote only the set of American deputies, the word Franse is in syntactic-contextual relation to afgevaardigden. If, on the contrary, the latter word helps to denote the set of French deputies only, the adjective Amerikaanse is in syntactic-contextual relation to the substantive. The latter possibility is only realized if en Amerikaanse is an intercalation. All three possibilities can be suggested, but not exactly expressed, by intonation. Furthermore, whereas in (7) the co-ordinator en can be deleted without destroying the cognitive content expressed by this group, the same deletion would change the meaning of (10) fundamentally.^9.

9. In my dissertation, Samentrekking in Nederlandse syntactische groepen (Leiden, 1968), I called this two-dimensional structure enkaptic (see e.g. pp. 83, 105). In a forthcoming article, I shall try to show the fruitfulness of two-dimensional representation of enkaptic wholes like John and Mary bought an ice-cream and a whisky-soda, and comparable sentences.