On the problem of topicalization in Dutch
Reinier Salverda

1. Introduction

In this paper I want to present a number of results and some of the problems of my study of topicalization in Dutch.1.)

The central question with which I will be concerned here is: How to give an adequate account of topicalization phenomena in Dutch declarative main clauses? Preliminarily, I take topicalization to be what is called inversion in traditional grammar.2.) That is, the investigation is concerned with declarative sentences in which not the subject, but some other constituent occupies the initial position3.), as e.g. in the following examples4.)

The literature suggests, that somehow the inversion construction is marked in contrast to the more neutral construction with the subject in initial position.5.) As another preliminary, I assume that judgements of adequacy involve criteria for observational, descriptive and explanatory adequacy.6.) That is, the account should be factually correct, it should provide a systematic description of what counts as grammatical and what not, and it should define a set of principles in linguistic theory with which to explain why (topicalized) sentences are constructed the way they are.

The third preliminary concerns the theoretical framework. My study of topicalization is conducted within the general framework of Chomsky's Extended Standard Theory (EST).7.) However, there is a certain tension between the original leading question formulated above and this theoretical framework. One of the central problems of the investigation,

consequently, is: How adequate and succesfull is EST as a theoretical framework for the account we want to give of topicalization in Dutch? In what follows, I will first sketch the case of topicalization in Dutch, presenting the facts and a proposal for their description. Then, I will go on to make some remarks on a number of more general, underlying problems I have come across while studying topicalization. Essentially, these are problems of explanatory adequacy, having to do with the conflict between a formalist and a functionalist perspective on linguistic theory.8.) I will argue for a position in which both are integrated.

All in all, then, my study of topicalization in Dutch has a double aspect throughout: it is an investigation in Dutch syntax as well as an exploration in theoretical linguistics, it attempts to account for a specific phenomenon in Dutch sentences as well as to clarify the general conditions and problems such an account is subject to.

2.1. Aspects of Topicalization in Dutch

For an analysis of topicalization in Dutch I think it is useful to distinguish the following aspects, all of which can be involved in our judgements on the grammaticality of topicalized sentences. The first aspect is that of word order and constituent structure. In Dutch declarative main clauses, the most frequent word order pattern is that in which the subject occupies the initial position before the finite verb.9.) Topicalized sentences are in clear contrast to this usual pattern, since they have another constituent than the subject in initial position. It appears, that in principle any constituent of the sentence may occupy this possition.10.) Thus we find sentences with initial Direct Object-NP as in (1), or with initial Adverbial Phrase as in (2), but also sentences with PP-Object (3), Dependent Clause (4), and Non-finite elements from the verbal predicate like participles, infinitives, adjectives ((5) - (7)) in initial position.

Thus, in general11.), we note that topicalized sentences show a considerable syntactic variety in initial position. The different types don't have the same frequency. I found12.), that approximately 60% of Dutch declarative main clauses begins with an NP-subject, roughly 30% has an initial adverbial phrase, and the remaining 10% is about equally divided between NP- and PF-objects on the one hand and elements from the verbal predicate like in (5) - (7).

The second aspect involved is that of intonation and sentence stress. In Dutch declarative main clauses the most frequent intonation pattern is the so-called ‘Hat-pattern’.13.) Usually, the main stress in this pattern is located near the end of the sentence. This stress we call ‘sentence stress’. An example of a sentence with this usual intonation pattern is (8):

(8)	Ik heb het gezien
	I have it seen
	Saya dudah ini melihat

However, the hat-pattern has many different possible realizations. Thus we find topicalized sentences in which the sentence stress takes its usual final position, but also cases in which the initial element carries sentence stress. In fact, a sentence like e.g. (3) may have various patterns, as can be seen in (9), (10) and (11):

(9)	Aan MARIE gaf hij een brief
(10)	Aan Marie gaf hij een BRIEF
(11)	Aan MARIE gaf hij een BRIEF

In general, then, elements in initial position can have varying types of stress. Sometimes, however, topicalization is only possible with special stress on the initial element, as e.g. in sentence (7).14.) The third relevant aspect is that of the interpretation of the sentence, Of special interest here is the relation between the initial element and the rest of the sentence. We can distinguish at least four different possibilities. The initial element may be interpreted as something about which the rest of the sentence says something as in (12), or it may say something itself about the rest of the sentence as in (13). Furthermore, the initial element can provide a kind of frame for the interpretation of the rest of the sentence as in (14), or it may qualify that interpretation as in (15):

Thus, here as before, we note that there is a variety of possibilities, not just one, single effect, associated with the initial position in Dutch declarative main clauses.

The fourth aspect we distinguish is that of the relation between a sentence and its context. The initial position provides a point of contact between sentence and preceding discourse. Elements in this position can entertain various relations with the context. Thus, for example, the initial element can be linked directly with something that has already been mentioned before: in (16) what is already known (‘Dat’) is put in first position, and the rest of the sentence contains new information:

But the element in initial position may also emphatically present a contrast with something that has been said before, as in (17), or alternatively, it may just simply connect the sentence with what precedes, as in (18), where the initial Daarom is neither known from nor in contrast with the preceding context.

These four aspects, together or on their own, may affect our judgements of grammaticality on topicalized sentences. In general, we may say that there are no 1:1-correspondences between the four aspects. For every aspect, we noted a variety of possibilities, which apparently can be combined freely. There are, however, some interesting cases of interaction and interdependence.

To begin with, we have already noted that sometimes topicalization is only possible if sentence stress comes along to the beginning of the sentences, as in (7). Likewise, the answer in (17) would be odd without initial sentence stress.

Furthermore, there are cases in which the context practically determines that a nonsubject has to be put in initial position: in (16) the alternative answer with neutral order Ik geloof dat wel is decidedly odd. Thirdly, we note that quite often topicalized sentences contain a negative element as in (4), (5) and (18), or a positive one like wel in (16), without which these sentences are strange. Perhaps this is an indication

that the topicalization construction serves to single out the initial element as the one that is being asserted or denied.

It is interesting to look more deeply into these cases, since they may tell us more about the specific properties of the topicalization construction. In this paper, however, I can only point to these cases of interaction.

So far, then, we have noted a number of characteristic aspects of topicalization in Dutch. Topicalized sentences may contrast with their non-inverted counterparts in one or more of the following respects: word order, stress pattern, interpretation and context relations. Quite often, these aspects can combine freely, but sometimes they interact and coincide. Distributionally, the construction is restricted to declarative main clauses and excluded in dependent clauses, cf. (19)15.)

All these properties are relevant if an observationally adequate account of the topicalization construction is to be given. That is, to be observationally adequate, any theory has to account for the data pertaining to the aspects mentioned above and their interrelations.

2.2. A proposal for description

Having presented the characteristic aspects of topicalization, the next question is: How do we describe these facts? Which concepts and rules do we need in our theory in order to account for the properties of this construction?

Here I want to discuss a set of concepts which I think are useful and necessary for an adequate description of topicalization.

Within the theory of grammar we have to distinguish a number of autonomous subsystems for the various aspects outlined in section 2.1. Each of these subsystems provides an account for its particular linguistic level. In addition to these partial accounts we need an overall integration component, which assembles them into one, coherent description of topicalization.16.)

In this section I want to discuss the relevant subsystems and their contribution to the description.

The first component we need is that of Formal Syntactic Structure. This component must account for the specific ordering of constituents in topicalized sentences and for the formal properties of topicalization. Intuitively, it is appealing to treat topicalization as a reordering process. Thus, first, we need rules that specify a base structure which accounts for the syntactic similarity between a topicalized sentence and its neutral counterpart. Then we need a rule of Fronting. Fronting is a general rule which transports constituents with varying syntactic functions to the open position before the finite verb.17.) This rule has the properties of a Root Transformation18.), since it can only prepose constituents to the initial position of main clauses. It cannot work in dependent clauses. However, in exceptional circumstances, the Fronting rule shares with Wh-movement

the property of fronting constituents from inside dependent clauses to the initial position of the main clause,19.) as in (20) and (21):

(20)	Dat gelezen had
	That book thought John that he - read had
	Itu buku merasa Yani bahwa dia - membaca sudah
(21)	Welk gelezen had?
	Which book thought John that he - read had
	Mana buku merasa Yani bahwa dia - membaca sudah

With this Fronting rule, we now have one general preposing operation in Formal Syntax which can put any type of constituent in initial position. Differences between the various types of preposing will, as a consequence, have to be accounted for in other subsystems of the grammar.

The second component is that of Phonological Structure. In this component we need rules that assign the various types of stress in the so-called ‘Hat-pattern’. The realizations of this hat-pattern may vary, but there is always only one main stress in the sentence. The rule which assigns this sentence stress we call Focusing, since it signals the perceptively most salient part of the sentence.20.) Usually, sentence stress is located near the end of the declarative, as in (8), (10) and (11). This we call the unmarked Focus position. Cases like (9) and (13), (17) have a marked Focus position, and consequently a contrastive effect.

With this analysis, we can make a distinction between the Focus-part and the Non-Focus-part of a sentence. It appears, that the second can often be left out21.), but not the first: Focus is an elementary constitutive factor of what we are used to call sentences.

In addition to the Focussing rule, we need rules for the assignment of other types of stress, like e.g. the initial one in (11). Here I will not go further into this matter.

The third component is that of Thematic Interpretation,22.) specifically the level of the presentation structure for the information in the sentence. Elsewhere,23.) I have sketched the terminological and conceptual chaos in this field. Still, I think some useful and interesting conceptualization has been done in the work of the Prague School24.) and in Functional Grammar25.). In this component I think we need the following concepts and rules in order to capture the various possible relations between the initial element and what follows in the sentence. First we need a rule that specifies the Topic, i.e. the item about which the sentence is a statement. This accounts for sentences like (12) in which the initial element is the Topic. Then we need a rule that specifies the Frame with respect to which the statement in the sentence is relevant. This accounts for sentences like (14) and (17), that have a Frame as initial element.

An interesting aspect of the concepts Topic and Frame is, that they may, but don't have to be present within the sentence. They may also be ‘in the air’, not put into words, but understood. Topic and Frame

therefore also don't have to be related to initial position, allthough quite often they are. With the introduction of these concepts, we leave the limits of sentence grammar behind: Topic and Frame are not formal entities and they are not necessarily part of a sentence. With these concepts we can now also account for the similarity between topicalized sentences and cases of Left Dislocation. Thus, in (22) we find an independent Topic, followed by a complete sentence; example (22) has the same thematic interpretation, but not the same syntactic structure as (12). The same holds for the Frame in topicalized (14) and left-dislocated (23):26.)

In this third component we also need a rule that specifies the counterpart of the Topic, viz. the Comment, i.e. that which is said with respect to the Topic, as e.g. in (12) heb ik gelezen, said with respect to Dat boek. Normally, the comment follows the topic in the presentation of the information in the sentence. The comment can, however, also be put in initial position. When it is, the sentence is marked, as e.g. in (13) and (17).

The important thing about the Comment is, that it is obligatory, in contrast to the Topic. The Comment appears to be the core of the declarative sentence, which is essentially a predicative entity.27.) The fourth component we need is that of Discourse Structure, which accounts for the relations between sentences and their context. From the discussion above, it is clear that Topic and Frame have a role to play in this component too, since - as in (22) and (23) - Topic and Frame may be located in the context that (immediately) precedes a sentence. In addition, we need the concepts Given and Non-Given - respectively to determine whether a sentence or elements in it are already known from the context or not. Note, that Topic and Given do not coincide. In (16) Dat is Given, but not Topic: it only refers to the Topic in the preceding sentence. Also, the Topic in (12) Dat boek does not have to be Given at all. Furthermore, Non-Given does not coincide with Comment. Thus, in (9) we see a focussed, contrastive Topic, followed by the Comment gaf hij een brief, which in this case is clearly Given. Psychologically, the order Given / Non-Given would appear to be unmarked, clearer and easier to understand than the order Non-Given / Given.28.) Initial elements in topicalized sentences may be either, or something else yet: Connecting Element like daarom in (18).

Finally, we need an overall, integration component, which takes care of the interrelations and combinations of the rules from the various subsystems just discusses. This component tabulates the results of the various autonomous components of the theory of grammar into one complete and coherent analysis of the construction under investigation.

For example, it may combine the rules of Fronting, Focussing, Topic-Comment-Assignment and Given-New-Specification in such a way, that the resulting sentence has a non-subject initial element that is at once Topic and Focus and Non-given, followed by a Comment which is Given, as in (9). Or alternatively, these rules may combine to produce a sentence in which the fronted constituent may be Given and have contrastive stress, whereas the Focus of the sentence is located near the end of the Comment as in (24)

(24)	Dat boek gaf hij aan MARIE
	That book gave he to Mary
	Itu buku memberi dia kepada Maria

Still other possibilities exist. It is therefore unjustified to take topicalization to be only a focussing operation29.) or only a topic preposing rule.30.) In fact, there is quite a variety here, since the fronting of a non-subject in declarative main clauses can be associated with Topic, Focus, Frame, Comment, Given, Non-given, or Connecting Elements.

The various concepts and rules discussed so far can be seen as an explication of the opposition ‘neutral’/‘marked’. They provide a set of dimensions along which varying degrees of markedness can be determined for declarative sentences. In this central integration component of the theory of grammar we can now construct a matrix for all the factors involved and their combinations. Thus we can break down the initial, traditional judgement concerning topicalization into a scale for all the possibilities it contains.

In this paper I cannot discuss in any detail the different components, concepts, rules and interrelations between them. I have limited myself to the discussion of a number of clearer cases. Still, from the analysis presented above, a picture of considerable complexity arises. Tentatively, we may now describe topicalization as a communication-oriented restructuring process in declarative sentences - a process in which various subprocesses combine to form sentences that facilitate the presentation of information, since their structure is adapted to the requirements of clarity, interest, relevance and appropriateness.31.) Together, the notions discussed so far - Fronting, Focus, Topic, Frame, Comment, Given, Non-Given, Connection - are sufficient and necessary to account for the various aspects and properties of the topicalization construction in Dutch. The analysis thus defines a condition of descriptive adequacy. That is, to be descriptively adequate, any theory has to be able to express the generalizations presented above in a coherent and explicit analysis of topicalization.

2.3. An explanatory perspective

Having come this far, we are still left with the question: Why? That is, we now face the problem of theory and explanation in syntax: What are the principles and processes by which sentences are constructed?32.) Why are topicalized sentences constructed as they are? Why do

they have all these properties? Why all this variety? Do we have an explanatory perspective on these questions? Specifically, do we have in our theory of grammar a set of principles with which we can explain the properties and the complexity of the construction under investigation?

With respect to these questions, I would like to present, first, a critical evaluation of two of the most serious linguistic theories at present, Chomsky's EST33.) and Dik's Functional Grammar (FG).34.) In my opinion, both are inadequate, so I will try and argue for a more integrated perspective.

I find Chomsky's EST inadequate for the following reasons. First, in his theory we note a very one-sided predominance of formal syntax as the only explanatory perspective over the other components of the theory. Explanatory principles, in order to be acceptable, have to be stated in terms of formal constituency. This formalist program leads to a systematic neglect of non-formal aspects, which - as we have seen for topicalization - can be very relevant to the construction we investigate. Or, if they are not neglected, they have to be reduced to the formal principles of constituent structure. This leads to misrepresentation, e.g. when Topic and Comment are taken to be constituents of syntactic deep structure35.), whereas they really have much more to do with the presentation of information. Secondly, we note in EST a strong preoccupation with formal theory construction as the only worthwile activity in scientific linguistics. In fact, everything becomes theorydependent, and no room is left for independent observation and analysis. As a consequence, EST can no longer be refuted. EST is committed to treating topicalization as a formal rule in autonomous syntax, with no relation to matters of Focus and Topic. If there is such a relation - as we noted for (7), (12), (16) and (17) -, EST has no way to account for this. All in all, then, we note two points of inadequacy, which follow directly from central tenets in EST. We conclude, that EST fails to take into account the full range of aspects associated with topicalization and that it also fails to develop a theory that can deal with the complexity of the phenomenon under investigation.

Dik's FG I find inadequate on quite different grounds. First, in this theory, the explanatory principles have to be functional. With respect to topicalization, FG distinguishes two functional possibilities: the element in initial position has either the pragmatic function of Topic or that of Focus.36.) And it is these pragmatic functions that determine how a declarative sentence is to begin.

However, the concepts used in FG to analyse the pragmatic level appear to be insufficient. FG identifies Topic with Given, and Focus with Comment, and thus cannot express some necessary distinctions and generalizations. like e.g. the ones we made above concerning sentences (9), (12) and (16). Also, for other types of topicalization than + Topic and + Focus, like e.g. those in (14) and (15), there is no room in FG.

Secondly, the overall functionalist program of FG leads to a neglect of the particular formal-syntactic properties of the topicalization construction, cases like (20) or Participle-fronting as in (5), which, allthough +Focus, is often excluded for structural reasons.37.)

Thus, both EST and FG fail to meet the criteria for adequacy. Both theories can only claim partial validity. Both also only treat com-

plementary aspects of topicalization. And they both make conflicting claims with respect to the goals and the leading perspective of linguistic theory. Consequently, they both only show half of the picture, EST the formal half, FG the functional half, leaving no room for the insight, that both formal and functional factors are crucially involved, in complex interaction, in the topicalization construction.

What is needed now, is a liberation from the limitations of this formalist/functionalist-dilemma. In this respect, I think it is interesting to try and relate linguistic structure to the goals and strategies of lingual acts.

In the analysis of topicalization, given in section 2.2, there are three points of contact for such an act-perspective. First, at the level of Focus, we can relate the topicalization construction to such lingual acts as ‘Calling attention to something important’ and ‘Highlight what is interesting’. Secondly, at the level of Topic, Frame and Comment, we can discern a general presentation strategy behind the topicalization construction. This strategy determines what is going to be said in general and what has to be asserted in particular. Also, it determines the selection, combination and sequencing of the parts of the information in the statement. Thirdly, at the level of Discourse Structure, we can relate topicalization to a strategy that aims to make it fit into the specific context in which it has to play a role. Thus, at these three levels, we can relate linguistic structure to a variety of goals and purposes in the presentation and communication of information: interest, clarity, relevance and appropriateness. Here, too, we note that the relations are complex: there is no 1:1-correspondence between linguistic structure and lingual act. Nevertheless, I think it is reasonable to conclude, that topicalization is a communication oriented process and that it will be interesting to look deeper into the specific properties of this orientation.

Literature