The paper surveys some (mostly secondary) literature on the evidence for verb movement in the old Germanic languages, among which foremost the claims of Kiparsky and Weerman that verb movement to the Complementizer (C) position is a late (language-specific) innovation which is not attested in the earliest Germanic sources. We rely on findings and conclusions by Eythórsson (1995) on Gothic, Runic inscriptions, and Old Norse, as well as Old English, Old High German, and Old Saxon. Eythórsson argues that already at the earliest stages there are cases which must involve movement of the finite verb from the Verb Phrase to a higher functional projection. It will be demonstrated that in a well-defined set of cases this position is C, but in other cases an IP-internal functional head position seems more likely as a landing-site for the verb. Whereas the verb does not move to C when complements are topicalized in Gothic and Old English, V to C movement is obligatory in topicalizations in the other old Germanic languages. An examination of Runic inscriptions reveals that topicalized complements triggered movement of the verb to C already in the earliest stage in the northern part of the Germanic linguistic area. Finally, the verb movement in the old Germanic languages is argued to support the asymmetrical hypothesis of Verb Second proposed by Travis and Zwart. It will, furthermore, be shown that modern assumptions with respect to topical negation in all historical Germanic stages cannot be reconciled with the fact that V-topicalization as such, i.e. both negated and unnegated, is a characteristic of a special narrative genre. It will finally be investigated in some detail what the categorial status of infinitivals, both with and without the infinitival preposition, as well as the perfect participle is in the diachronic development. All of this has weighty theoretical consequences for the theory of PRO as well as the understanding of verbal diathesis. As a by-product evidence emerges that the theory of Linear Fields (‘Felderlehere’) as developed by Erdmann and Drach (or, by the same token, by Diderichsen) is observationally inadequate and requires modification in terms of UG. That modern assumptions with respect to topical negation in all historical Germanic stages cannot be reconciled with the fact that V-topicalization as such, i.e. both negated and unnegated, is a characteristic of a special narrative genre. It will finally be investigated in some detail what the categorial status of infinitivals, both with and without the infinitival preposition, as well as the perfect participle is in the diachronic development. All of this has weighty theoretical consequences for the theory of PRO as well as the understanding of verbal diathesis. As a by-product evidence emerges that the theory of Linear Fields (‘Felderlehere’) as developed by Erdmann and Drach (or, by the same token, by Diderichsen) is observationally inadequate and requires modification in terms of UG.

 

Former attempts of establishing the rules of word order as valid in the most ancient Indo-European languages with a view to a reconstruction of Proto-Indo-European sentence structure have often failed to consider the structural divergences of prosaic and poetic texts as well as interlinguistic influences that are significant for non-original (i.e., translated) texts. In the present paper, several aspects of these shortcomings will be discussed on the basis of textual materials taken from Old Indic (Vedic), Old Germanic, Latin and Greek sources. The paper will focus on the question whether there is a functional and/or historical connection between the accentual behaviour of finite verbal forms in main and subordinate clauses in Old Indic on the one hand and the difference of word order which is characteristic for German on the other hand.

 

Several scholars have claimed that in syntax one can neither set up correspondence sets nor reconstruct; Harris and Campbell (1995, Chapter 12) have argued that both are possible. This paper provides numerous examples of syntactic correspondence sets from the Kartvelian language family and from the North East Caucasian family, and it is argued that such sets provide a reliable basis for the reconstruction of syntax.

The presentation of correspondence sets raises the issue of what degree of identity we should require in such sets, both in reconstructing general patterns that apply across a language, and in reconstructing restricted, lexically governed patterns. Focus constructions in North East Caucasian illustrate the need for care when dealing with constructions that are common or easily borrowed.

A second problem that has been raised by some scholars is that if we have a correspondence set that contains different patterns, in syntax (unlike phonology) there is no criterion for determining which pattern is older. Several criteria are discussed here, including known phonological correspondences, dialect data, and syntactic relics.

The question is then addressed, "If reliable reconstruction is possible, why have so many scholars been convinced that it was not?" Answers to this focus on problems encountered in the reconstruction of Indo-European.

In this paper, I will air some views on the prospects for syntactic reconstruction from a Principles and Parameters point of view. The prospects for syntactic reconstruction correlate to some extent with how UG-driven one takes grammar change to be. Ironically (in view of the pessimism expressed in Lightfoot (1979)), the prospects for reconstruction seem to be the most favourable if we assume a) that any synchronic state represents the output of a possible grammar, and b) that the grammar of stage n +1 is based on robust evidence in the output of stage n, in line with Lightfoot (1991). On these assumptions, in combination with what we know about the pathways of change (e.g. reanalysis is often to a functional specifier or head after a movement stage, but not often the other way around), a restrictive theory of grammar should be of considerable help in reconstructing stages sparsely attested or non-attested, since it narrows down greatly the range of possible grammars for the earlier stage, as producing the input for the grammar of the earliest attested stage.

I discuss issues concerning reconstruction as posed in the empirical domain of what we grandiosely call THE Wackernagel position (as if there is one single magic position that may aspire to that term). I restrict discussion to those elements which are, for Gothic, called the second position particles, in essence a bunch of short adverbs, which follow the "first" word or constituent. We seem to be dealing here with a panGermanic phenomenon: adverbs like Gothic an, OE a, onne ‘then’, panGermanic nu ‘now’, Gothic auk, OE eac ‘also’, occur in the earliest West-Germanic stages in a position between that for the pronominal subject and the DP subject. Similar facts are found in modern and historical Scandinavian. The facts for Gothic are analysed in Ferraresi (1997). I will present an analysis for the early West-Germanic facts which is by and large valid for the modern panGermanic (except English) phenomenon, and clearly cuts across the the OV/VO divide, as we would expect. I also present an analysis for (sparsely attested) Gothic which is a modification of Ferraresi (1997). Comparing between these stages, I will discuss issues of reconstructability, as well as venture into reconstruction of Proto-Germanic.

Ferraresi, Gisella (1997) Word order and Phrase structure in Gothic. Stuttgart PhD thesis.

Lightfoot, David (1979) Principles of Diachronic Syntax. CUP.

– (1991) How to set Parameters. Cambridge, Mass: MIT Press.

 

Nonfinite subordinate clauses typically develop in one of two ways: by deverbal nominals acquiring infinitival status, and by reanalysis of constituents of matrix clauses with participial predicatives as subjects of subordinate clause headed by the participle. In this talk I examine instances of the second of these scenarios for clues about the mechanism of syntactic reanalysis, as a contribution to the program of grounding syntactic reconstruction more firmly in a theory of change.

Reanalysis is commonly held to occur when some of the data encountered by learners is ambiguous and they project from it a grammar that differs from the grammar of the speakers whose language they are acquiring. Whether implemented in terms of parameter-setting, cue-based, or constraint-ranking models of acquisition, this view of reanalysis fails to provide a plausible triggering mechanism for the change. This conceptual difficulty is particularly acute in cases such as those that form the subject of my study, where reanalysis affects fundamental aspects of grammatical structure, so that innovating speakers must be assumed to have overridden salient positive data that is inconsistent with the grammar they nevertheless end up acquiring.

Instead, I propose that reanalysis is a process of grammatical optimization motivated by internal structural factors. Reanalyses resulting in new types of participial subordinate clauses offer support for this view. In each case, the proportion of ambiguous data at the stage prior to the reanalysis is arguably insufficient to lead a learner astray. On the other hand, each reanalysis is motivated by prior syntactic changes that made the old analysis of the participle as a predicative complement synchronically unmotivated. For example, the rise of the locative and genitive absolute constructions in Sanskrit is triggered by a series of shifts in the uses of the system of adverbial cases, and the rise of participial "clause substitutes" in Finnish is triggered by the loss of the predicative construction itself, independently observable in the case of non-participial predicatives.

The implication for syntactic reconstruction is that reanalyses must be analyzed in the context of the entire grammatical system, and that they must be justified not so much by ambiguous surface structures as by the structural causes that drive them.

 

The possibilities for reconstructing a realistic proto-syntax, whether at the level of a grammar or a language, are very limited. Change is not systematic over the long-term and consequently the two main techniques for reconstruction, the comparative and internal methods, do not get us far.

If historically related languages are effectively identical in some respect, it may be plausible to attribute that common property to the proto-system. However, where the languages differ in some respect, e.g. in word order properties, one can draw conclusions about the proto- system only if one knows that property *a* may be replaced by property *b* over time, but not vice versa. Such hypothetical trajectories of change tend not to be well-founded, and that fact also undermines the reliability of any analog of so-called internal reconstruction.

Things have APPEARED to be different in research programs which deal in terms of long-term trajectories. Recent examples are the Greenberg-based "typological" theories of the 1970s, grammaticalization theories, and theories which attribute change to biases built into Universal Grammar. All of these approaches are flawed and it seems more appropriate to view change as "chaotic", reflecting "sensitive dependence on initial conditions:" minor differences in primary linguistic data (PLD) may entail grammatical re-analyses.

In that case, reconstruction of a realistic proto-syntax will be possible only if somebody has a theory of how PLD change from one generation to another. In reality, this is too contingent to predict. If we are willing to settle for accounts which explain change when it happens but which do not predict change, then we must assume that reconstrucrtion of a proto-syntax can only be an arbitrary device to state historical relationships, roughly the view of Meillet of reconstruction in general.

Most reconstructions constitute myths and they are based on mythical notions about the nature of change.

 

In this paper I will discuss some conceptual foundations of a research program for a very restrictive theory of syntactic change. In particular I will propose to explore the possibility of the most restrictive conceivable theory of diachronic syntax, namely that syntax by itself is diachronically completely ‘inert’. In other words it will be suggested that linguistic change may only originate as an interface phenomenon, in the sense of Chomsky’s Minimalist program, perhaps just for reasons concerning the relation between language and the external world (classical interference or pressures from the conceptual and articulatory-perceptual systems). Syntactic change (e.g. parameter resetting) would only happen as a totally predictable reaction by a deterministic language acquisition device to a superficial (interface) change in the primary data. Such a research program, call it the Inertial Theory (IT), excludes the intervention of probabilistic models in the development of syntax and largely trivializes the sociolinguistic problem of convergence of a speech community on the same innovations. Also, IT views syntax as one of the most diachronically conservative domains and thus quite suitable for long-range historical explanations. As any non trivial hypothesis, this approach is faced with plenty of surface counterevidence and is likely to be only partly correct. In order to test its value,it is necessary to show that a number of cases of syntactic change, apparently recalcitrant to the previous restrictive scheme, can eventually be reduced to independent changes in the phonology/lexical semantics of the languages involved which may have triggered the syntactic reshaping. Some example of this procedure will be presented probably from the history of Romance nominal structures and the Germanic complementizer system.

 

The thesis, that Indo-European contained no subordinate clauses, is still maintained by Indo-Europeanists today. Treated in this paper are the possibilites of expressing the content of virtual subordinate clauses, i.e. of subject, object, adverbial and attribute clauses, in Indo-European languages. Whereas even recent studies pertaining to subordinate clauses in Indo-European languages have been semasiological in nature – i.e. the definitions of individual conjunctions are investigated – in this study onomasiological principles are implemented. To be shown are firstly, how content equivalent to that of subordinate clauses can otherwise be expressed, and secondly, what interpretative differences might exist between the indiviual possibilities of expression. Variants in "Konkurrenz-Syntax" are thus the subordinate clause (including its conjunction, its verb position and its use of modality), the infinitve, the participle, the absolute construction, the gerund, the gerundive, and the (prepositional) construction with an abstract. The languages included in the study are Armenian, Latin, Greek, Old Indian, and Germanic.

 

Syntactic reconstruction may, pace Lightfoot (1979), be possible. What is needed, as Harris & Campbell (1997:353) point out, is a way of determining syntactic correspondences. The principles-and- parameters model of cross-linguistic variation is able to provide this. Suppose that we can define a finite set of binary parameters which defines the variation allowed by universal grammar. Parameter values are expressed as members of the set {1, 0}. In this way, we can conceptualise reconstruction of the value of parameter p as follows:

daughter 1: p = 0 daughter 2: p = 0 daughter 3: p = 0 daughter 4: p = 1

Following the traditional methodology of reconstruction, we would naturally assume p = 0 in the parent language. An example might be word-order in the ancient Indo-European languages. Daughters 1, 2 and 3 would represent the prevalent OV pattern (found in Gothic, Latin, Sanskrit and elsewhere), while daughter 4 represents VO Old Irish. Here p is the OV parameter (p = 1, VO; p = 0, OV). p = 0 in, say, Latin is the continuation of the Indo-European value ("continuation" thus refers to a potentially variant property which is stable across grammars in the diachronic domain; it is the case where a parameter value is "correctly" set by acquirers over a number of generations). Following Hale (1996:162), we can define a proto-language as a "set of grammars which are non-distinct in their recoverable features," the recoverable features in question being parameter values. We discuss concrete examples of this type involving aspects of the C-system in Indo-European. Older Indo-European languages have an infixation-like process affecting pronouns and other material, illustrated in (1):

These examples have the abstract form (2):

Force and Fin make up a split-C structure (see Rizzi (1997)). X was a pronoun in OIr and Latin; in Gothic it was an interrogative particle. The IE C-system thus seems to have had the following properties: (i) it attracted V, (ii) it attracted pronouns, (iii) it attracted particles. Pronouns were never initial; particles vary in having to be, being unable to be and optionally being; V was initial only where there was no initial particle. We will describe these patterns in the three daughter languages just mentioned and make inferences using the methodology described above regarding the likely IE situation.