16. Cross-Linguistic Perspectives on Syntactic Change.pdf

16. Cross-Linguistic Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:39 PM

16. Cross-Linguistic Perspectives on Syntactic Change

ALICE C. HARRIS

Subject

Linguistics » Historical Linguistics

Key-Topics

syntax

DOI:

10.1111/b.9781405127479.2004.00018.x

In this chapter I seek to characterize briefly an approach to universals of diachronic syntax that depends

crucially upon a rich cross-linguistic corpus. 1 I do this by stating some of the aims of the study (section 1),

briefly describing differences between the approach taken here and others (section 2), and describing the

method followed (section 3). Section 4 sets out an example from Georgian, which helps with the

characterization of reanalysis and actualization (section 5). Section 6 is devoted to an extended example,

which illustrates the application of this method in one area of syntax, the change from biclausal to

monoclausal structure.

1 Goals of the Study of Syntactic Change

Different approaches to language can be distinguished in part in terms of their goals. Among my goals in

studying syntactic change are the following:

i to characterize syntactic change accurately;

ii to identify and characterize universals of syntactic change;

iii to explain syntactic change;

iv to build a theory of change.

Characterizing syntactic change includes a consideration of general questions, such as: Is syntactic change

regular? Is it directional? It includes a description of the mechanisms of syntactic change. In addition, a

complete characterization comprises a description of the changes that actually occur in natural languages.

HC applies inductive methods in searching for universals of change, seeking the general rule on the basis of

specific cases. Examining instances of the “same” change in diverse languages, we can focus on elements of

that change that are the same or similar and eliminate from consideration elements that vary from one

language to another. This method is described in greater detail in section 3.

There are many kinds of explanation, and among the most effective is demonstration of a relationship

between the familiar and that which is (or was) unfamiliar. It may be that the better way to view goal (iii) is

that we seek simply to understand all aspects of syntactic change, and we include here attention to the

causes of change. 2

It may be too early to state a general theory of change, but it is at least possible in our current state of

understanding to distinguish language-particular from universal aspects of syntactic change, to state

generalizations about classes of change, and to identify the kinds of syntactic change that are possible in

natural language and, at least by implication, the kinds that are not.

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747918

Page 1 of 19

16. Cross-Linguistic Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:39 PM

natural language and, at least by implication, the kinds that are not.

2 Approaches to the Characterization and Explanation of Syntactic Change

Logically there are two basic ways to approach the study of change. One approach begins with a theory and

examines what that theory tells us about what we should expect to find in language change. The best-

known theory-driven approach of this sort is found in Lightfoot (1979, 1991). What have we learned from

these studies? The centerpiece of Lightfoot (1979) was the Transparency Principle, “a rather imprecise,

intuitive idea about limits on a child's ability to abduce complex grammars” (Lightfoot 1981: 358), which was

proposed to characterize the point at which change takes place. In Lightfoot (1991) we find the notion of

“degree-Ø learnability,” likewise a hypothesis about the way a child learns language. But the hypothesis of

imperfect learning cannot account for all syntactic change, since many diverse languages retain the source

construction beside the reanalyzed structure (see Harris and Campbell 1996). And if we are looking for a set

of general principles that limit syntactic change or statements of universals of syntactic change, we come

away from Lightfoot (1979, 1991) empty-handed. Lightfoot adroitly avoids commiting to anything of

substance by arguing that there are no constraints on change other than the theory of Universal Grammar.

Naturally, Universal Grammar sets upper limits on change, but the doctrine that it adequately characterizes

syntactic change would imply, for example, that it is possible for any sanctioned construction to become any

other, without limit. Even if there were no limits on syntactic change other than those imposed by Universal

Grammar, one could still state valid generalizations. Yet we come away from these studies with no

universals, with no constraints, with no hypotheses that can be tested. It is my position that the theory of

Universal Grammar is as yet incompletely stated, and that studies of universal properties of syntactic change

will contribute significantly to developing it further. 3

A second approach is data-driven and seeks to develop generalizations based on the corpus of actual

changes. Many studies of the diachronic syntax of individual languages or individual families, including my

own studies (e.g., 1985, 1991, 1994, 1995), are intended in this spirit as contributions to the general

corpus. There is a wealth of data available on attested changes (i.e., changes during the historical period) in

some languages of the Indo-European, Semitic, Uralic, Kartvelian, Dravidian, and Sino-Tibetan families,

among others. Even some languages outside these families are attested for a long enough period that

syntactic change can be carefully tracked; for example, a comparison of Classical Nahuatl with those of the

modern Nahuatl dialects known to be its descendants provides attestation of change. Among the problems

for the would-be constructor of theories is that it requires knowledge of the language to analyze the texts,

to read much of the secondary literature, and to avoid the pitfalls of incomplete understanding of the

synchronic systems.

Among data-driven approaches, some limit themselves to particular aspects of change. For example, while

we have learned a great deal from recent work in grammaticalization, that approach is primarily centered on

features of words and morphemes and on the transition from the former to the latter. For example, in a

grammaticalization approach to the case study in section 4 below, emphasis would be on the transition from

verb to auxiliary; in contrast, it is my aim to treat the structural change involved, as well as the verb-to-aux

transition. Similarly, recent functionalist studies contribute to our understanding of certain types of change,

but they provide no general characteristics of change. Finally, several important recent papers have provided

valuable studies of the gradual implementation of particular changes in syntax (e.g., Kroch 1989a; Fischer

and van der Leek 1987; Naro 1981; Naro and Lemle 1976). What these studies do not identify clearly is the

mechanism that gets these processes started and constraints on that mechanism. 4 In our work, we seek to

provide an overall framework in which contributions to various individual aspects of syntactic change can be

correlated with others to make sense of diachronic syntax.

3 Cross-Linguistic Perspective

The method of cross-linguistic comparison developed in HC is part of an overall framework for the

description and explanation of data from a wide variety of languages, and on this basis we develop a theory

of morphosyntactic change. The method begins by comparing the “same” changes in very different

languages. Characteristics that are found in language after language are candidates for universals, and from

them we develop hypotheses which can then be tested against additional data. We reason that

characteristics that do not occur in all instances must be language-specific. This method begins by

comparing changes that are as closely matched as possible for input structure, output structure, and

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747918

Page 2 of 19

16. Cross-Linguistic Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:39 PM

comparing changes that are as closely matched as possible for input structure, output structure, and

meaning-function. Generalizations are made at this level, and the comparison proceeds to a higher level,

where the previous set of changes are compared with other sets which have already been compared

internally. Again generalizations are drawn, and, if appropriate, the comparison continues, throwing an ever-

wider net.

An example which is elaborated below in section 6 is the simplification of biclausal structures into

monoclausal ones. The changes from independent modal verbs to modal auxiliaries in different languages

are compared, and conclusions are drawn. At a second level, these results are compared with conclusions

based on comparisons of similar changes involving other independent verbs, such as ‘have, hold, keep’ used

in perfects, and ‘be’ in progressives. At a third level, other transitions from biclausal to monoclausal

structure (i.e., not involving creation of auxiliaries) are compared with the results previously obtained.

A second example, not included below, involves the operation of word order change. We found typological

approaches limiting, and we look instead at the changes that actually occur. At a first level we compare

verbs and auxiliaries that are not adjacent in the input to the change and are adjacent after the change. At a

second level we compare with these results other changes from discontinuous to continuous constituency. A

further level adds comparison of changes of the relative positions of head and dependent (HC, 195–238). 5

Cross-linguistic comparison is set within a theory that recognizes only three mechanisms of syntactic

change: reanalysis, extension, and borrowing. Other phenomena that might be described by others as

mechanisms of change are, in our view, usually a specific instance or type of one of these. Reanalysis is a

mechanism which changes the underlying structure of a syntactic pattern and which does not involve any

modification of its surface manifestation. 6 We understand underlying structure in this sense to include at

least (i) constituency, (ii) hierarchical structure, (iii) category labels, and (iv) grammatical relations. Surface

manifestation includes morphological marking, such as morphological case, agreement, and gender-class.

Extension is a mechanism which results in changes in the surface manifestation of a pattern and which does

not involve immediate or intrinsic modification of underlying structure. Borrowing is a mechanism of change

in which a replication of the syntactic pattern is incorporated into the borrowing language through the

influence of a host pattern found in a contact language.

Other aspects of this theory that are essential for a complete understanding of cross-linguistic comparison

are (i) that syntactic change is regular, in the sense that it is rule-governed and not random (HC, 325–30),

(ii) that research to date has not shown any kind of syntactic change to be absolutely unidirectional, though

many changes are known to proceed usually in one direction (HC, 330–43), and (iii) that reanalysis depends

upon the possibility of multiple analysis 7 (HC, 81–9).

4 An Example: Georgian unda

This example from Georgian will be used in sections 5 and 7 to make several points. As will be immediately

clear, the change is very similar to one in English.

In the historical period the Georgian modal ‘want’ has become an auxiliary, expressing a range of

modalities, including necessity, intention, and obligation. 8

Old Georgian hnebavs ‘wants’ was an independent verb that could have a nominal object or a sentential

object expressed in the subjunctive, in the aorist, or in any of several non-finite forms. Examples with the

subjunctive are given in (1) and (2):

(1)

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747918

Page 3 of 19

16. Cross-Linguistic Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:39 PM

(2)

The present tense forms of this verb in Old Georgian included m-nebavs ‘I want it,’ g-nebavs ‘you want it,’

h-nebavs ‘s/he wants it,’ and the imperfect (past) mi-nda ‘I wanted it,’ gi-nda ‘you wanted it,’ u-nda ‘s/he

wanted it.’ The imperfect forms came to be used for the present tense by the eleventh or twelfth century,

and a new imperfect was created: mi-ndoda, gi-ndoda, u-ndoda (Sar ǰ vela[]e 1984: 412–13). Thus, the forms

used today for the present tense of ‘want’ are so-called past-presents. The verb ‘want’ was and is one of a

number of verbs, traditionally called inversion verbs , which govern a syntactic pattern in which the

experiencer (here, the one who wants) is in the dative and conditions indirect object agreement, and the

stimulus (here, that which is wanted) is in the so-called nominative case and conditions subject agreement.

Thus, the m-, g-, h - and the mi-, gi-, u - prefixes isolated in the forms above in general mark indirect

object agreement, in this instance agreement with the experiencer.

(1) and (2) illustrate the pattern in (3), which occurred in both Old and Middle Georgian (from the twelfth

century):

(3)

The form unda cited in (3) was imperfect tense in Old Georgian, but later present. (1) and (2) show that in

Old Georgian the initial subject of unda (the experiencer, mas in (3)) did not have to be coindexed

(coreferential) with an argument of the verb of the subordinate clause; when the initial subject of the matrix

was coindexed with an argument in the subordinate clause, the latter was generally omitted.

The biclausal structure represented in (3) was reanalyzed as the monoclausal pattern in (4):

(4)

At the same time, the meaning of unda changed from ‘want’ to a range of modalities including epistemic

necessity and deontic obligation. In this innovative usage it ceased to be conjugated, but exists in this single

form (derived from and identical to the third person experiencer, third person singular stimulus form). The

original biclausal construction continues to exist side by side with the new, maintaining its original meaning,

its original structure (with an optional complementizer and with the option of a non-coreferential subject in

the complement clause), and its original complete paradigmatic variation (with the tense adjustment from

imperfect to present described above), as illustrated in part in (5a), (6a), and (7a) below. Thus in the modern

language there is an invariant auxiliary unda ‘should, ought, must’ beside a third person lexical verb form

unda ‘s/he wants it,’ which still alternates paradigmatically with minda ‘I want it,’ ginda ‘you want it,’ and

the plural forms. 11 The independent verb ‘want’ is illustrated below in (5a), (6a), and (7a), and the derived

auxiliary in (5b), (6b), and (7b):

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747918

Page 4 of 19

16. Cross-Linguistic Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:39 PM

(5)

(6)

(7)

Table 16.1Two case patterns in the subjunctive

Morphological class Subject case Direct object case

Narrative

Nominative

Following reanalysis as a monoclausal structure, other changes in the clausal syntax were also made. In (4),

after reanalysis, the initial subject is still expressed in the dative, the case required by the syntax of the old

verb ‘want’ as an inversion verb; later the pattern required by the (main) verb was extended to this

construction. In Georgian, case marking of subjects varies according to the morphological category of the

(main) verb. The patterns in table 16.1 illustrate those found in (5b), (6b), and (7b). 12 Class 1 generally

contains transitive verbs, and class 2 a subset of intransitives.

These patterns are required in simple sentences, as illustrated in (8):

(8)

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747918

Page 5 of 19

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: