Object positions
- comparative syntax in a cross-theoretical perspective

1. General research goals      1.1 A cross-theoretical perspective      1.2 A comparative perspective 2. The central empirical areas 3. The subprojects      3.1 Functional and formal linguistics      3.2 Functional linguistics      3.3 Formal linguistics List of references

In linguistics today, two theoretical paradigms would seem to dominate, the formal paradigm and the functional paradigm. Both approaches have many adherents worldwide and also if we just look at Scandinavia. In Denmark, however, the interest has primarily been focussed on the functional approach.

It is our opinion that if Danish linguistics is to assert itself on the international level now and in the future, then the dominant functional approach needs a formal and theoretical opponent which is both qualified and constructive. We also believe that the formal approach needs to take the challenge posed by functional linguistics seriously, a challenge which is only rarely taken seriously in international formal linguistics.

Our project applies the two approaches in parallel to a concrete empirical area (object positions), in that the project is divided up into two subprojects, a functional one and a formal one. By comparing different accounts of the same empirical data offered by the two approaches, it will be possible to investigate what and how much the two different approaches have in common, and also what distinguishes one from the other. It will furthermore be possible to see whether and to what extent each approach can be inspired and influenced by how the other approach goes about solving a particular problem. We are convinced that such synergy effects will be both considerable and inspiring, and that the project thus may be influential for subsequent work within each of the two approaches.

Both the formal and the functional approach are concerned with linguistic form, e.g. how a word is pronounced, what it means, or where it occurs in the sentence. Formal linguistics is primarily interested in the linguistic form itself, i.e. in the internal structures of language. Functional linguistics is primarily interested in the communicative function that a linguistic expression has in the world outside language, i.e. in the connection between language and external factors. The basic assumptions of the two approaches may be formulated as follows:

As should be clear, it would not be fair to say that the two approaches simply exclude each other. It is possible to agree with both approaches, e.g. by holding the opinion that certain linguistic phenomena are best explained by making reference to communicative function, whereas other linguistic phenomena are best explained without making reference to communicative function.

It is ultimately an empirical question whether a given property of a language or a given difference between two languages is best accounted for with (functionalism) or without (formalism) reference to communicative function. In other words, when much more is known about languages and linguistics, it might be possible to tell which of the two approaches results in the explanation that by means of the fewest extra assumptions yields the highest number of predictions which are testable and which are not falsified in such tests. But as it undoubtedly will take a long time to reach such a point, we believe that in the meantime linguists of both persuasions should be prepared not just to show respect for each other's work but also to work with and learn from each other, instead of resorting to what might be called extreme formalism or extreme functionalism, which might be characterised as the negation of the other approach:

The formal approach: Linguistic form can be characterised independently of communicative function.	Extreme formalism: Communicative function is completely irrelevant for linguistic form.
The functional approach: Communicative function can determine linguistic form.	Extreme functionalism: Linguistic form can not possibly be characterised independently of communicative function.

The project is not only cross-theoretical, as discussed above, but also comparative. In addition to comparing languages empirically, we believe that comparative linguistics should also seek to account theoretically for as many surface differences as possible by deriving them from as few underlying abstract differences as possible. In this way it may be established which aspects of a given language, e.g. Danish, are also found in other languages and which aspects are specific to a single language.

Such a typological perspective makes it possible not only to establish typological connections and predictions (e.g. of the kind "only languages which have X, also have Y"), but also to explain and justify these theoretically.

By comparing different languages (e.g. Danish vs. English) and different stages of the same language (e.g. Old Norse vs. modern Danish) we obtain a good basis for establishing what constitutes possible (and impossible) types of language variation.

In this way a solid foundation is set up for many types of applied linguistics, ranging from computational linguistics to research into learning difficulties. Thus theory and empirical research may be integrated.

Here there is also a clear link to the strategy plan 2003-2007 of the Danish Humanities Research Foundation, Humanistisk forskning - en nødvendighed, `Humanities Research - a necessity', (cf. not only page 19 about "the description of the Danish language in relation to other languages" but also page 28 about "the view of the human being"), in that comparative linguistics in this way strives to find out which kinds of variation exist between languages, and which kinds do not exist, and thus makes a contribution to our knowledge about the scope and the limitations of the human brain.

The concrete manifestation of the comparative aspect is that both the functional and the formal subprojects will consider not only modern Danish, but also related languages like Middle Danish, Old Danish and Old Norse, other Scandinavian languages including Icelandic and Faroese, other Germanic languages like English, Yiddish, and German, and finally also more distant languages like French and Hungarian.

These ambitions, from a language point of view, are not unrealistic, considering the research experience of the four linguists involved in the project. To mention only some representative works: Tavs Bjerre's Ph.D. dissertation (University of Southern Denmark, 2003) on complex predicates in English, Danish and Hungarian; Eva Engels' Ph.D. dissertation (University of Potsdam, 2004) on adverbial positions in English, German and French; Henrik Jørgensen's doctoral dissertation (Aarhus University, 2000) on the pronouns of Danish, Norwegian and Swedish and their dialects; and Sten Vikner's habilitation dissertation (University of Tübingen, 2001) on verbal positions in Danish, Faroese, Icelandic, English, German, Yiddish, Afrikaans and others.

The empirical areas that are the focus of attention in both subprojects are the different possible positions that an object may occupy inside the clause. This section will briefly sketch some of the basic data which will be considered in both subprojects.

A basic difference concerning object positions is the division of languages into the type object-verb (OV languages) and languages of the type verb-object (VO languages). The OV languages include e.g. German, Dutch, and Frisian, the VO languages include e.g. the Scandinavian languages and English (Vikner 2001b). In the discussion of the relative placement of the object and the verb, it is customary to consider the position of the verb in embedded clauses, which for different reasons is seen to represent the most basic position (also because it is the same position that non-finite verbs have in main clauses, Heltoft 1986, Platzack 1986). The OV/VO difference appears from the following two examples:

The OV languages are very complex as far as object positions are concerned. Thus, for example, the positions of indefinite and definite objects may differ. Both types of objects occur to the left of the verb, but the natural position of definite objects is to the left not just of the verb but also of sentence medial adverbials (like e.g. ohne Zweifel and nie), cf. (2a) and (2b), see also Engels (2004a):

In contrast to this, the difference between indefinite and definite objects does not play the same role in the word order of the VO languages, cf. (3a) and (3b):

The difference between (2a) and (2b) is often referred to as ”scrambling”. This is a phenomenon characteristic of the OV languages. However, in the Scandinavian languages, a phenomenon exists that could be seen as a limited version of scrambling. We are here referring to what is known in Danish as ”letledsreglen”, `the rule of light objects', and which only involves super-definite objects, i.e. the definite pronouns (Jørgensen 2000a:87, 2000b, Vikner 1994):

Further requirements in connection with the application of the light object rule in Danish are that the clause in question must be a main clause and that the main verb must be the finite verb. Similar requirements apply to the corresponding construction in Icelandic, except that the requirement that the object should be a pronoun does not apply:

Therefore one name, ”object shift”, is used for the Danish (/Nowegian/Swedish) construction in (4b) and the Icelandic construction in (5b) taken together.

A different type of variation concerning object positions is found in the so-called verb particle constructions, i.e. constructions containing expressions of the type throw away, look up, etc. In such constructions, Danish requires the object to precede the particle, Swedish requires the object to follow the particle, and English and Norwegian allow both (Jørgensen 2000b:95, Vikner 1987):

In causative constructions (Vikner 1987), Danish, Swedish and English are parallel, as long as the embedded clause contains a subject (e.g. Mary in (8c)):

If the subject, on the other hand, is missing, then Danish has the object precede the verb, Swedish has the object follow the verb, and English simply cannot do without the subject: Both (9c) and (9d) are impossible, as indicated by the asterisk in front of the examples.

Yet another difference in object positions has to do with negation. In negated clauses with an indefinite object, this object may either be positive if the sentential negation is expressed independently at the same time (ikke ... nogen penge in (10a) = not ... any money in (11a)), or it may be negative provided that the sentential negation is not expressed independently (ingen penge in (10b) = no money in (11b)), Jørgensen (2000a:93):

The difference between Danish and English is that the negative object moves leftwards into the position of the sentential negation in Danish, (10b), but has to stay in the normal (postverbal) object position in English, (11b):

As will hopefully be apparent, the position of the object displays a series of complex variations which interact in interesting ways with other syntactic phenomena, such as definiteness, adverbials, causative constructions, negation etc. This makes object positions a very suitable topic for the comparison of functional and formal linguistic explanations.

We believe it to be very important to try to reduce the distance between theory and empirical data as much as at all possible, by making sure that theories and hypotheses have a solid foundation in empirical observations and also that empirical observations are always seen in a theoretic perspective.

In section 1.1, we mentioned the two theoretical paradigms that dominate linguistics today, the formal paradigm and the functional paradigm. Functional linguistics in Denmark (cf. e.g. Engberg-Pedersen et al. 1996 and Togeby 2003a,b) goes back to an absolute classic in Danish grammar, Diderichsen (1946). We want to compare this approach to formal linguistics in the form of generative linguistics, primarily Government & Binding Theory, Minimalism, and Optimality Theory (Chomsky 1981, 1995, Haegeman & Guéron 1999, Radford 2004, Kager 1999, Legendre, Grimshaw & Vikner 2001).

Our intention is to develop and compare the two approaches by running two parallel and cooperating subprojects on object positions, a functional one with Henrik Jørgensen as principal investigator and Tavs Bjerre as research assistant, and a formal one with Sten Vikner as principal investigator and Eva Engels as research assistant.

In our view, Aarhus University offers an ideal location for such a coordinated project, as it houses both a significant cluster of functional linguists (at the Scandinavian Institute and at the Dept. of Linguistics within the Institute of Anthropology, Archaeology and Linguistics) and a significant cluster of formal linguists (at the Dept. of English within the Institute of Language, Literature and Culture).

The experience with functional linguistics within the project consists both of teaching experience (Bjerre and Jørgensen have both taught using Togeby 2003b) and research experience, see e.g. Jørgensen (1991, 1999, 2000a,b, 2001). Furthermore, Vikner (1999) discusses about Diderichsen's sentence model, and the corresponding topological model for German is the foundation of Wöllstein-Leisten et al. (1997), a textbook co-authored by Vikner.

Functional linguistics, as mentioned above, attaches decisive importance to the communicative function of language. The function of linguistic entities is to communicate a particular content. The basic idea is therefore that language should be examined in the light of our knowledge about the way consciousness processes various types of input, and that considerations to do with other people being able to understand what is said play a decisive role for linguistic form. In other words, functional linguistics is concerned with the connection between linguistic and non-linguistic cognitive processes and with the communicative function of language.

Within the functional subproject on object positions, we will, among other things, investigate the following three theoretical themes:

Case theory: Theories about semantic cases and their theoretical foundations are abundant within functional linguistics, and to some extent also within formal linguistics (Gruber, Jackendoff m.fl.), in spite of their clearly semantic character. The project primarily focusses on semantic objects that refer to entities that take part in the action denoted by the verb, the completely unmarked type of object, which must be assumed to be possible with any type of predicate. Fillmore calls this ’objective’; in the terminology of Widell (1996) and Togeby (2003b), which we intend to follow, it is called E-role (”for that which exists”, Togeby 2003:78). Many other types of case theory tend to subdivide semantic cases ad infinitum, and it may be difficult to delimit which criteria are necessary to set up (or abandon) a particular semantic case (cf. Jørgensen 2000b, chap. 3). Compared to this, the terminology of Widell/Togeby has the advantage that it is founded on a clear semantic intuition and that the system has developed criteria that make it possible to test the results against concrete observations.

Semantic case theory thus has a central function in the subproject in that the data will consist of E-roles. We expect to be revising and elaborating the operational definitions of E-roles given in Togeby (2003b).

Information structure and reference: As we have already seen (in section 2 above), the referential status of an object is decisive for whether a certain object position is used or not: definite objects behave differently from indefinite ones, generic objects differently from specific ones, and pronominal objects again differently from all the others, as the examples in section 2 above show. Corresponding problems arise in connection with sentential objects, which also seem to have a number of possible positions, the use of which depends on the status of the sentence with respect to presupposed content, cf. Jørgensen (2001). Also here there is a significant potential for investigations into the relation between considerations of form and function.

Sentence structure and information structure: Eventually, the project will also throw light on a well-established basic assumption within functional grammar, the assumption of a connection between on the one hand familiarity, topic-status, and grammatical subject and on the other hand unfamiliarity, comment-status, and grammatical object. This theory is already present as a basic assumption within the Prague Functional School, e.g. Jan Firbas. Presumably this assumption, which has never been ”proven”, may be made significantly more probable (or potentially disconfirmed), precisely by being systematically confronted with a formal theory concerned with the same data. We see a significant motivation for our project precisely in the possibility for getting to terms with this very tenacious assumption.

The experience with formal linguistics within the project consists both of teaching experience (Vikner since 1984) and research experience, see e.g. Engels (2004a,b), Vikner (1987, 1994, 1995, 2001a,b).

Generative theories have in common that they take it that only part of human linguistic knowledge is acquired, whereas another part of this knowledge is innate. In other words, certain aspects of language are the way they are because the human brain is structured the way it is. Thus such theories have an interesting explanation to offer for why certain characteristics do not vary between different languages: Non-attested types of variation as well as universal linguistic features may be derived from the innate part of human linguistic knowledge.

Optimality theory (cf. Legendre, Grimshaw & Vikner 2001 and also Vikner 2001b) analyse linguistic variation as a result of conflicts between different generalisations or constraints. Such constraints may be violated, and thus it may be explained why different languages have different ways of solving a conflict between two constraints. Which solution is chosen to a given conflict, depends on which rules are given a high priority by a particular language. An example could be that in both Danish and French embedded clauses, a finite main verb is preferably next to both the subject and the object:

In certain cases, the verb cannot both be next to the subject and to the object, and a conflict thus arises between the two constraints. Here it may be seen that Danish assigns higher priority to the verb being next to the object, whereas French assigns higher priority to the verb being next to the subject:

Optimality theory thus builds on the assumption that different versions of the same sentence compete. In the above example, there are two different versions of the Danish sentence, viz. Jeg tror ikke at Peter ofte læser avisen, (13a), and Jeg tror ikke at Peter læser ofte avisen, (13b), and here Danish prefers the first version, whereas French prefers what corresponds to the second version. The form of every sentence is a result of which constraints are observed and which are violated. A well-formed sentence is the version that only violates constraints with low priority, i.e. constraints with a lower priority than the constraints violated by competing versions of the same sentence.

In other generative theories, e.g. Minimalism (Chomsky 1995) and the somewhat older Government & Binding Theory (Chomsky 1981), parameters are taken to be at the root of language variation. Parameters are to be seen as distinct from principles: Whereas principles determine invariable (i.e. universal) linguistic properties, parameters determine the possibilities of variation. Thus a parameter concerning object positions is often suggested, where the two parametric options are object-verb or verb-object.

Within the formal subproject on object positions, we will, among other things, investigate the following three theoretical themes (which have clear parallels to the functional themes in section 3.2):

Case: To which extent is the distribution of object positions related to case assignment? Here it is essential to use a comparative approach, as Danish has no case differences outside the pronominal system. Within formal linguistics, it is often assumed that all noun phrases are assigned some form of (abstract) case, regardless of whether or not this has any morphological manifestation. Investigations into object positions, object types and case may contribute to throwing light on the correlations between the case of the object and its position, or potentially to question the validity of such correlations.

The project will also shed light on case assignment differences between languages, e.g. it could be suggested that verbal particles do not assign case in Danish, (6a), which is why the object has to move, whereas the objects do not move in Swedish, (6b), because the verbal particle bort, `away', assigns case to it.

Projections: Because generative linguistics is based on the idea that language and the ability to speak is specific to humans, and that this is due to certain innate properties in the brain, which are the same for all human beings, different object positions should all potentially be available to all languages. In other words, clause structure contains positions which are not necessarily exploited e.g. in Danish, but which might be filled e.g. in Icelandic or in an earlier stage of Danish. The mechanisms that govern this distribution may then depend on parameters (as in Minimalism) and/or ”constraint competition” (as in Optimality Theory).

An example could be differences and similarities between ”scrambling” and ”object shift”: Are these two the same type of movement (e.g. are they blocked under the same circumstances?), and is the position to which the object moves the same in the two types of movement?

Mapping: To which extent is there a semantic effect connected with the different object positions? Diesing (1997) and Vikner (2001a) suggest that the answer is "to a considerable extent" as far as indefinite objects in German, Yiddish and Icelandic are concerned, but how does this look for other types of objects and for languages with less morphological case than these three?

Bjerre, Tavs: 2003, “Syntactically and Semantically Complex Predicates”, Ph.D. dissertation, Syddansk Universitet.

Chomsky, Noam: 1981, Lectures on Government and Binding, Foris, Dordrecht.

Chomsky, Noam: 1995, The Minimalist Program, MIT Press, Cambridge Mass.

Diderichsen, Paul: 1946, Elementær Dansk Grammatik, Gyldendal, Copenhagen.

Diesing, Molly: 1997, "Yiddish VP Order and the Typology of Object Movement in Germanic" in Natural Language and Linguistic Theory 15.2, 369-427.

Engberg-Pedersen, Elisabeth, Michael Fortescue, Peter Harder, Lars Heltoft, & Lisbeth Falster Jakobsen (eds.): 1996, Content, Expression and Structure: Studies in Danish Functional Grammar (Studies in Language Companion Series 29), John Benjamins, Amsterdam.

Engels, Eva: 2004a, "Adverb Placement. An Optimality Theoretic Approach", Ph.D. dissertation, Universität Potsdam.

Engels, Eva: 2004b, "Optimizing adverb placement in gap constructions" in J. R. Austin, Stefan Engelberg & Gisa Rauh (eds.), Adverbials: The interplay between meaning, context, and syntactic structure, John Benjamins, Amsterdam, 67-101.

Haegeman, Liliane & Jacqueline Guéron: 1999, English Grammar - A Generative Perspective, Blackwell, Oxford.

Heltoft, Lars: 1986, "Topologi og syntaks. En revision af Paul Diderichsens sætningsskema" in Lars Heltoft & John E. Andersen (eds.), Sætningsskemaet og dets stilling - 50 år efter (Nydanske Studier og Almen Kommunikationsteori, NyS 16/17), 105-130.

Jørgensen, Henrik: 1991, "Om de danske personlige pronominer" in Danske Studier 1991, 5-28.

Jørgensen, Henrik: 1999, “Nogle noter om en semantisk-hierarkisk klassifikation af nominerne i sætningen”, Almen Semiotik 15, 150-170.