Seminar Alternative Syntactic Theories - Summer Term 2002
Generative and Functional Grammar
Different Approaches to Word Order
2. Word Order in Generative Grammar
2.1. General Remarks
2.2. Principles and Parameters
2.2.1. The Lexicon and X-Bar Theory
2.2.2. Case Theory
2.2.3. Move Alpha
2.2.4. Bounding Theory
3. Word Order in Functional Grammar
3.1. Functional Relationships
3.2. Universals and Hierarchies
3.3. Constituent Ordering Principles
3.4. Placement Rules
3.5. The Multifunctional Approach
4. Final Remarks
That language is a very complex phenomenon is certainly a truism, but that consequently we should not be astonished about the fact that there are a variety of theories and approaches to the study of language is already less so. Each theory may chose its special angle to discover part of the mysteries of language, each approach has its own questions, assumptions and goals - to grasp language in its complexity in one fell swoop is probably impossible. However, when we try to evaluate different theories, we sometimes seem to forget that each linguistic work only covers fractions and we are tempted to compare apples with oranges. At least to the novice in linguistics it appears sometimes that quite unnecessary polemics are going on between different schools. A point in case may be the relationship between functional linguistic theories and the generative approach to language.
I have chosen the topic of word order to exemplify how linguistic features may be viewed at under very different methodological and theoretical provisos without necessarily falsify the other′s arguments. For example, one may like to analyse word order under the aspect of the processing difficulties one may see in the "unidimensionality" of the "temporal linearity of verbal communication".1 One may approach it from the side of language contact and explore the mechanisms by an "areal method".2 Accordingly, in the first case, one may come to the result that word order is a fundamental property of language or, in the latter case, a language specific arbitrary convention. If one considers word order as a fundamental property of language one may like to classify languages according to word order as has been done by Greenberg, Hawkins and many other typologists. Or, one may explore word order under the aspects of spoken or written language, or style and register and one will then probably come to differing results.3 One may stress pragmatic, psychologic or structural aspects, or one may like to approach the problem diachronically. At first glance the outcomes may differ considerably, still - they may, but not always need to be contradictory but just be the answers to different questions.
I have chosen Simon Dik′s Functional Grammar (FG) and the Chomskyan approach with major reference to the Theory of Principles and Parameters (PP) to explain two different approaches in some more detail. Dik′s project is: "How do speakers and addressees succeed in communicating with each other through the use of linguistic expressions?"4, while Chomsky asks: "What is the system of knowledge incorporated in the mind/brain of a person who speaks and understands a particular language?".5 Dik postulates: "The main function of a natural language is the establishment of communication between NLUs"6 whereas the Chomskyan paradigm takes the viewpoint: "Language is not a tool for communication (although it can be used as such) and communication does not rely on language (although it usually uses it)".7 Although the Chomskyan statement sounds very radical - can we really say it is untrue? Nor should we too easily jump at Dik′s definition, as his focus on the communicational aspect may be slightly too limited. Language may also function as a device for making sense of the world or for thinking etc. At least the discussion of what the functions of language are and which function is primary has kept philosophers busy for about 2500 years.
Perhaps a danger with functional approaches lies in a certain ′plausibility at first sight′ which may tempt one to abstain from closer inspection. I presume a majority, including linguists, is satisfied with the notion of communication as the main task of language and may then overlook that there are further phenomena which have to be taken into account. The sequence of tenses in reported speech may serve as an example: the functional explanation for an adjustment to the deictic centre seems to be logical and convincing - yet not all languages with a tense system do adjust and change the tense of the original utterance accordingly. How can this be explained functionally?8 Thus I think that not even functional approaches, although they often appear to be all-encompassing, cover all aspects and additions from other theories may be helpful to close the gaps.
From the different starting points of GG and FG it becomes only too obvious that the respective elaboration of the theories must necessarily lead in different directions and that different methodological procedures are required. The stress on the communicational aspect yields the view that syntax proper is subservient to semantics which is again subservient to pragmatics.9 This also entails that Dik mainly concentrates on structuring the informational content of linguistic expressions and that grammatical issues in the narrow sense like morphology or word order figure very low in his hierarchically organised model. Morphosyntax is considered to be rather language specific than general and is dealt with by traditional empirical means.
Not so the Chomskyan approach: for his project the autonomy of syntax is essential per definitionem, as it were. The structure of language which enables us to make "infinite use of finite means"10, as Humboldt put it, is the target of interest. This aim requires the study of an autonomous structure stripped off from all other ′intervening′ human behaviours. Yet, one must not misunderstand the postulation of an abstract and autonomous syntax as the negation of any other, for instance pragmatic influences on the surface form of language or of regarding them as of minor importance in general, even if they are called "peripheral"11. These features are "peripheral" in the sense that they are not the focus of enquiry of this approach but serve to provide "evidence concerning the internal system of the mind/brain [...] what is [...] to discover".12 Besides, research into aphasia and first language acquisition seem to yield evidence in favour of the independence of grammatical from communicative faculties in the human mind as there are cases where those faculties have been found to be impaired.13
In contrast to functional approaches, hypotheses postulated by generativists lack all first-sight plausibility. They operate on an abstract level and require some serious preoccupation before a thorough understanding can be obtained (which, however, then seems to open up a fascinating simple coherence).14 Furthermore, given its target, the Chomskyan approach necessarily faces a number of strong methodological problems as nobody can ′look into ones head′. It therefore has to apply a deductive-hypothetical method which has to be checked against empirical data and accordingly to be revised again and again until the model generates data which agree with the empirically found ones.15 Thus the project necessarily entails re-shaping of former assumptions which has repeatedly served as a point of criticism. Moreover, in the meantime a great number of linguists all over the world work in this field. They may follow slightly different approaches and work on special problems of particular languages. Thus the theory is not represented homogeneously and it is difficult to keep track with new developments.
2. Word Order in Generative Grammar (GG)
2.1. General Remarks
The notion of wellformedness has always been linked to word order as one crucial aspect for judging the correctness or non-correctness of a sentence. As such, word order has been ascribed a major role in providing us with information about our structured, unconscious linguistic knowledge. The importance given to word order may have had its roots in the fact that Noam Chomsky and the majority of the linguists working with GG have been anglophones and - as is widely recognised - English codes much information by the order of its constituents. Yet the view that ′GG is only valid for English′ is clearly exaggerated today. Many studies have been carried out for German, Italian, Korean, Japanese and many other languages since the early 1970s at the latest. Even Chomsky is said to have worked first with Hebrew, but turned to English as "it is not easy to do theoretically interesting syntactic work in a language one does not speak natively".16
Although the works by non-anglophone linguists brought about changes and refinements in some aspects, and also diminished to some extend the importance of word order in favour of constituency17, no linguist even outside the generative paradigm seems to have the opinion that there are languages with totally free word order, i.e. that word order could be fully replaced by other mechanisms of coding information. Typological research confirms a certain universal structural importance of word order as only a restricted number out of all theoretically possible variations has been found to apply in natural languages.18 On the other hand, the multifaceted attempts of generative grammarians to account for relatively free word orders may seem discouraging in the sense that it is difficult to make out a clear conception yet.19 Due to this general practical problem with GG and in view of the size of this essay, I will have to content myself with explaining only very basic notions of PP in respect to English word order and will not be able to offer any in-depths discussion of special problems. However, the principal difficulty of Standard Theory and Phrase Structure Rules concerning split, i.e. discontinuous constituents has long been overcome.
2.2. Principle and Parameters20
The Theory of Principles and Parameters as developed since the early 1980s is a modularised system of principles that are supposed to operate universally according to a certain language specific parameterisation. The parameterisation is supposed to be set unconsciously by the child while acquiring its first language. The structuring principles are thought to be innate or at least a certain device for them is supposed to belong to the genetic endowment of all human beings. Principles and parameters are triggered to develop thanks to linguistic input provided by the environment and thus account for the solution of the logical problem of language acquisition. However, the theoretical description of the exact shape of the principles and the setting of parameters is still a work in progress - and has in the meantime seen some further developments and re-evaluations under the Minimalist paradigm.21
The main modules of PP are formed by X-Bar theory, Theta theory, Case theory, Bounding theory, Binding theory, Control theory and Government theory. These are supposed to interact which each other and thus to constrain the generation as well as the structural interpretation of well-formed sentences. The principles of Theta theory and X-Bar theory project a so-called d-structure from the lexicon. S-structure is derived by the functioning of Case, Bounding, Binding, Control and Government theory.22 This finally leads to the interfaces to other domains of language use, i.e. the Phonological Form PF (articulation) and the Logical Form LF (semantic and pragmatic interpretation).
2.2.1. The Lexicon and X-Bar Theory
On the one hand, much information for the linguistic structure to be generated is coded in the lexicon. It contains indications as to the phonological matrix, the category (Verb, Noun etc.), the sub-categorisation frame (e.g. what complements does a transitive verb take), and a "grid" for thematic roles.23 An example of a lexical entry may be the following:
suggest: +V, (AGENT, theme) = category + thematic roles
NP = possible sub-categorisation
CP = possible sub-categorisation
Another structuring device is X-Bar theory. X-Bar theory is the grid that allows to project phrases, i.e. larger meaningful units, from lexical as well as functional head words. It thus forms a matrix for possible positions of words and makes their relationship among each other visible thanks to tree structures. The general principle is that a head word X forms a first projection together with its complement, i.e. its inner argument as selected in the lexicon. The next step is the connection of the first projection level X′ (hence the name) with a specifier to the maximal projection X′′ or XP - the fully fledged phrase. The specifier is seen as an external argument which is not linked to a syntactic frame and may thus not surface in certain cases. The asymmetric relation of specifier-head and complement-head is thus reflected in X′-Bar theory. X′ and X′′ may be enlarged by adjuncts which do not change the level of projection as they are not chosen by the lexicon but simply add further information, e.g. adjectives to nouns, adverbials to verbs or relative clauses.24 The generalised principle of X-Bar theory is the following:
Xn _ .... Xn-1 ... 25
X stands for any category eligible to function as a head word, i.e. verb, noun, adjective, preposition as contentive lexical heads, complementisers, auxiliaries and other tense/agreement features, or determiners and pronouns26 as functional heads. The dots indicate the possible position for any complement or specifier. Properties of the head remain visible at any level of projection (= projection principle). Complements and specifiers are maximal projections of their heads respectively.
The general tree structure parameterised for English may look like the following one.
The heads of each projection figure on the left and complements to the right. The latter are subcategorised for by the lexicon. Specifiers appear to the left side of the intermediate projection level X′. As specifiers have to be maximal projections of their head words, the example phrase marker is a simplified version to allow for a better overview. Actually each specifier should be extended by its constituents, (i.e. intermediate projection with its specifier, if applicable, and head plus possible complement). This schema could be extended by adjunctions to either the X′ or XP level. For instance the adjective "nice" could be adjoint to the N′-projection of "picture" without entering a new level of projection.
A German matrix would differ in the position of I° branching to the right instead and VP branching to the left respectively. Thus the difference in word order of different languages is supposed to be caused by functional categories. Lexical categories are found to be relatively uniform across languages and thus the hypothesis has been postulated that functional heads are the place for parametric variation.27
Functional heads become necessary if we want to account for agreement, tense or embedding, and possible "landing sites" for constituents, i.e. word order. The IP is the maximal projection of the inflectional head I or Infl and corresponds to the sentence level of a main clause. The complementiser phrase CP becomes indispensable with embedded subordinated clauses. However, the position of specifier CP or C° are also available as a landing site for question words, cleft arguments and inverted auxiliaries. In more recent works, the structure of IP has further been split into its components tense, subject agreement and object agreement (as required in some perfect constructions in French and Italian)28 or in gender as is required by the Russian past tense.
Assuming a principle of economy, the projections are made only as high as required by the lexical entities. So there is no necessity to project a CP branch for main clauses, or it is supposed that children, who at a certain early stage of language acquisition do not use inflections, project only as high as VP. This proposal may explain the lack of nominative case assignment to syntactic subjects like: "me talk" or "my close it"29. Agreement and nominative case is assigned by Infl, not by V. Once children use tenses and consequently have to project up to IP, nominative case is found to be allocated coherently to the subject according to Radford (1996)30 and their word order becomes more coherent with the adult one.
2.2.2. Case Theory
This just mentioned consideration of nominative case assignment leads to the impact of Case theory on the problem of word order. Case theory states that each phonetically realised NP must bear a structural case. That means that each NP must be put into a position where case can be assigned to it.31 Case theory can motivate movement as can be seen with passive constructions. The passivised verb (or similarly the ergative verb) necessarily assigns the role of theme, but not the one of agent by the rules of Theta theory. Usually the theme is put in complement/object position as an internal argument of the head word. However, case is said to be "absorbed" by those types of verbs.32 The direct projection from the lexicon would then look like the following (= d-structure): was formed the sentence.
However, the predication principle requires a subject, and a noun phrase requires a case assignment. Thus the sentence should move to a position were both is possible. On the other hand, the projection principle requires that phrases may not simply be cancelled from their original position, but that they be preserved at all levels. Furthermore the thematic grid of the verb requires a complement that cannot be left empty. Hence the moved phrase (or a moved head) will leave a coindexed trace which fulfils the requirements of structural consistency and the original relation remains visible at the s-structure level. In subject position case can be assigned by Infl and the predication principle is satisfied by this movement, thus yielding the s-structure: [the sentence]i was formed ti..
2.2.3. Move Alpha
This movement is called move alpha - an important component for the question of word order, although not specifically a module of PP theory. Move alpha is triggered and restricted by the principles of GB and thus cares for the positioning of heads or maximal projections within the sentence structure. It is the mechanism that yields the abundance of word orders found in natural languages. Not even English is as fixed in its word order as it is commonly asserted. Green for instance found 12 instances in English (some of which, however, are purely literary) where the order SVO seems to be apparently replaced by VSO.33 That means that even English needs plentiful use of move alpha, not only the languages of freer word orders.
Move alpha is the generalisation and simplification of the transformation rules as developed in Standard Theory, whose function it was to make the grammatical relations between derived phrase markers visible and to yield well-formed senctences.34 However, the transformation rules proved to be too strong and specific (and thus complicated) over time. As a result they were replaced by move alpha which still serves the purpose of showing the relationships between d-structure, s-structure and PF and LF thanks to the traces left or other empty categories.
2.2.4. Bounding Theory
One of the most important principles affecting move alpha is the subjacency condition within Bounding theory. It has been observed that in natural languages constituents may not move too far away from their initial (d-structure) position and Bounding theory tries to capture the ruling mechanisms for these restrictions. Movement seems to be only allowed within certain syntactic domains and at the most one so-called bounding barrier may be crossed. This accounts, for instance, for the fact that an English embedded clause may be moved away as long as it refers to the head of a phrasal projection, but not if it refers to a complement. Freidin gives the following example:35
[NP an author of [NP an article on string theory] CPi] [VP [VP just arrived] [CPi who I want to meet]]
This is a correct sentence, the "distance" from the CP trace to its actual position has just to surmount the barrier from VP to NP. Whereas the sentence:
*[NP an author of [NP an article on string theory CPi ]][VP[VP just arrived] [CPi which I want to read]]
is not well-formed because the relative clause would have to cross one more barrier from its referent article which is more deeply embedded.
Movements can be realised one after another leaving a chain of traces which each must correspond to the subjacency principle, i.e. that "members of each link of a certain chain"36 must no be farer apart than one bounding category. Bounding categories are subject to language specific parameterisation. It seems that in Italian an IP is no bounding category whereas it is in English. However, a CP has bounding effects in Italian. The general principle operating is that of subjacency, the parameters for Bounding categories are the field of further research for particular languages.
It goes without saying that the above given account of the treatment of word order in GG is much too superficial and not at all complete. Theta theory and Government theory have been briefly mentioned as to have their impact, but also Binding and Control theory should not be left out in a complete account as each theory has a more or less strong effect on the linear order of words. However, I hope to have shown that in GG
a) word ordering mechanisms are considered to be universally valid and therefore word order is seen as a fundamental property of linguistic organisation; and
b) word order is dealt with "together with all other aspects of structure, and is not considered as an independent issue."37
This view seems to stand in direct opposition to Simon Dik′s Functional Grammar.
3. Word Order in Dik′s Functional Grammar (FG)
The role word order plays in FG is substantially different from that in GG. Word order is principally regarded as a language specific phenomenon, and in contrast to the all-pervasiveness in GG, FG as "any model of grammar which is based on the premise that linear order is not an underlying property of sentences [ ...] must [...] possess a set or sets of linearization (word order) rules."38 Thus FG has a clearly defined component, the Placement Rules within the module Expression Rules, which is responsible for putting words into the right linear order after the definition of their functional relationships in the other modules of FG, i.e. the Fund, a sort of extended lexicon, and the main body of FG, the Underlying Clause Structures (UCS). The UCS form a very interesting structuring module for the semantic and pragmatic content of a sentence taking so-called predications of the Fund as its starting point. However, the UCS is only of little relevance to the present topic. So the most important part of FG will hardly be touched in the following in favour of the focus on word order.
3.1. Functional Relationships
Functional relationships which are partly defined in the Fund and on different levels of the UCS may, however, have their effect on word order. Dik distinguishes three types of functional relationships: semantic, pragmatic and syntactic functions. Semantic (agent, goal, recipient etc.) and pragmatic functions (topic, focus, etc.) are regarded to be universally operating in all natural languages. The syntactic functions subject and object, however, are re-interpreted by Dik as primary and secondary vantage points. The term "perpectivizing" functions, as also used by Dik, may therefore be more appropriate.39
Dik argues that a subject is only a relevant function if a language has the possibility of expressing synonymous passive constructions like: The child threw the ball vs. The ball was thrown by the child. However, languages in which such constructions are not possible - for instance, if an inanimate patient like ball cannot occupy a subject position and the passive has to be circumscribed by constructions like It was the ball that the child threw, the function of subject is not considered to be relevant, the semantic functions as defined in the Fund will then satisfy all structural and functional requirements. Thus, according to Dik, there are languages with no subject assignment - Hungarian is said to be a case in point.
Similarly, object functions are only necessary if such constructions like Mother gave a birthday present to her son vs. Mother gave her son a birthday present are possible. The two objects son and birthday present may alternate without any change of meaning (i.e. without any change of semantic function assignment) or any additional marking - son is considered to have been raised from indirect object to direct object. Accordingly also the "second vantage point"40 has alternatives and consequently is a relevant syntactic differentiating feature. Given that such constructions are not possible in German - the marking for the dative case remains visible no matter which order of the object constituents is chosen - German is not considered to be a language with object assignment.
Siewierska, however, draws attention to problems with the subject/object assignment. The passive/active constructions or raising from indirect object to direct object are not synonymous in all languages and thus do not "necessarily depict the same state of affairs in the FG sense."41 Moreover, some languages like Latin may show object raising only with tri-transitive verbs, but not with di-transitive ones.42
Following Dik, subject and object assignment are not language universals. Yet if subject and object assignment is found, patterned consequences for word order become apparent across languages, i.e. subject precedes object.43 On the one hand this is a logical corollary of Dik′s definition of what subject and object stand for, namely the primary and secondary vantage points respectively. As Siewierska rightly maintains it would hardly be accountable for the secondary perspective to appear before the first one.44 Languages which seem to counter-exemplify this order are found either to have no subject and/or object assignment in terms of FG interpretation, which might also be circular reasoning, or their pragmatic functions seem to be prevalent, which seems a bit arbitrary.
Subject and object are means of giving additional information in a grammatically coded way and are viewed as being complementary to semantic functions. As they seem to have a greater power for marking, as can be seen by the passive constructions, we find subject and object in Dik′s general pattern for word order, and semantic functions coded as argument 1, argument 2 etc. in rather language specific formalisations.
3.2. Universals and Hierarchies
Dik draws from a number of cross-linguistic, empirical data from typological research to postulate his Placement Rules. He thus follows an inductive method. The collection of data allows for a very appealing overview of the obviously operating mechanisms responsible for the sequencing of words in natural languages.
In the wake of Greenberg′s pioneering typological works of the early 60s, Dik distinguishes different types of universals according to the parameters absolute/statistical and unconditional/implicational. Absolute unconditional universals are rather rare and of little practical use. These are statements like "all languages distinguish between vowels and consonants"45. Dik, however, maintains that "in the actual practice of typological research, many of the most interesting hypotheses take the form of statistical implicational universals, for which we will also use the term "tendency".46 The subject/object distinction as interpreted by Dik may count as an example.47
Many implicational universals are found to form a chain, like if A then B, if B then C etc. whereby C is the most general of the properties. Such chains may be regarded as forming hierarchies of occurrence (i.e. all languages will have property C, some also will have property B and still fewer of them will additionally have property A) and frequency (in a language characterised by all three properties, property C will occur more often than property B, and B more often than A). Hierarchies concerning word order may be classified into three main groups:
a) dominance hierarchies cover the way humans perceive the world from their own deictic centre;
b) formal hierarchies cover the length and syntactic complexity of linguistic entities;
c) familiarity hierarchies cover "the speaker′s individual interests as manifested in discourse via parameters such as topicality, givenness, definiteness, etc."48
The mechanisms of formal hierarchies favour short entities being placed before longer or more complex entities. Dominance hierarchies are characterised by the preference for place entities denoting humans before other animates, these before inanimate forces and these again before other inanimate things. For example, a sentence like: The storm destroyed the roofs of many houses will be found to be either the unique possibility of expressing this information in a particular language or will at least be more frequently used in a language where the alternative way of putting it: The roofs of many houses were destroyed by the storm is possible. The inanimate force wind is placed before the inanimate thing roofs, as well as the short noun phrase the wind precedes the heavier noun phrase the roofs of many houses. The third group - the familiarity hierarchies - favours a preceding position for topic before comment, or given element before new element etc., i.e. the pragmatic functions seem to follow a certain cross-linguistic pattern.49
However, already from this short enumeration it might become visible that different hierarchies may conflict with each other. Just let us change the above mentioned example sentence to A wild, howling storm blowing with a speed of 100 km/h and more destroyed many houses. There we have the heavier construction being the inanimate force - the formal and the dominance hierarchy would each favour different positions. Dik solves this problem by his multifunctional approach which is the subject of chapter 3.5.
3.3. Constituent Ordering Principles
Dik derives general and specific principles (GPs and SPs) of constituent ordering from the empirical data collected. They may apply to both clause or phrase levels. The principles are often classified according to the position of dependent words to their respective ′heads′ or ′centres′, i.e. the centre of a clause will be the predicate, that of a noun phrase the noun, etc.50 A hierarchy among ordering principles, however, cannot be postulated.51 Examples of GPs are iconic ordering, i.e. the meaning is mirrored in the sequence of words (veni, vedi, vici is a case in point); the linearity principle which captures the tendency of dependent words to remain in a fixed order, no matter where the head may be posited; the centripetal orientation designates the tendency to maintain the relative distance of a certain dependent to its head word; the domain integrity - i.e. that interruptions of phrases/clauses by other phrases/clauses are avoided, head proximity, functional stability, i.e. the same functional entities tend to have the same distribution, pragmatic highlighting, i.e. special positions for salient information etc. Among the SP we find such principles as the tendency of heads to either follow or precede their dependents (leading to a classification of prefield and postfield languages respectively); of co-ordinators to prefer a middle position; of clauses to have a special initial position P1; of demonstratives, numerals and adjectives to prefer centripetal orientation or of relative constructions to prefer the postfield. Dik postulates 9 general and 12 specific principles to formalise his placement rules.
3.4. Placement Rules52
Placement rules match the functional relationships within a clause as assigned in the UCS via ordering templates to the proper linear form of an actual utterance of a specific language. To be able to do this, Dik derived from all principles and hierarchies a formalised general word order pattern
P2 P1 (V) S (V) O (V), P353
whereby S(ubjec) and O(bject) may be replaced by A(rgument)1 and A(rgument)2 for languages without subject and object assignment. (V) symbolises alternative positions of the verb according to language specific requirements, or by requirements caused by the illocutionary content of a clause. For instance, English sentences position their finite verbs in different places according to whether they are declarations or questions. According to the variety of possible actual word orders, each language and each illocution operate with a specific ordering pattern.
As already mentioned, P1 is a special position apparently found in all languages. Dik postulates rules according to which this P1 position may be filled:
(R0) P1 must contain one and only one constituent.
(R1) Place P1-constituents in P1, where P1-constituent = question word, subordinator, or relative pronoun.
(R2) else, place constituent with GivTop, SubTop or Foc function in P1 (optional).
(R3) else, place X in P1, where X= some satellite or a dummy element. 54
These rules are supposed to apply in an orderly way, i.e. if there is a question word, this word has to be placed in P1 and no other element may ′claim′ this position. (R1) and (R3) are considered obligatory, (R2) is optional.
An interesting interpretation by Dik of the Verb-2-position in German, Dutch and other languages is his view that both languages belong to the basic pattern VSO, but that P1 is often taken by the subject in accordance with rule (R2).
P2 and P3 positions are available for "left-" or "right-dislocated items"55 which are considered to stand outside the "predication proper"56. In Dik′s terminology P2 and P3 are available for theme and tail respectively. However, as Siewierska shows for languages with very free word order, P2 and P3 or even more special positions become necessary to generate all possible word order variations as allowed in the respective language.57
Another point that may be criticised58 is that the work of the placement rules mainly concentrates on the clause level, although some of the ordering principles also touch the order within phrases. Dik himself admits that due to the greater cross-linguistic variety within phrases "fewer convincing or at least promising explanatory principles have been advanced."59 Hawkins however, thinks that "the biggest problem [...] is found in the ordering of the arguments of the verb at the sentence level. This is typically the locus of considerable word order freedom."60 whereas "within the noun phrase and the adposition phrase word order freedom is much less extensive".61 Yet maybe we should simply acknowledge that Dik′s project is to enquire into the mechanisms of communication, i.e. his main interest more or less starts with the clause level and is rather directed towards greater units, not so much towards smaller ones. As I said before, probably no theory can be totally exhaustive.
3.5. The Multifunctional Approach
So far I have concentrated on Dik′s attempt to formalise word order and the necessary preliminaries, i.e. the hierarchies and principles. These formalisations are more the result of a pure collection of empirical data than a theory. Although this method may not yield satisfactory explanations, it surely is a good tool for working on the phenomenon.
Nevertheless, Dik gives a theoretical background which he calls his multifunctional approach. In view of the multiplicity of criteria to place words in a certain sequence, he states that the phenomenon of word order is subject to a number of different principles influencing the linearisation.
I [..] assume that constituent ordering patterns should be described and explained in terms of a number of interacting and possibly competing principles and preferences. This is what I understand by a "multifunctional" theory of constituent ordering.62
The notion of interacting principles is surely nothing new. Hawkins refers to them in a very comparable manner63, Chomsky - though more specifically - lets the modules of PP interact with each other, and in many other fields outside of linguistics such a mechanism has become a truism. Nevertheless Dik′s multifunctional approach gives the inductively found principles a plausible framework and allows for an integrative view of phenomena which otherwise appear to be random. Dik formulated the following assumptions concerning the interaction of word order principles:
A1 The actual constituent ordering patterns found in a language are the resultant of a number of interacting principles.
A2 Each of these principles is in itself functionally motivated: it is a "natural" principle with respect to some parameter of "naturalness".
A3 But two such principles do not necessarily define the same ordering preference. One principle may, for good reasons, prefer the order AB while another my, for equally good reasons, prefer BA.
A4 Therefore, no language can conform to all the ordering principles at the same time and to the same degree.
A5 The actual "solution" for constituent ordering in a given language will thus contain an element of compromise, and will to that extent be characterised by a certain amount of "tension".
A6 Shifts in the relative force of the different principles may lead to (sometimes radical) changes in constituent ordering.
A7 Where such changes relieve tension with respect to one principle, they may create new tension with respect to another.
A8 There is consequently no optimal, stable solution to the constituent ordering problem.64
Despite the fact that "naturalness" is a very imprecise term, these assumptions show the way for further research - the forces at work and their respective importance are still not clearly enough defined to form a theoretical base capable of generating all correct word order patterns, as FG strives to accomplish - an aim that FG shares with GG.65
4. Final Remarks
I think it has become clear that the two theories operate on different levels and that therefore a comparison is actually out of question. Dik′s FG on the whole is a very plausible theory founded on the notion of communication as the main function of language. His treatment of word order, however, is peripheral in view of the main body of his theory. Yet he incorporated an immense variety of empirical findings across languages in his Expression Rules. By suggesting a mechanism of competing forces, he accounts for the multiplicity of possible word orders of natural languages and provides a theoretical framework. His approach to word order is of a strongly empirical nature and although it may not explain much, it very well describes the phenomenon and meets the criterion of typological adequacy. The strengths of Dik′s FG, however, lies in the concept of a layered informational content of a sentence as postulated in the UCS which plausibly models possible mechanisms of human communication.
In contrast, the Chomskyan approach with its very abstract assumptions and its deductive method may appear absurd at first sight (as it is probably the fate of most deductive theories until empirical data give evidence of their correctness). GG wants to explicitly describe the intuitions of a speaker about grammatically correct sentences which are the preconditions for all speech or writing acts.66 As such it seems to be the more suitable theory to explain the syntactic feature word order. In spite (or perhaps because) of its elementary approach, GG is more restricted and focused in its field of research than FG. As a result it strives for a much more concrete and explicit account in regard to syntax in the original sense of the word.67 In regard to the other domains of linguistics, Chomsky is said to have conceded himself that after the study of the formal constraints on language, one should pay attention to its semantic functions.68 As such the reservations against GG Dik feels to be necessary to state in the beginning of his book seem to me rather superfluous.
By and large most of the criticism against GG I have come across does rather argument in the vein of a Yuri Gagarin who is said to have rejoiced about proving God′s non-existence because he did not meet him in the orbit, i.e. critics often criticise claims that are not actually made by GG.69 As I stated in the beginning, it is necessary to first define the project of a theory and only then to pass judgement along those lines. Generativists criticising other generativists is a more suitable ground for exploring weaknesses and strengths, yet the full understanding of the mechanisms discussed requires a deeper knowledge of the subject than I can offer at present. Therefore and in view of the bulk of literature pro and contra GG, I am not able to render any concrete criticism of GG in this essay but want to adopt a Pyrrhonian attitude - I will refrain from judgement but study further.
Lastly both theories have their fascinating points. FG′s plausibility and comprehensiveness is striking in regard to its purpose, though - as I already mentioned above - morphosyntax is somewhat neglected. On the other hand, if GG reaches its aim, we will know something about the "inner forces" which steer our linguistic system, it is to a certain extent like a Faustian enterprise. Despite the probability of further revisions of today′s assumptions, one may perhaps more broadly come to regard GG as a complementary rather than a contradictory theory to the functional approach as it is also seen by Dürscheid, Helbig and Siewierska. As I see it the one linguistic school inquires into the microcosm while the other strives to explore the macrocosm of language.
Andersen, Paul Kent. Word Order Typology and Comparative Constructions. Current Issues in Linguistic Theory, 25. Amsterdam, Philadelphia: John Benjamin Publishing Company, 1983.
Atkinson, Martin. "Generative Grammar: The Minimalist Program" in Brown, K. and J. Millers, eds. Concise encyclopedia of syntactic theories. Oxford, New York: Pergamon/Elsevier, 1996, p. 137 - 147
Beckmann, Barbara Joe. Underlying Word Order - German as a VSO Language. Frankfurt a. M.: Peter D. Lang, 1980.
Black, Cheryl A. A step-by-step introduction to the Government and Binding Theory of Syntax. Summer Institute of Linguistics - Mexico Branch and University of North Dakota. 1999. www.sil.org/americas/mexico/ling/E002-IntroGB.pdf , 17.07.2002.
Chomsky, Noam (1999). "On the Nature, Use, and Acquisition of Language" in Ritchie & Bhatia (eds.). Handbook of Child Language Acquisition. San Diego: Academic Press, p. 33-54.
Chomsky, Noam. Strukturen der Syntax. The Hague, Paris: Mouton, 1973.
Chomsky, Noam. The Minimalist Program. Cambridge, Mass., London: MIT Press, 1995.
Coulmas, Florian. Direct and Indirect Speech. Berlin, New York, Amsterdam: Mouton de Gruyter, 1986.
de Geest, Wim and Dany Jaspers. "Government and Binding Theory" in Droste, Flip G. and Johne E. Joseph, eds. Linguistic Theory and Grammatical Description. Current Issues in Linguistic Theory, 75. Amsterdam, Philadelphia: John Benjamins Publishing Company, 1991, p. 23 - 62.
Dik, Simon C. The Theory of Functional Grammar. Ed. Kees Hengeveld. 2 vols. Berlin, New York: Mouton de Gruyter, 1997.
Dürscheid, Christa. Syntax - Grundlagen und Theorien. Studienbücher zur Linguistik. Wiesbaden: Westdeutscher Verlag, 2000.
Freidin, R. "Generative Grammar: Principles and Parameters" in Brown, K. and J. Millers, eds. Concise encyclopedia of syntactic theories. Oxford, New York: Pergamon/Elsevier, 1996, p. 119 - 137.
Grewendorf, Günther and Wolfgang Sternefeld (eds.). Scrambling and Barriers. Amsterdam/Philadelphia: John Benjamins Publishing Company, 1990.
Grewendorf, Günther. Aspekte der deutschen Syntax. Eine Rektions-Bindungs-Analyse. Tübingen: Gunter Narr Verlag, 1988.
Grewendorf, Günther. Minimalistische Syntax. Tübingen and Basel: a. Francke Verlag, 2002
Hawkins, John A. Word Order Universals. New York, London: Academic Press, 1983.
Helbig, Gerhard. Geschichte der neueren Sprachwissenschaft. Leipzig: VEB Bibliographisches Institut, 1973
Koktova, Eva. Word-Order Based Grammar. Trends in Linguistics. Studies and Monographs 121. Berlin, New York: Mouton de Gruyter, 1999.
Linke, Angelika, Markus Nussbaumer and Paul R. Portmann. Studienbuch Linguistik. Reihe Germanistische Linguistik, 121. Tübingen: Max Niemeyer Verlag, 1996.
Newmeyer, Frederick J. Grammatical Theory. Chicago and London: The University of Chicago Press, 1983.
Oxford English Dictionary (OED). Vol XVII. Oxford: Clarendon Press, 1989.
Radford, Andrew. "Towards a Structure-Building Model of Acquisition" in Clahsen, Harald (ed.). Generative Perspectives on Language Acquisition. Amsterdam/Philadelphia: John Benjamins Publishing Company, 1996, p. 43 - 90
Radford, Andrew. Syntactic theory and the structure of English. A minimalist approach. Cambridge: CUP, 1997.
Siewierska, Anna. Word Order Rules. London, New York, Sydney: Croom Helm, 1988.
Ten Hacken, Pius. "Chomskyan Linguistics and the Sciences of Communication" in Studies in Communication Sciences 2/2. 2002 p. 109 - 134.
1 Koktova, passim
2 Anderson, passim
3 cf. Siewierska 8 ff
4 Dik 1
5 Chomsky(1999) 38
6 Dik 5; NLU = Natural Language User
7 Ten Hacken 128
8 cf. Coulmas 17 f
9 cf. Dik 8
10 Chomsky (1999) 37
11 Ten Hacken 124
12 Chomsky (1999) 49
13 cf. Newmeyer 12
14 This in fact, is often mentioned as a criticism of GG. However, I cannot see how this argument can hold in an academic discourse. Many theories, or even established natural laws, are not understandable without a certain effort of learning, but this does not normally render them less true.
15 cf. Linke, Nussbaumer, Portmann 104
16 Newmeyer 68
17 cf. Grewendorf (1988) 45 f
18 cf. Anderson, Dik, Hawkins
19 e.g. cf. Beckman concerning the question ′Is German a SOV, SVO or VSO language?′, or Grewendorf & Sternefeld concerning "Is scrambling base-generated or the result of a movement. Does it apply on the level of d-structure or s-structure?′ etc.
20 also called Government & Binding (GB)
21 cf. Atkinson and Grewendorf
22 Binding and Government theory are also supposed to work on the level of LF (cf. Freidin 128 f and 134 ff)
23 cf. de Geest & Jaspers 30
24 This view, however, is debatable. Jackendorff in his first presentation of X-Bar theory projects until X′′′ by adjuncts (cf. Hawkins 183), whereas Freidin, de Geest and Black do not.
25 cf. de Geest & Jaspers 34
26 Determiners and pronouns as functional heads were introduced only in the late 90s, cf. Radford (1997) 47 f
27 "Die Beobachtung, dass substantielle lexikalische Kategorien in den verschiedenen Sprachen uniforme Eigenschaften aufweisen, während die grammatischen Eigenschaften funktionaler Kategorien von Sprache zu Sprache variieren, führte zu der Hypothese, dass nicht die Prinzipien der universalen Grammatik, sondern vielmehr die funktionalen lexikalischen Kategorien als der Lokus parametrischer Variation anzusehen sind. Sprachvariation ist dieser Hypothese zufolge [...] auf die nicht-substantiellen Elemente des Lexikons beschränkt." Grewendorf (2002) 14
28 cf. Grewendorf (2002) 38 ff
29 cf. Radford (1996) 50
31 This touches the grounds of Government theory which, however, I will not be able to cover in this essay.
32 cf. Grewendorf 24
33 cf. Siewierska 122
34 cf. Helbig 284
35 Freidin 124 f, the notation is simplified
36 Freidin 125
37 Siewierska 1
38 Siewierska 1
39 cf. Dik 26 f
40 Dik passim
41 Siewierska 110
42 cf. Siewierska 112 ff
43 Dik′s specified principle SP2; Dik 405
44 cf. Siewierska 116
45 Dik 28
46 Dik 29
47 However, I see a problem with the statistical evaluation. That a certain feature may occur only in a few languages, or in many languages of the world may also be attributed to historical, not only to linguistic facts.
48 Siewierska 29
49 cf. Siewierska 30
50 Dik uses special terms as developed in the main body of his theory. As I have not introduced Dik′s terminology, I will use more general terms. However, one must be aware of the fact that Dik′s designation of a ′first operator′ and a ′phrasal term′ may in some respect differ from the general term ′noun′ or ′noun phrase′.
51 cf. Dik 399 ff
52 cf. Dik 405 ff
53 cf. Siewierska 104
54 Dik 421, GivTop = Given Topic, SubTop = what is assumed to belong to the general knowledge of the world by the interlocutors, Foc = Focus, satellite = mostly adverbial expressions which enrich the informational content by indications of manner, place, time, possibility etc.
55 Siewierska 104
57 cf. Siewierska 123 ff.
58 cf. Dürscheid 182
59 Dik 435
60 Hawkins 11 f
62 Dik 395
63 cf. Hawkins 113
64 Dik 396
65 cf. Dürscheid 192
66 cf. Helbig 281
67 "The arrangement of words (in their appropriate forms) by which their connexion and relation in a sentence are shown." OED 487
68 cf. Helbig 279
69 I have already hinted at some of those points of criticism above, e.g. that sociolinguistic, pragmatic or psychologic aspects were ′perifpheral′ etc.