47
NONSPECIFIC FREE RELATIVES AND (ANTI)GRAMMATICALIZATION IN ENGLISH AND GERMAN TORSTEN LEUSCHNER Abstract One of the most contentious issues in historical syntax at the moment is the problem of directionality, in particular its theoretical status in the debate over grammaticalization. Grammaticalization is usually said to proceed unidirectionally from less grammatical to more grammatical. Under a recent proposal by Haspelmath (2004), genuine violations of this constraint are called "antigrammaticalization", and the rest "retraction". The present paper takes up this distinction with respect to the problem of clause- integration in complex sentence constructions in English and German, focusing on a potential example of antigrammaticalization, viz. the historical relationship between nonspecific free relatives and concessive conditionals. Concessive conditionals are always adjuncts and hence much more disintegrated than free relatives synchronically, but since there is also typological evidence which suggests that concessive conditionals are historically derived from free relatives, the development from one to the other could conceivably have involved a decrease of clause-integration rather the increase that one would expect under the unidirectionality constraint. Using present-day and historical data, the present paper argues that the historical link between free relatives and concessive conditionals in English and German does not in fact involve antigrammaticalization at all, and only a limited amount of retraction. The impression of antigrammaticalization arises partly because concessive conditionals have become partially dissociated from free relatives in a process which itself bears some hallmarks of grammaticalization and which, on a larger scale, can be seen as part of the formation of the prototype traditionally known as "hypotaxis". 1. Introduction One of the most contentious issues in historical syntax at the moment is the problem of directionality, in particular its theoretical status in the debate over grammaticalization. 1

Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

NONSPECIFIC FREE RELATIVES

AND (ANTI)GRAMMATICALIZATION

IN ENGLISH AND GERMAN

TORSTEN LEUSCHNER

Abstract

One of the most contentious issues in historical syntax at the moment is the problem of directionality, in particular its theoretical status in the debate over grammaticalization. Grammaticalization is usually said to proceed unidirectionally from less grammatical to more grammatical. Under a recent proposal by Haspelmath (2004), genuine violations of this constraint are called "antigrammaticalization", and the rest "retraction". The present paper takes up this distinction with respect to the problem of clause-integration in complex sentence constructions in English and German, focusing on a potential example of antigrammaticalization, viz. the historical relationship between nonspecific free relatives and concessive conditionals. Concessive conditionals are always adjuncts and hence much more disintegrated than free relatives synchronically, but since there is also typological evidence which suggests that concessive conditionals are historically derived from free relatives, the development from one to the other could conceivably have involved a decrease of clause-integration rather the increase that one would expect under the unidirectionality constraint. Using present-day and historical data, the present paper argues that the historical link between free relatives and concessive conditionals in English and German does not in fact involve antigrammaticalization at all, and only a limited amount of retraction. The impression of antigrammaticalization arises partly because concessive conditionals have become partially dissociated from free relatives in a process which itself bears some hallmarks of grammaticalization and which, on a larger scale, can be seen as part of the formation of the prototype traditionally known as "hypotaxis".

1. Introduction

One of the most contentious issues in historical syntax at the moment is the problem of

directionality, in particular its theoretical status in the debate over grammaticalization.

Grammaticalization is often defined as leading from less to more grammatical (e.g. Hopper /

Traugott 2003), hence it is not surprising that authors who seek to question the validity of

grammaticalization as a framework for theorizing on language change often bring up

counterexamples which they say violate alleged directionality constraints (see e.g. several

papers in Campbell, ed, 2000).

The traditional candidate term for such changes is "degrammaticalization". Among the

diverse definitions of degrammaticalization discussed by van der Auwera (2002), there is one

which covers changes contrary to unidirectionality (ibd.: 21, def. 4):

(1) [Degrammaticalization is] the undoing of a grammatical formative into something other

than a grammatical formative, or [...] into a grammatical formative with a weaker

degree of grammatical function.

1

Page 2: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

Discarding the multiply polysemous term "degrammaticalization", Haspelmath (2004) coins a

new label for this class of phenomena: "antigrammaticalization". He does not formally define

antigrammaticalization, but in view of his own definition of grammaticalization as cited in (2)

(ibd.: 26), he would presumably define antigrammaticalization as something like (3):

(2) Grammaticalization is a diachronic change by which the parts of a constructional

schema come to have greater internal dependencies.

(3) Antigrammaticalization is a diachronic change by which the parts of a constructional

schema come to have weaker internal dependencies.

The term "antigrammaticalization" is intended to designate reversals of potential

grammaticalization paths, i.e. types (not necessarily tokens) of changes that go "against the

general direction of grammaticalization" (ibd.). For instance, if (4) represents an attested path

of grammaticalization, (5) constitutes its reversal and hence antigrammaticalization:

(4) grammaticalization: A1 > A2

(5) antigrammaticalization: A1 < A2

Given a visualization of grammaticalization paths as leading from left to right as in (4),

"antigrammaticalization" can be described as the acquisition of a variant further to the left

along a given grammaticalization path. This distinguishes genuine counterexamples to

unidirectionality like (5) from changes which only appear to violate unidirectionality and

therefore do not to represent antigrammaticalization but something else, which Haspelmath

calls "retraction" (2004: 15). Retraction does not involve the new acquisition of a less

grammaticalized variant but the loss of a more grammaticalized variant. For instance, if the

variant A2 in (4) were lost again (the loss being represented here by brackets), this would

qualify as a case of retraction:

(6) retraction: A1 (> A2)

Retraction thus forms a "grammaticalization chain in which a right-hand member becomes

obsolete" (ibd.: 33).

2

Page 3: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

The empirical relevance of the distinction between antigrammaticalization and

retraction lies in the fact that very few alleged counterexamples to unidirectionality conform

to (5). Rather, according to Haspelmath, most alleged counterexamples to unidirectionality

qualify as retraction, not antigrammaticalization (2004: 35), reducing the number of attested

violations of unidirectionality to a handful at a stroke. This trend will be confirmed by the

present paper, which however seeks not only to affirm the directionality constraint on

grammaticalization but also look beyond it by presenting a case-study of an apparent instance

of antigrammaticalization from the domain of complex sentence constructions. The

phenomenon in question (originally raised by Haspelmath himself in a paper co-written with

König, 1998) has to do with the historical relationship between nonspecific free relative

clauses and a variety of adverbial clauses called concessive conditionals (cf. Leuschner

2003). Example (7) represents a nonspecific free relative, (8) a concessive conditional:

(7) Whatever he says is nonsense.

(8) Whoever he is, he can't just walk in like this.

What makes these clause-types interesting from the point of view of directionality is first of

all the fact that they differ dramatically with regard to the type and degree of syntactic

dependency between the subordinate clause and the matrix clause: whereas free relatives tend

to be more or less deeply embedded in their matrix clause, in which they prototypically

function as arguments, concessive conditionals are adjuncts and part of the syntactic

periphery (d'Avis 2004 on German). On the other hand, they are introduced by the same kind

of WH-based subordinator in English (and also German), and this naturally leads one to

suspect that there is some diachronic link between them. This idea is supported by

typological evidence: in languages with special (WH-based) pronouns for free relatives,

concessive conditionals use these special pronouns, not the bare WH-words (Haspelmath /

König 1998: 606). This is the case, inter alia, in Greek (FR = free relative):

(9) ti? 'what?' > ó-ti 'FR-what' > o-ti-thipote 'FR-what-soever' (ibd.)

Similar patterns are reported from Bulgarian, Slovene, Georgian and other languages which

have special markers for free relatives (ibd.). And while there are no free-relative markers as

such in Germanic, relative particles are involved in the formation of concessive conditionals

3

Page 4: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

in some ancient or conservative languages like Old Norse / modern Icelandic (sem, er) and

Gothic (-ei) (see Leuschner 2001).

In order to substantiate the hypothesis that free relatives are a diachronic source of

concessive conditionals, Haspelmath / König (1998) raise the possibility that concessive

conditionals could have arisen from free relative clauses through a shift of the latter towards

the periphery of the sentence via left-dislocation:

(10) a. Whatever she writes is brilliant.

> b. Whatever she writes, it is brilliant.

> c. Whatever she writes, producers want to read it.

> d. Whatever she writes, producers queue up to buy the movie rights.

Among the various grammaticalization paths suggested by Haspelmath / König (1998: 625)

to account for the totality of recurrent sources for concessive conditionals in their sample

(which also include conditionals, interrogatives, lexical expressions of indifference,

agreement and permission, etc.), this particular path is special because it is the only one

whose directionality violates general expectations of grammaticalization. After all, the shift

leading from clasual embedding to increasingly loose adjunction in (10) would constitute a

sharp decrease in clause-integration and hence a genuine counterexample to unidirectionality

(ibd.: 623) – i.e. antigrammaticalization.

In what follows, I will discuss data from the history of free relatives in English and

German (and occasionally also Dutch) to suggest that the directionality hypothesis (10) is too

simplistic: in fact, both concessive conditionals and free relatives as we know them today

developed from adjoined free relatives, not embedded ones, and since there is therefore no

antigrammaticalization involved, both the typological evidence and unidirectionality can be

accomodated. I begin with a brief overview of nonspecific free relatives in English and

German and their relationship to specific free relatives on the one hand and concessive

conditionals on the other (section 2), followed by a demonstration that the historical

development of clause-integration in free relatives did not start from tight embedding as in

(10)a. but from more or less loose adjunction, i.e. from a situation in which the difference

between the two construction types was much less distinct than it now appears to be (section

3). With the directionality issue relegated to the background, the rest of the paper

concentrates first on further changes which from the 14th century onwards (and even later in

German) produced the increasing dissociation of concessive conditionals and free relatives

4

Page 5: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

whose results we encounter in the present-day languages (section 4), and finally expands its

perspective to argue that these changes are embedded in larger trends in the development of

complex sentence constructions (section 5), viz. the formation of the prototype of complex

sentence constructions which is traditionally known in the languages in question as

"hypotaxis" (cf. Leuschner / Van den Nest 2005).

2. Nonspecific Free Relatives and Concessive Conditionals

2.1. Basic Distinctions

As the name suggests, nonspecific free relatives (henceforth: NFRs) are the nonspecific

variety of the free relative. Free relatives are usually analysed as relatives whose relativizer

has been merged with the antecedent (hence "fused relative", Huddleston / Pullum 2002:

1068, and similar terms elsewhere in the literature). The following are two versions of a free

relative in English, first specific, then nonspecific:

(11) a. (specific:) What he says is nonsense.

b. (nonspecific:) Whatever he says is nonsense.

Authors writing in English (e.g. Trotta 2000: 134) sometimes refer to the a. version as

"definite" free relatives and to the b. version as "indefinite" free relatives, but this is

misleading. In fact both versions are indefinite in the sense that they fail to identify any

instantiation for x in the propositional schema

(11)' He says x. x is nonsense.

The default interpretation of (11)a. is 'He is saying something and it is nonsense'; on this

reading, the free relative is interpreted as referring to a uniquely identifiable utterance or at

most to a small, determinate set of utterances that the listener can be expected to be familiar

with (Tredinnick 1995: 254). If -ever is added to the WH word as in (11)b., on the other hand,

the sentence acquires an "exhaustiveness presupposition" (Huddleston / Pullum 2002: 761),

i.e. the set of utterances referred to is interpreted as large and indeterminate, encompassing

any occasion in the present, past or future that the addressee may choose to think of. The

5

Page 6: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

intended effect can also be brought out by paraphrasing NFRs as a series of conditionals with

identical consequents:

(11)'' If he says a, it is nonsense; if he says b, it is nonsense; if he says c, ...

The particle used for this effect in English is -ever (occasionally -soever), corresponding to

immer 'always, ever' and / or auch 'also' in German, to ook 'also' and / or maar 'only, just' in

Dutch, and to similar particles in other languages (Haspelmath / König 1998, Leuschner

2000, 2001; cf. below).

Exactly the same analysis holds for the other construction in which the subordinators in

question are used, viz. (a subtype of) concessive conditionals (henceforth: CCs). CCs are

essentially conditionals with several antecedent values instead of one:

(12) prototypical conditional: if x, then q

concessive conditional: {if a, b, c, ...}, then q

This can be demonstrated by paraphrasing a given CC as a series of if-conditionals just as in

(11)' above, e.g.:

(8)' If he is a, he can't just walk in like this; if he is b, he can't just walk in like this; if he is

c, he can't just walk in like this; if he is d, ...

The difference between CCs and NFRs (at least in prototypical cases) lies in the external

relations they contract: whereas CCs are adjuncts and syntactically part of the periphery

(d'Avis 2004), NFRs (like free relatives generally) are embedded as arguments in the main

clause. At least in English, NFRs can be used in all the usual syntactic functions, i.e. not only

as subject as in (7) and (11)b. but also as direct object (cf. examples later) and, as in (13) and

(14), as indirect object and prepositional object:

(13) This helps whoever is delivering the baby to hold the head gently until the force of the

contraction passes. (LOB)

(14) [...], and she would get in the car and drive to wherever he was working, to take him a

fresh hot meal. (BRO)

6

Page 7: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

CCs, by contrast, are never arguments and may even lack any shared information gap with

the matrix clause. Compare:

(15) (NFR) Whatever she gave him he devoured.

'She gave him x. He devoured x.'

(16) (CC) Whatever she gave him, he grumbled.

'She gave him x; he grumbled.'

In (15), the relative clause as a whole provides the content of the gap in the matrix clause,

hence the free relative, which is governed by the verb devour, is read as denoting food. The

CC in (16) has no such interpretation (Huddleston / Pullum 2002: 763f.).

2.2. Topological Disintegration and Functional Ambiguities

With the above observations in mind, we can describe the relationship between specific free

relatives, nonspecific free relatives and concessive conditionals in terms of three parameters:

syntactic function in the matrix clause, integration into the matrix clause, presence or absence

of nonspecificity marking. According to these parameters, free relatives are prototypically

embedded in the matrix clause and well-integrated with the latter, but distinguished by the

obligatory presence of nonspecificity marking in NFRs. CCs share nonspecificity marking

with NFRs but are adjoined and disintegrated. The intersection of the three parameters in

present-day English is shown schematically in Table 1 below.

Table 1: NFRs and Related Constructions in Present-Day English (schema)

specific free relatives:

nonspecific free relatives:

concessive conditionals:

syntactic relationship withmatrix clause embedding (peripheral)

adjunctionintegration into matrix clause integration disintegration

nonspecificity marking? no yes

7

Page 8: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

The peripheral status of CC protases in the sentence is more manifest in German (and also

Dutch) than in English because this clause-type violates the verb-second constraint, which

says that the finite verb normally occupies the second structural slot in declarative sentences.

This includes cases where the constituent in the first slot is a subordinate clause, as in the

following NFR:

(17) Wer immer einen solchen Vergleich anstelle, [schrieb Schröder an Bush,] habe in

seinem Kabinett keinen Platz. (Deutsche Welle radio)

'Whoever draws such a comparison [wrote Schröder to Bush] has no place in his

cabinet.'

Sentences with an (initial) CC subordinate clause, by contrast, show what Zifonun et al.

(1997: 2322) call "topologische Desintegration", i.e. the protasis fails to occupy the so-called

forefield of the verb:

(18) Was immer wir auch June vorwerfen könnten – Zusammenarbeit mit J.F. Traber

gehört bestimmt nicht dazu! (MK)

'Whatever accusations we could raise against June – cooperation with J.F. Traber is

certainly not one of them!'

Under the verb-second constraint, one would expect the subject to change places with the

verb at the beginning of the main clause; CCs however do not fill the forefield, so there is no

inversion. The first example is from German, the second from Dutch:

(19) a. Was er auch tut – nichts kann unser Projekt retten / *kann nichts unser Projekt

retten.

b. Wat hij ook doet – niets kan ons project redden / *kan niets ons project redden.

'Whatever he does, nothing can save our project.'

8

Page 9: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

Topological disintegration in CCs is usually explained as a manifestation of iconicity, i.e. as

the formal correlate of the fact that the protasis expresses conditions which are irrelevant to

the situation described in the apodosis and that the latter is therefore separately assertible, at

least in standard cases (König / van der Auwera 1988). It is often supported by punctuation,

viz. dashes or colons, indicating that the subordinate clause forms a separate tone group from

the main clause. This tendency towards phonological disintegration can also be observed

when the protasis is sentence-final (in which case the verb-second rule does not apply), and

also in languages like English that are not verb-second in the first place.

Topological disintegration also plays a role in resolving functional ambiguities that

can occur with certain NFRs. The following are NFRs of time and place that function as

arguments and are therefore unambiguously embedded:

(20) Her lamb went wherever Mary went. (Ranger 1998: 140)

(21) [...], and she would get in the car and drive to wherever he was working, to take him a

fresh hot meal. (BRO)

NFRs of time and place may, however, fail to fill an argument slot in the matrix clause, i.e.

function as adjuncts, and this can lead to ambiguities with CCs. Thus, clauses like

(22) We have [...] to [...] eradicate it wherever we find it. (LOB)

may have distinct readings as free relatives and as CCs in English (Quirk et al. 1985: 1102;

cf. Huddleston / Pullum 2002: 764, Ranger 1998: 215):

(22)' (free rel.) 'We have to eradicate it at every place where we find it.'

(22)'' (CC) 'We have to eradicate it, no matter where we find it.'

The difference carries over to integration: the CC reading is likely to be marked by a stronger

intonation break, which may be represented by a comma (Quirk et al. 1985: 1102). If the

clause is preposed, verb-second languages like German and Dutch express the distinction by

means of word order. Here are two German examples (cf. Haspelmath / König 1998: 577f.,

Zaefferer 1987: 283f. fn. 9):

9

Page 10: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(23) (free rel.) Wo immer auch in einer Erzählung das Band des Satzes aufgelöst ist,

empfinden wir die Stelle als lyrisch [...]. (MK)

'Wherever in a narrative the bond of the sentence has been dissolved we

experience the wording as lyrical.'

(24) (CC) Wo immer sich aber auch die Lage und die Erzählung dramatisch zuspitzt,

die Macht der Spannung wird wieder gebrochen [...]. (MK)

'But wherever the situation and the narrative are driven to a dramatic

climax, the power of the tension is broken again.'

Only in (23) does the wo clause share an information gap with the matrix clause:

(23)' The bond of the sentence has been dissolved in a narrative at x; at x we experience the

wording as lyrical.

This difference is reflected in the integration of the clause into the forefield of the verb in the

main clause in (23), whereas the protasis in (24) does not trigger V/2 and remains peripheral

to the main clause. The corresponding distinction may be marked by punctuation in English,

but in actual practice the presence or absence of a comma is anything but a reliable guide.

3. Between Adjunction and Embedding

3.1. English (and Dutch)

Before we tackle the question of whether NFRs and CCs are diachronically related by a

development contradicting the expected directionality from less to more grammatical, i.e. by

antigrammaticalization from embedding to adjunction, let us see first of all whether the cline

(10) suggested by Haspelmath / König (1998) with material from English is a reasonable

proposition in view of present-day evidence, i.e. whether present-day variation in NFRs in

this language seems to support the supposed diachronic development. According to

Haspelmath / König (1998: 577), (a) in (25) shows an embedded NFR in topic position and

subject function, (b) shows the same clause left-dislocated, and (c) can be analysed as the

10

Page 11: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

same clause again but dislocated out of another, non-topic position where it has direct-object

function:1

(25) (a) full embedding: Whatever she writes is brilliant.

(b) adjunction with left-dislocationTP: Whatever she writes, it is brilliant.

(c) adjunction with left-dislocationN-TP: Whatever she writes, producers

want to read it.

And then there is also pattern (d), where there is no shared variable at all and hence also no

(overt) anaphor. Let us call this pattern "full disintegration" for want of a better term:

(25) (d) adjunction with full disintegration: Whatever she writes, producers

queue up for the movie rights.

Not all of these structures are well represented in the corpus, however. Embedded nonspecific

free relatives as in (a) are common, and as already suggested above, they occur in the corpus

as subject, direct object, indirect object and prepositional object. The following is a variant of

(a) with the free relative in direct object function and topicalized:

(26) Whatever Pinturischio has been asked to do to date, he has accomplished [0] in

effortless style [...]. (LOB)

Type (c) too is perfectly common, and so is (d) which, however, we will leave aside for the

time being. But there is only one example of (b), and it does not have it as anaphor but that:

(27) Whatever was the science in the high school course for the time being, that was my

favorite study. (BRO)

These initial observations do not bode well for the antigrammaticalization hypothesis:

although a cline between CCs and nonspecific free relatives can be established at least in

principle in English, only full embedding and adjunction with left-dislocationN-TP, i.e. (a) and

1 I leave aside here the question of where the borderline falls between those versions of the construction that are "still" free relatives and those that are "already" UCCs. See Leuschner (2003: Ch. 4) for detailed discussion.

11

Page 12: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(c), are really common, whereas the alleged transitional pattern, adjunction with left-

dislocationTP (b), is disfavoured.

How could this distribution have come about? According to the directionality

hypothesis (10), any grammaticalization path linking NFRs to CCs would have started from

the fully integrated kind of construction (a) that is current in English today. But this is

disproved by the diachronic evidence: in fact, early attestations of WH-ever type

subordinators and and its predecessor construction (the Old and early Middle English swa

WH swa pattern) fail to show any evidence of such fully embedded free relatives until well

into the Middle English period. Instead, any potential free relatives in topic position are

usually adjoined rather than embedded until at least the 14th century. Here is an example of

this kind of construction, corresponding to the (b) structure of present-day English:

(28) Whoso myghte by þe grace of Godd go þis way, he sulde noghte erre. (Hampole, ca.

1340; Allen 1980: 207)

'Whoever were to follow this path by the grace of God, he would not err.'

And there are also numerous examples of (c), with the anaphor at various distances from the

topic position:

(29) and swa hwæs swa hie rihtlice biddað for ðinum naman & for ðinum gearningum hig

hyt onfoð (Allen 1980: 114)

'and whatever they ask for thy name and thy merit, they receive it'

(30) What se hæfde richedom, he hime makes wræeche mon. (12th. cent.; ibd.: 207)

'Whoeveri had riches, hej made himi a poor man.'

(31) Who so euere honourith me, Y shal glorifie hym. (Wyclif, ca. 1382; Visser 1972: 910)

In other words: whereas (c) co-exists with (a) in present-day English, it co-existed with (b) in

(Old and) Middle English. Not until the 14th century do we begin to find evidence of the step

from adjunction to embedding. Here is one of the earliest fully integrated examples:

12

Page 13: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(32) Who-so seyth hem sothes is sonnest iblamed. (Langland, ca. 1378; Allen 1980: 207)

'He who tells them the truth is the first to be blamed.'

Interestingly, this change is confirmed by parallel data from Middle Dutch. In Middle Dutch,

just as in English, there is no evidence of the embedding pattern (a) before the 14th century.

The following examples show first an instance of (b) and then an early instance of (a), both

from the same 14th century text (Bossuyt 1987: 46):

(33) wie dat dade, hi verboerde .iij .lb.

'whoever were to do that, he would loose 3 pounds'

(34) wiet [= wie dat] dade, verboerde van haringhe .v .s

'whoever were to do that would loose 5s. of herring'

As far as the alleged shift from embedding to adjunction is concerned, the appearance of

antigrammaticalization is thus simply an optical illusion: there was at first no embedding but

only adjunction, involving various kinds of left-dislocation; one of them, (b), developed into

(a), producing a sharper distinction between (a) and the remaining types of adjunction. The

change from (b) to (a) does not, however, go against the general direction of

grammaticalization, on the contrary: it implies an increase in clause-integration, so we

actually have evidence, not of antigrammaticalization, but of genuine grammaticalization

from weaker to greater internal dependency.

4.2. German

In present-day German, we can identify the same cline as in English, though with

characteristic differences. The following are examples of (a) to (c) from the corpus:

13

Page 14: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(a) embedding (~ whatever she writes is brilliant):

(35) Wer immer einen solchen Vergleich anstelle, habe in seinem Kabinett keinen Platz.

(Deutsche Welle radio)

('Whoever draws such a comparison [wrote Schröder to Bush] has no place in his

cabinet.')

(b) adjunction with left-dislocationTP (~ whatever she writes, it is brilliant):

(36) Er ruft seine Märtke mit Namen von Blumen [...], aber was er auch sagt, es ist nur ein

Rauschen. (MK)

'He calls his Märtke by the names of flowers, but whatever he says, it is only a rustle.'

(c) adjunction with left-dislocationN-TP (~ Whatever she writes, producers want to read it):

(37) Die Geschichte lehrt: was immer Menschen Großes hervorgebracht haben, Menschen

haben es alsbald wieder zugrunde gerichtet. (MK)

'History teaches (us): whatever great things Humankind has created, humans have soon

destroyed them again.'

As to the presence of members of the cline in the corpus, the picture for German is similar to

English, except that here it is the embedded type (a) that is rare: in the whole of MK, there

are only 7 embedded NFRs. This means that the formal distinction between different degrees

of disintegration is less sharp in present-day German than in English, and again this does not

bode well for the antigrammaticalization hypothesis, this time because the alleged starting-

point (a), though not completely absent, is conspicuously rare.

This is not the only problem, however. Not only do NFRs seem to prefer

disintegration, free relatives generally (whether specific or nonspecific) are often prevented

from being embedded on syntactic grounds in German. This has to do with the well-known

and cross-linguistically established syntactic phenomenon called "matching" (Groos / van

Riemsdijk 1981; cf. Trotta 2000: 140f.). In German, "matching" is the requirement that the

14

Page 15: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

function of the WH element in the relative clause be identical to the function of the relative

clause in the matrix clause, as required by the matrix verb in terms of surface case (see

Zifonun et al. 1997: 2270-2272, who do not however use the term "matching"). This means

that embedding is unavailable in cases like (41) (ibd.):

(38) Wer solche Vorschläge macht, dem sollte man nicht trauen. (Zifonun et al. 1997:

2271)

'One shouldn't trust anyone who makes such proposals.'

(38') Wer solche Vorschläge macht, *sollte man nicht trauen.

Wer is subject in the relative clause, but the verb in the matrix clause requires a dative, so the

wer clause violates the matching requirement; the solution is to left-dislocate the WH clause

and replace it with an anaphoric pronoun in the required case (dem).2 Analogous examples

are attested from older English:

(39) Whamm se ðu seost ðatt Godes Gast inn aness cullfress heowe of heoffne cumeðð

uppon himm & uppon himm bilefeðð, he fullhtneðð all ðatt fullhtnedd [sic] is. (Allen

1980: 207)

'Whomever you see such that God's spirit comes upon him in the shape of a dove and

remains upon him, he baptizes all that is baptized.'

Constructions like these contribute to making the present-day German pattern similar, to a

certain extent at least, to the situation in Middle English, with evidence of (b) and (c) but

little presence (though not total absence, either) of (a).

As in English, the historical development which led to the present sharp distinction

between embedded and adjoined patterns in German did not involve antigrammaticalization.

In Middle High German, free relatives were invariably of the pattern (b), with or without

there being any requirement for matching (Stolze 1888: 73, Patocka 1998):

2 The conditions under which matching may or may not be relaxed in German do not concern us here. See Zifonun et al. (1997: 2271f.) on "Rektionsgradienz" or case-hierarchies, and Eisenberg (2004: ...) for more recent discussion of the phenomenon with references.

15

Page 16: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(40) Wer daz übel lâzen wil, der muoz alle toetlîche sünde lâzen. (Berthold von Regensburg,

14th century; Roetteken 1884: 57)

'who(ever) wants to renounce evil (he) must renounce all deadly sin'

(41) wer umb rat bittet und rates nicht folgen will, dem ist nit zu raten (Ackermann aus

Böhmen, 15th century; Stolze 1888: 72)

'who(ever) asks for advice without intending to follow it, (him) cannot be given

advice'

Disintegration of the type not forced by matching as shown in (40) more or less disappeared

from the written language in the course of the 16th century in German (Lötscher 1995: 53),

constituting a clear increase in clause-integration. The difference with English lies in the fact

that German retained the matching requirement, as this prevented (b) type disintegration from

disappearing to the same extent as it did in English. Present-day instances of disintegration

are therefore not the result of some historical development which started from embedding,

but on the contrary a residue of the adjunction, forced by matching, as shown in (41) (ibd.).

4. The Diachronic Dissociation of CCs from NFRs

4.1. Preliminaries

Recalling our apparent dilemma between, on the one hand, the typological findings (which

suggest that NFRs are historically primary and CCs secondary) and, on the other hand,

expectations arising under the unidirectionality constraint, we now find that the dilemma has

turned out to be due to a mere optical illusion. In fact, the change did not start from the fully

embedded NFRs that we find (at least to a certain extent) in the present-day languages but

from a situation of adjunction. This yields two significant results. On the one hand, both

typology and unidirectionality have been accommodated. We are therefore justified in

treating the medieval versions of present-day CCs as derived from adjoined free relatives; all

we need to postulate is that the adjunction could be of various degrees of closeness and that

in some peripheral cases, the user required the free-relative construction to express mainly the

adverbial relationship of concessive conditionality. This situation is well represented by

16

Page 17: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

constructions like the following, which have no overt anaphor (or only a very distant one) and

are readily recognizable as the predecessors of present-day CCs:

(42) And to swa hwilcere leode swa we cumað, we cunnon ðære gereord. (Allen 1980: 119)

'And to whichever people we come, we know their language.'

(43) swer sîn (= des Schwertes) hete gegert ze koufen, an der koste was er wol tûsend marke

wert (Nibelungenlied; Kuhlmann 1891: 22)

'whoever would have wanted to buy the sword, as far as the price was concerned it was

certainly equivalent to a thousand marks'

(44) So wie so vallet in scraven forfait ende bi scepenen wert verwonnen; men sal tesinen

hus gaen. (13th cent.; Weerman 1987: 61f.)

'Whoever falls due for payment to the count and is sentenced by sheriffs, he will be

looked for at his house.'

Clearly, more fine-grained historical research is needed to describe the early evolution of the

interclausal relationship up to the point represented in these examples. Such a study would

presuppose a much clearer picture of the overall evolution of free relatives in the languages in

question, encompassing also the usage types mentioned immediately above; unfortunately

such a thorough picture is not yet available on any large scale (though cf. Naganawa 2004 for

a survey of free relatives in one important Old High German writer, Otfrid), and it certainly

cannot be attempted here. It is also clear, on the other hand, that the unidirectionality problem

is likely to be marginal in such a study, and so it makes sense to concentrate here on later

developments, i.e. those that led to the increasing dissociation of NFRs and CCs in the

present-day languages. This will not imply, however, that the directionality issue will be lost

sight of altogether: after all, even with antigrammaticalization out of the way, there is still a

very real possibility that the increasing dissociation of NFRs and CCs could have involved

retraction.

17

Page 18: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

4.2. The WH-'ever' Subordinators

Recall that the relationship between NFRs, CCs and specific free relatives can be described

on the basis of three formal parameters: the syntactic relationship of the subordinate clause

with the matrix clause (embedding vs. adjunction), integration (or otherwise) of the

subordinate clause into the matrix clause, and the presence or absence of nonspecificity

marking. We have so far concentrated on the first two, and to put it briefly, the result is that

the three constructions did not start out from embedding but from adjunction and various

degrees of disintegration; as the adjoined constructions abandoned dislocation to become

fully integrated in English (as they did in German, albeit to a lesser extent), CCs were left

with peripheral adjunction. Let us now focus briefly on the third relevant parameter, viz.

nonspecificity marking, and more particularly on the use of particles inside the subordinate

clause; discussion of a clause-external strategy, viz. the distinction between several types of

pronominal anaphors, will follow further below.

In surveying the use and historical development of nonspecificity-marking through

particles inside the concessive-conditional protasis, we will not concern ourselves with what

motivates the particles individually (English -ever, German immer and auch, separately and

in combination) or how they contribute to the "exhaustiveness presupposition" of the

constructions in which they occur (cf. Leuschner 2003). What interests us here is the

relationship with the earlier 'so WH so' construction and the changes which have made

nonspecificity-marking particles (more or less) obligatory in the new construction. As for

English, this development can be roughly represented as follows (cf. Haspelmath / König

1998: 607):

(45) so WH so > (so) WH so (...) (ever) > WH (so) ever > WH-(so)-ever

But this picture, though superficially correct, is misleading insofar as it could suggest that

present-day nonspecificity marking arose as a direct successor to or replacement for 'so WH

so'. In fact, present-day nonspecificity-marking arose as a strategy for strengthening a

particular interpretation of 'so WH so' clauses, viz. nonspecificity, in appropriate contexts

through particles which were based partly on temporal adverbs with the suitable free-choice

meaning element (English æfre 'ever', German iemer 'always, ever' – cf. Leuschner 1996);

later, as the original 'so WH so' construction declined, the nonspecificity marking particles

were left in place.

18

Page 19: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

More or less the same development took place in German, though with some

important differences. The German version of (45) can be summarized schematically as

follows:

(46) so WH so > so WH (so) > sWH (...) (immer / auch) > WH (...) (immer / auch) > WH

immer / WH (...) auch

One of the differences is that the first 'so' was lost first in English but the second in German;

in German, the first 'so' survived as a prefix s- but was then also lost in the course of Middle

High German. Due to this difference, we still have residues of the second 'so' in present-day

English (viz. occasional compounds with WH-soever, showing how the so tended to get

caught in between the WH word and the increasingly obligatory ever in Middle and Early

Modern English), but no trace at all of the two 'so' in present-day German. Another difference

is that there is only one nonspecificity-marking particle in English, viz. -ever, but several in

German, mainly immer and auch, and that these particles may also occur in combination

(immer auch, auch immer; cf. Leuschner 2000). Earlier forms of the same particles occurred

regularly in classical Middle High German, as in the following example:

(47) diu schamt sich des, swâ iemer wîbes scham geschiht (Walther v. d. Vogelweide; BMZ,

s.v. iemer 3 e)

'he is ashamed of it, wherever dishonour is done to a woman'

In contrast to English, there was no early tendency towards obligatorification, and as a result,

present-day German has no clear equivalent to the well-delineated paradigm of WH-ever

words in present-day English. What German and English do have in common, on the other

hand, is a tendency for the particles to shift to the left inside the clause. Immer has been

affected quite substantially by this trend in German (ibd.), but the same development has

gone much further in English, where nonspecificity-marking ever is now obligatorily

univerbated with the WH word. By contrast, German immer is still syntactically part of the

"middle field" of the WH-clause. Hence we often find WH-clauses which are structurally

parallel to Middle English in older German and very occasionally (as here) also in present-

day German:

19

Page 20: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(48) Die Wahrheit ist, daß der Plakathersteller – welcher Nation er immer sei – nichts

taugt. (Die Zeit; Métrich 1981: 216)

'The truth is that the guy who made the posters – of whatever nationality he be – is

useless.'

All in all, however, cases where immer is in the right-hand part of middle field, behind the

subject, are now extremely rare. Rather, immer is in principle restricted to the extreme left of

the middle field, i.e. to adjacency with the WH word. But it is also clear that it has not as yet

been reanalysed as part of the subordinator position, as shown by the fact that immer is never

univerbated with the WH word (Leuschner 2000).

4.3. Pronominal Anaphors

Not surprisingly, the distinctions and developments in clause-internal nonspecificity marking

which we have just discussed show that the distinction between specific free relatives, NFRs

and CCs is part of a larger diachronic dynamic in the languages in question that cannot be

adequately captured by a too exclusive focus on the question of antigrammaticalization. This

is also confirmed by an aspect of their external syntactic behaviour which differentiates CCs

and NFRs together from other free relatives and has not been mentioned so far: the

distinction between different types of pronominal anaphors, called "DEM-LD" and "PERS-

LD" by Jansen (1980).

With this distinction, we return to the issue of left-dislocation. The use of personal

pronouns in some left-dislocation constructions and demonstrative anaphors in others

corresponds to the traditional distinction between two distinct types of left-dislocation, viz.

left-dislocation proper on the one hand and "hanging topics" on the other (see Altmann 1981

and, for more recent discussion, Frey 2004 and Boeckx / Grohmann 2004), and hence also to

different degrees of functional and syntactic distance between the dislocated element and the

matrix clause (Lötscher 1995: 38). As the name suggests, hanging topics are only very

loosely bound to the next sentence: the hanging topic may form a separate prosodic unit, the

anaphorical pronoun need not be in topic position but may be further inside the clause

(though Frey 2004 argues that left-dislocation proper may have an anaphorical pronoun

inside the matrix clause, too), and in extreme cases there may not be any overt anaphorical

20

Page 21: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

link with the matrix clause at all but only a pragmatic link, the dislocated element being

interpreted as topic for the matrix clause (Lambrecht 1994: 193). In addition, the anaphorical

pronoun used with hanging topics in German may be a personal pronoun (er, sie, es etc.)

rather than a demonstrative one (der, die, das etc.).

Any cursory glance at NFRs and CCs shows that both are much more similar to

hanging topics than to left-dislocation proper. CCs in particular may be heavily disintegrated

prosodically, and they may be linked to a fairly distant anaphor – if any. Crucially for us, the

anaphor itself is almost invariably a personal pronoun in German (and Dutch). Compare:

(36)' Er ruft seine Märtke mit Namen von Blumen [...], aber was er auch sagt, es / *das ist

nur ein Rauschen. Er ist ein Baum.

(37)' Die Geschichte lehrt: was immer Menschen Großes hervorgebracht haben, Menschen

haben es / *das alsbald wieder zugrunde gerichtet. (in MK with es)

'Whatever happens in the world is either a mechanism at the end of a tangle of cross-

references and causalities, or it is carried along by what drives the latter.'

By contrast, specific free relatives obligatorily a demonstrative pronoun as anaphor:

(49) Wer zu spät kommt, den / *ihn bestraft das Leben. (attr. M. Gorbatchev; Lötscher 1995:

41)

'Who(ever) arrives too late is punished by Life.'

At least in German, the only possibility to have a demonstrative anaphor with nonspecific

relatives appears to be when a nonspecific free relative with wer 'who' (and possibly was

'what') is dislocated out of topic position:

(50) Wer auch immer so etwas sagt, er / der muss Beweise haben. (based on Leirbukt 1995:

151; Leirbukt mentions only er)

'Whoever says such a thing (he) must have proof.'

21

Page 22: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

(51) Wer auch immer ihren Boden betritt, er / der fällt unter ihr Gesetz. (based on Schanen

1993: 152 n. 1; Schanen mentions only der)

'Whoever trespasses on their terrain (he) falls under their laws.'

But even then, the two versions are functionally distinct: there is a stronger intonation break

in the version with e- and also a palpable difference in meaning, though this is not easy to

describe. Generally speaking, left-dislocation can be seen as a thematizing or focusing device

and as such as iconic of the succession of two steps in the communicative act, viz. focusing

and predication (Lötscher 1995: 37). But whereas demonstrative anaphors (here: the German

d-series) are associated with "thematische Kohärenz" between the focusing and the

predication, pronominal anaphors (here: the German e-series) are associated with re-focusing,

thematic discontinuity and backgrounding of the referent of the dislocated constituent (ibd.:

38). This is nicely demonstrated by Frey (2004: 217f.), who cites the following minimal pair:

(52) a. (left-dislocation proper:) Maria wird morgen mit Hans nach Paris fahren. *Der

Hans, der ist sehr zerstreut in letzter Zeit.

b. (hanging topic:) Maria wird morgen mit Hans nach Paris fahren. Der

Hans, er ist sehr zerstreut in letzter Zeit.

As Frey points out (ibd.), the demonstrative anaphor in version a. is inappropriate as part of a

move to foreground what is for all means and purposes a new discourse topic; the personal

anaphor in b. is much better. The typical function of sentence-initial CC protases in discourse

consists in precisely the same combination of fore- and backgrounding (Leuschner 2003).

Consider again example (36) above, and compare the following constructed versions:

(36)'' Er ruft seine Märtke mit Namen von Blumen: "Rotklee, Luzerne" und "Honiggras",

aber was er sagt, ist nur ein Rauschen. Er ist ein Baum.

(36)''' Er ruft seine Märtke mit Namen von Blumen: "Rotklee, Luzerne" und "Honiggras",

aber was er sagt, ??das ist nur ein Rauschen. Er ist ein Baum.

22

Page 23: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

Version (36)'' is acceptable: it may lack the dynamism of the original, but we understand the

resumption and the backgrounding. (36)''', by contrast, is odd, because das suggests that the

relative clause presents a new thematic focus that now needs to be carried over into the

matrix clause.

These observations are interesting for two reasons. On the one hand, they suggest a

motivation why embedded nonspecific free relatives are so rare in German: because the

function of the protasis is associated with uses that are better carried out in a dislocated form

(cf. the distinct act of backgrounding), and with personal, not demonstrative, anaphors. On

the other hand, the distinction between the two kinds of anaphor in German also underwent

an interesting historical development which displays clear evidence of what Haspelmath calls

"retraction". At first, e.g. in the 8th-century German of Otfrid, the personal e-series (>

modern er 'he' etc.) clearly predominates over the demonstrative th-series (> modern d-, e.g.

der 'that-one.MASC') in 'so WH so' free relatives (Naganawa 2004). A similar picture holds for

Middle Dutch (Weerman 1987), where the personal h-series (> modern hij 'he' etc.) is used

more frequently than the demonstrative d-series (> modern die 'that one' etc.):

(53) wie dat dade, hi verboerde .iij .lb. (13th cent.; Bossuyt 1987: 46)

'whoever were to do that, he would loose 3 pounds'

(54) Sowie bin der vreide ieme quetst of mesdoet; hi sal de mestat ghelden vierschatte.

(1254; ibd.: 45)

'Whoever wounds or hurts someone in time of peace, (he) must pay fourfold for his

crime.'

(55) So wie die valschede doet in lakenen; die verburd dat laken. (13th cent.; Weerman

1987: 58)

'Whoever commits fraud in the cloth trade, he loses the cloth.'

With respect to variation between personal and demonstrative anaphors, Old High German

and Middle Dutch contrast with Middle High German. In classical Middle High German and

23

Page 24: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

for some time afterwards, the d-series predominates to the near-exclusion of e-anaphors

(Lötscher 1995: 52, cf. also Naganawa 2004):

(56) swaz iu ieman taete, daz waer mir inneclîchen leit (Nibelungenlied; Kuhlmann 1891:

22)

'whatever someone had done to you.PL, it would cause me pain'

(57) swerz hât an sîme lîbe, der sol vil gar wol sîn bewart (ibd.: 20)

'whoever is in bad health (he) will be looked after very well'

(58) Swederz daz ander überziugen mac, daz ez sîne ê an im gebrochen habe, daz ziuhet sich

mit rehte von im. (Berthold von Regensburg; Roetteken 1884: 58)

'Whichever of the two can convince the other that he (i.e. the other) has violated his

honour, (that one) is justified to withdraw.'

The predominance of the d-series represents a strong tendency in Middle High German texts

of all genres, especially during the classical period and after, to preserve topic-initial word

order in the main clause, with a demonstrative pronoun as topic (Lötscher 1995: 52). It

extends, at least to some extent, not only to left-dislocated clauses but also to extrapositional

constructions, i.e. those where the subordinate clause is sentence-final and any pronoun

cataphoric. In such cases not only personal pronouns occurred but also demonstrative one, as

in this examples (cf. also Naganawa 2004):

(59) Den lât er niemer mêre heilen swen er verwundet. (Berthold von Regensburg;

Roetteken 1884: 57)

'(That one) he never allows to get well again whom he hurts.'

As far as I am aware, such extrapositional constructions never occur in older English with

swa WH swa and its successors; in German, too, the topic-first requirement of classical

Middle High German was soon relaxed again. This change clearly qualifies as retraction on

24

Page 25: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

Haspelmath's definition: while the d-anaphors continued to be available, they were no longer

systematically observed or required as part of grammatical rule.

5. Wider Perspectives

Taken together, the changes reviewed above can be seen as a kind of conspiracy which

produced the fairly clear distinction between NFRs, CCs and specific free relatives in

present-day English and (to a lesser extent) German – and hence indirectly also the illusion of

antigrammaticalization. On the one hand, the free relative constructions went from adjunction

to embedding, while on the other hand the nonspecific constructions together (viz. NFRs and

CCs) became gradually more clearly distinguished formally from specific free relatives.

Among the contributing processes were the grammaticalization of nonspecificity-marking

subordinators in English and (to a lesser extent) German, and the emergence of a division of

work between DEM-LD and PERS-LD in German which followed the partial retraction of

the former and the re-emergence of the latter after classical Middle High German.

Apart from the conspiracy itself, another interesting aspect of the historical relationship

between NFRs and CCs turns out to be the embedding of their evolution in larger

grammatical developments of the languages in question. To begin with, the increase in

clause-integration in free relatives in English is paralleled in other relative constructions. For

instance, (37) and (38) are parallel to two examples cited by Hopper / Traugott (2003: 197f.)

from Old and Middle English to illustrate the development of clause-integration in bound (i.e.

non-free) relatives. According to O'Neil (1977), on whose findings their account is based,

such relative clauses "started out essentially as adjuncts, that is, as paratactic clauses close to

the end of the sentence"; from there the path to integration was opened "via a stage of

topicalization which moved certain relative clause structures to the left of the sentence"

(Hopper / Traugott 2003: 197). At this stage, constructions like (60), where the relativized

subject is topicalized, could occur:

(60) Ure ieldran þa þe þas stowa ær hioldon, hie lufodon wisdom. (ca. 880; ibd.: 198)

'Our forebears who previously possessed these places, they loved wisdom.'

25

Page 26: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

From here, full embedding was the next step, and this started taking place substantially in

later Middle English (ibd: 198f.):

(61) Thilke penance that is solempne is in two maneres. (Chaucer, ca. 1390; ibd.: 199)

'Such penance that is ceremonial is of two kinds.'

English relatives thus became more integrated from later Middle English onwards as the

dislocated structure receded in favour of the embedded; and our integrated free relative in

(32) above is of course from the same period as the integrated bound relative in (64). In turn,

relatives of any kind share their increase in clause-integration with other subordinating

constructions, as shown by data from English and its Germanic sister languages. This link is

particularly clear in German, where adverbial clauses developed from loose adjunction to

tighter embedding in the course of the 16th century (Axel 2004) – exactly the period when

the tendency for free relatives to abandon left-dislocation first showed up. Both are thus

likely to be part of the same historical process which led to greater clause-integration and, in

verb-second languages, to a strengthening of the verb-second character (cf. Burridge 1993

with references). Note furthermore that the retention of case-marking in German has also

played a role, working in favour of matching and against consistent clause-integration.

Conversely, the clear syntactic distinction between NFRs and CCs that we have in English

can be ascribed partly to the loss of the need for matching.

Finally, it is worth noting that the insistence on the embedding of an alleged case of

antigrammaticalization in wider changes in the languages in question is well in line with

other recent debates concerning directionality in grammaticalization. A prominent example is

the apparent degrammaticalization of the s-genitive from inflectional marker to clitic in

English and Mainland Scandinavian: cited as a classic violation of unidirectionality by

opponents of grammaticalization for about a decade (including several in Campbell, ed.,

2000), and still acknowledged as a legitimate instance of antigrammaticalization in

Haspelmath (2004: 29 – though cf. the more cautious stance taken by Traugott 2001: 6), this

matter, too, now seems to have been settled against antigrammaticalization. This has been

achieved by showing that the change in question falls out naturally from a wider restructuring

of the case-marking system in both English (Rosenbach 2004) and Mainland Scandinavian

(Askedal 2003). Similarly, the historical development of free relatives can be seen as

embedded in a larger trend which led to the formation, from medieval times onwards, of the

26

Page 27: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

prototype of complex sentence structures in Germanic which is traditionally known as

"hypotaxis" (Leuschner / Van den Nest 2005). This macro-trend involved several subtrends

of which two are relevant here: an increase of clause-integration (which again is in line with a

general strengthening of the verb-second character of the languages concerned) and the

emergence of inventories of semantically well-defined subordinators (Kortmann 1997, De

Groodt in prep.).3 These two trends intersect in NFRs, especially in English, where NFRs are

syntactically well-integrated and distinguished by semantically appropriate subordinators (the

indefinite-nonspecific WH-ever paradigm), and they are also reflected in CCs. The latter

share the same subordinators but make systematic use of syntactic disintegration, a niche

which is functionally well-motivated and became salient thanks to the increase in clause-

integration in the language at large.

REFERENCES

Abraham, Werner (ed.)2004 Focus on Germanic Typology. (Studia Typologica 6.) Berlin: Akademie-Verlag.

Allen, Cynthia L. 1980 Topics in diachronic English syntax. New York: Garland.

Altmann, Hans1981 Formen der ‘Herausstellung’ im Deutschen. Rechtsversetzung, Linksversetzung, Freies Thema

und verwandte Konstruktionen. (Linguistische Arbeiten 106.) Tübingen: Niemeyer.Askedal, John Ole

2003 “Grammaticalization and the historical development of the genitive in Mainland Scandinavian”, in: Barry J. Blake – Kate Burridge (eds.), 21-32.

Axel, Katrin2004 “The syntactic integration of preposed adverbial clause on the German left periphery: a diachronic

perspective”, in: Horst Lohnstein – Susanne Trissler (eds.), 23-58.Belletti, Adriana –Brandi, Luciana –Rizzi, Luigi (eds.)

1981 Theory of markedness in generative grammar: Proceedings of the 1979 GLOW Conference. Pisa: Scuola Normale Superiore.

Blake, Barry J. – Burridge, Kate (eds.)2003 Historical Linguistics 2001: Selected Papers from the 15th International Conference on

Historical Linguistics, Melbourne, 13-17 August 2001. (Current Issues in Linguistic Theory 237.) Amsterdam – Philadelphia: Benjamins.

Boeckx, Cedric – Grohmann, Kleantes K. 2004 “Left dislocation in Germanic”, in: Werner Abraham (ed.), 139-152.

Bossuyt, Alain1987 “Headless relatives in the history of Dutch”, in Willem Koopman – Frederike van der Leek – Olga

Fischer – Roger Eaton (eds.), 33-54.Burridge, Kate

1993 Syntactic change in Germanic: Aspects of language change in Germanic with particular reference to Middle Dutch. (Current Issues in Linguistic Theory 89.) Amsterdam – Philadelphia: Benjamins.

Campbell, Lyle (ed.)2000 Grammaticalization: A critical assessment. Theme issue of Language Sciences 23, 93-340

Cinque, Guglielmo1977 “The movement nature of left dislocation”, Linguistic Inquiry 8: 397-412.

3 The third sub-trend is the grammaticalization of word-order inside the subordinate clause (cf. Burridge 1993 on Dutch), but since this parameter does not contribute to the distinction between NFRs and CCs, we leave it aside here.

27

Page 28: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

De Groodt, Sarahin prep. Die historische Entwicklung adverbialer Subordinatoren aus dem kausal-konditional-konzessiven

Bedeutungsspektrum im Westgermanischen. Ein Beitrag zur Theodistik im Rahmen der funktionalen Sprachtypologie. Ph.D. diss., German Dept., Ghent University.

Ernst, Peter – Patocka, Franz (eds.)1998 Deutsche Sprache in Raum und Zeit. Festschrift für Peter Wiesinger zum 60. Geburtstag. Vienna:

Edition Praesens.Fischer, Olga – Norde, Muriel – Perridon, Harry (eds.)

2004 Up and down the cline: The nature of grammaticalization. (Typological Studies in Language 59.) Amsterdam: Benjamins.

Frey, Werner 2004 “Notes on the syntax and the pragmatics of German Left Dislocation”. In: Horst Lohnstein –

Susanne Trissler (eds.), 203-233Groos, Anneke – van Riemsdijk, Henk

1981 “Matching effects in free relatives: A parameter of core grammar”, in: Adriana Belletti – Luciana Brandi – Luigi Rizzi (eds.), 171-216.

Haspelmath, Martin2004 “On directionality in language change with particular reference to grammaticalization”. In: Olga

Fischer – Muriel Norde – Harry Perridon (eds.), 17-44.Haspelmath, Martin – König, Ekkehard

1998 “Concessive conditionals in the languages of Europe”, in: Johan van der Auwera (ed.), 563-640.Hopper, Paul – Traugott, Elizabeth Closs

2003 Grammaticalization. Second edition. Cambridge: C.U.P.Huddleston, Rodney D. – Pullum, Geoffrey K.

2002 The Cambridge grammar of the English language. Cambridge: C.U.P.Jansen, Frank

1980 “Developments in the Dutch left-dislocation structures and the verb-second constraint”, in: Elizabeth C. Traugott – Rebecca Labrum – Susan Shepherd (eds.), 137-149.

Koopman, Willem – van der Leek, Frederike – Fischer, Olga – Eaton, Roger (eds.)1987 Explanation and linguistic change. (Current Issues in Linguistic Theory 45.) Amsterdam:

Benjamins.Kortmann, Bernd

1997 Adverbial subordination: A typology and history of adverbial subordinators based on European languages. Berlin – New York: Mouton de Gruyter.

Kuhlmann, Hermann 1891 Die Konzessivsätze im Nibelungenliede und in der Gudrun mit Vergleichung der übrigen

mittelhochdeutschen Volksepen. Ph.D. Diss., Kiel.Lambrecht, Knud

1994 Information structure and sentence form: Topic, focus and the mental representations of discourse referents. (Cambridge Studies in Linguistics 71.) Cambridge: C.U.P.

Leuschner, Torsten2000 “‘..., wo immer es mir begegnet, ... – wo es auch sei.’ Zur Distribution von Irrelevanzpartikeln in

Nebensätzen mit w- auch immer”, Deutsche Sprache 28: 342-356.Leuschner, Torsten

2001 “Nebensatzkonnektoren des Typs 'W-Wort + Partikel(n)' (Deutsch wer auch immer usw.) im Germanischen. Eine intragenetische Typologie aus areallinguistischer Sicht”, in: Studia Germanica Gandensia 2001/2, 3-26.

Leuschner, Torsten 2003 Between discourse and syntax: The syntacticization of concessive conditionals in English,

German and Dutch. Ph.D. Diss., Freie Universität Berlin.Leuschner, Torsten – Van den Nest, Daan

2005 “The diachronic formation of hypotaxis as prototype: Processes and consequences in West Germanic”, lecture in the workshop Prototypes and Grammaticalization – Grammaticalization as Prototype? (convened by Tanja Mortelmans and Torsten Leuschner) at New Reflections on Grammaticalization 3, Santiago de Compostela, 18th July, 2005. Abstract available for download at http://www.usc.es/ia303/Gramma3/abstracts.html. DOA 09/2005. To be published in a workshop-based theme issue of Language Sciences, edited by the convenors.

Lohnstein, Horst – Trissler, Susanne (eds.)2004 The syntax and semantics of the left periphery. (Interface Explorations 9.) Berlin – New York:

Mouton de Gruyter.

28

Page 29: Syntactic Functions - Focus on the Periphery€¦  · Web viewNonspecific Free Relatives . and (Anti)grammaticalization . in English and German. Torsten Leuschner. Abstract. One

Lötscher, Andreas 1995 “Herausstellung nach links in diachroner Sicht”, Sprachwissenschaft 20: 32-63.

Marillier, Jean-François (ed.)1993 Satzanfang – Satzende. Syntaktische, semantische und pragmatische Untersuchungen zur

Satzabgrenzung und Extraposition im Deutschen. (Eurogermanistik 3.) Tübingen: Narr.Métrich, René

1981 “Les syntagmes dits concessifs introduits par un élément en w-”, Verbum 4: 191-256, 363-379.Naganawa, Kan

2004 “Zu den verallgemeinernden Relativpronomen bei Otfrid und im Iwein”, Zeitschrift für Sprachwissenschaft 29: 321-344

O'Neil, Wayne 1977 “Clause Adjunction in Old English”, General Linguistics 17: 199-211.

Patocka, Franz 1998 “Zur Linksversetzung im Mittelhochdeutschen”. In: Peter Ernst – Franz Patocka (eds.), 611-621.

Ranger, Graham 1998 Les constructions concessives en anglais: une approche énonciative. (Special issue of

Linguistique.) Paris: Ophrys.Roetteken, Hubert

1884 Der zusammengesetzte Satz bei Berthold von Regensburg. Ein Beitrag zur mittelhochdeutschen Syntax. (Quellen und Forschungen zur Sprach- und Culturgeschichte der germanischen Völker 53.) Straßburg – London: Trübner.

Rosenbach, Anette 2004 “The English s-genitive: A case of degrammaticalization?”, in: Olga Fischer – Muriel Norde –

Harry Perridon (eds.), 73–96.Rothe, Paul

1895 Die Conditionalsätze in Gottfried's von Straßburg Tristan und Isolde. Ph.D. Diss., Halle.Schanen, François

1993 “Funktionen der ‘vor-ersten’ Stellung”, in: Jean-François Marillier (ed.), 145-160.Stolze, Karl

1888 Der zusammengesetzte satz im Ackermann aus Böhmen. Ein beitrag zur mhd. syntax. [sic] Ph.D. Diss., Bonn.

Traugott, Elizabeth C. 2001 “Legitimate counterexamples to unidirectionality”, lecture at Freiburg University, October 17.

Available for download at http://www.stanford.edu/~traugott/papers/Freiburg.Unidirect.pdf. DOA 05/2005.

Traugott, Elizabeth C. – Labrum, Rebecca –Shepherd, Susan (eds.)1980 Papers from the 4th International Conference on Historical Linguistics. (Current Issues in

Linguistic Theory 14.) Amsterdam: Benjamins.Trotta, Joe

2000 WH-clauses in English: Aspects of theory and description. (Language and Computers 34.) Amsterdam – Atlanta: Rodopi.

van der Auwera, Johan2002 “More thoughts on degrammaticalization”, in: Ilse Wischer – Gabriele Diewald (eds.): New

Reflections on Grammaticalization. (Typological Studies in Language 49.) Amsterdam – Philadelphia: Benjamins, 19-29.

van der Auwera, Johan (ed.)1998 Adverbial constructions in the languages of Europe. (Eurotyp 3.) Berlin – New York: Mouton de

Gruyter.Visser, Fredericus Theodorus

1972 An historical syntax of the English language, part 2: Syntactical units with one verb, continued. Leiden: Brill.

Weerman, Fred 1987 “Modern Dutch could be Middle Dutcher than you think (and vice versa)”, in: Willem Koopman –

Frederike van der Leek – Olga Fischer – Roger Eaton (eds.), 55-75.Wischer, Ilse –Diewald, Gabriele (eds.)

2002 New reflections on grammaticalization. (Typological Studies in Language 49.) Amsterdam – Philadelphia: Benjamins.

Zifonun, Gisela – Hoffmann, Ludger – Strecker, Bruno 1997 Grammatik der deutschen Sprache. 3 vols. (Schriften des Instituts für deutsche Sprache 7.1-3.)

Berlin – New York: de Gruyter.

29