15. Variationist Approaches to Syntactic Change.pdf

15. Variationist Approaches to Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:38 PM

15. Variationist Approaches to Syntactic Change

SUSAN PINTZUK

Subject

Linguistics » Historical Linguistics

Key-Topics

syntax , variation

DOI:

10.1111/b.9781405127479.2004.00017.x

The development of modern syntactic frameworks and the growth of research in the field of comparative

syntax have enabled the rigorous investigation of syntactic change. In one sense, diachronic syntax can be

regarded as a form of comparative syntax, where the comparison is between two different stages of the

same language rather than between two different cotemporaneous languages or dialects. In the terminology

of the Principles and Parameters framework (Chomsky and Lasnik 1993), the difference between two stages

of a language can be regarded as a difference in the values of one or more parameter settings; and the goal

of the diachronic syntactician is to explain how and why parameter settings change. 1 I will present evidence

in this chapter to support the hypothesis that parameter settings do not change abruptly, but rather that

change proceeds via competition between two alternative parameter settings during periods of syntactic

variation.

The term “variationist” when describing approaches to syntactic change is best understood as referring to

methodology rather than to a specific framework or a general philosophy. When the systematic syntactic

variation exhibited by languages during periods of change is analyzed quantitatively, generalizations emerge

which enable us to describe the time course of syntactic change, and therefore to begin to understand and

explain how change starts and how it progresses. The most important of these generalizations are the

following three, which will be discussed and illustrated in the remainder of this chapter. It should be

emphasized that these generalizations are not untested hypotheses, but rather empirical results supported

by the analysis of historical data.

First, a distinction must be made between two types of syntactic variation. The first type is controlled by

prosodic constraints and information structure, and frequently involves a simple alternation in constituent

order. This type of variation is diachronically stable, and it does not necessarily lead to or play a direct role

in syntactic change. 2,3 It is commonly found both in modern languages and in the written records of

languages no longer spoken; examples include object shift in the Modern Scandinavian languages (Bobalijk

and Thráinsson 1998; Diesing 1996; Jonas 1996), heavy constituent shift in Old English (Colman 1988;

Pintzuk 1998a, 1998b; Pintzuk and Kroch 1989), and postposition in Early Yiddish and Ancient Greek

(Santorini 1993 and Taylor 1994, respectively). The second type of syntactic variation involves the use by

individual speakers of two distinct grammatical options in areas of grammar that do not ordinarily permit

optionality. This type is diachronically unstable, with the new option competing with the old one and

gradually replacing it. 4 This type of variation has been labeled the double base hypothesis in regard to

variation in underlying (base) structure (Santorini 1992), and more generally grammatical competition (Kroch

1995). It has been found to be characteristic of almost all syntactic changes that have been qualitatively and

quantitatively studied in detail. See section 2 for further discussion, and section 3 for an example of

grammatical competition in Old English.

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747917

Page 1 of 15

15. Variationist Approaches to Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:38 PM

The second generalization is that syntactic change is gradual, and may continue for several hundred years or

more. This observation is of course not new. But when syntactic variation is analyzed as grammatical

competition, our picture of the time course and the nature of syntactic change must be revised. Consider the

change from object-verb (OV) word order to verb-object (VO) word order in the history of English. Many

studies of Old English syntax (e.g., van Kemenade 1987; Koopman 1990; Lightfoot 1991) claim that Old

English was uniformly OV in underlying structure, and that variation in surface word order was a result of

optional movement rules which derived VO order from OV structure: leftward movement of the finite verb

(verb second) and rightward movement of the object from pre-verbal to post-verbal position (postposition).

It is suggested that speakers used these movement rules with increasing frequency over the course of the

Old English period, with the result that VO surface word order reached near categorical status by the

beginning of the Middle English period. Children from that point on acquired a VO grammar, because the

surface word order of the language to which they were exposed was almost entirely VO. According to this

view, the underlying grammar during the Old English period was stable, although there was variation in

surface word order; the grammatical change occurred abruptly at the beginning of the Middle English

period. This picture of the change from OV to VO is challenged in Pintzuk (1996b, 1998a, 1998b), where it is

demonstrated that VO underlying structure was a grammatical option in Old English, and competed with OV

structure during the Old English period and most of the Middle English period. 5 When syntactic variation and

change is understood in this way, we can see that the new grammatical option (in this case VO structure)

does not simply replace the old one (OV structure) at the end of a long period of variation; rather the new

option is acquired and both options are used, with the old option finally lost at the end of the period of

competition. The gradual nature of syntactic change is thus simply a reflex of the gradual nature of

grammatical competition.

The third generalization is that during a period of change, when two linguistic options are in competition,

the frequency of use of the two options may differ across contexts, but the rate of change for each context

is the same. While some contexts may favor the innovating option and show a higher overall rate of use, the

increase in use over time will be the same in all contexts. This generalization was first proposed by Kroch

(1989a) and called the Constant Rate Hypothesis, and is now known as the Constant Rate Effect due to its

overall applicability. 6 It will be discussed in more detail in section 1.

When the syntactic variation found in historical texts is analyzed using quantitative methods based on those

originally developed for sociolinguistic research, we typically find “orderly heterogeneity” (Weinreich et al.

1968): the variation is systematic and the patterns are revealing. When the distributions of forms are

analyzed in detail, either during a single historical stage of a language or over a longer period of time, the

results can provide support for the choice of one grammatical analysis over the other (Pintzuk 1999; Taylor

1994), permit the tracking of syntactic variation and change over time (Kroch 1989a; Pintzuk 1996a, 1999;

Santorini 1993), uncover dialect differences (Haeberli 2000; Kroch and Taylor 2000), and lead to insights

into the nature and organization of the grammar (Kroch 1989a; Pintzuk 1998b). The quantitative methods

used to analyze the variation range from the simple examination and comparison of distribution frequencies

to the statistically more complex variable rule analysis (see Cedergren and Sankoff 1974; Guy, this volume;

Sankoff 1988; among many others). Some of these methods will be illustrated in section 3.

Although the variationist approach to syntactic change is not by necessity tied to any particular grammatical

framework, most researchers who use the methodology are generative syntacticians who are in accord with

the assumptions of the grammatical approach to syntactic change (see Lightfoot, this volume, and the

references cited there): they assume a rich, highly structured Universal Grammar, consisting of invariant

principles that hold of the grammars of all languages and parameters that are set by triggers in the

language learner's linguistic environment. And they share the view that language change and language

acquisition are intimately connected, and that there can be no separate theory of language change. Much of

the research discussed in this chapter was carried out in a Principles and Parameters framework, but this is

not, of course, a requirement of the variationist approach to diachronic syntax. The changes that are

investigated involve phenomena that distinguish modern languages from each other (e.g., the order of verbs

and their complements, the behavior of clitics, the verb-second constraint) and therefore can be expressed

in any syntactic framework, including Principles and Parameters, Minimalism, Head-Driven Phrase Structure

Grammar, Lexical Functional Grammar, and Construction Grammar.

1 The Time Course of Linguistic Change

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747917

Page 2 of 15

15. Variationist Approaches to Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:38 PM

Suppose that, within a group of historical texts with a range of dates of composition, we can identify one

particular linguistic change that we want to study, in which a new form alternates with and eventually

replaces an older form in a variety of linguistic contexts. For each text, we can count the number of times

each of the two forms appears in each context. We can then plot the frequency of the new form 7 against the

dates of the texts and examine the time course of the change. Many investigators (Altmann et al. 1983;

Bailey 1973; Kroch 1989a; Osgood and Sebeok 1954; Weinreich et al. 1968; among others) have suggested

that this type of change - the gradual replacement of one form by another in the language of speakers over

time, perhaps over the course of many generations - follows an S-shaped curve, as shown in figure 15.1 .

The replacement of old forms by new ones occurs slowly at the beginning of the period of change, then

accelerates in the middle stage, and finally, at the end of the period, when the old form is rare, tails off until

the change reaches completion.

Figure 15.1 S-shaped curve of linguistic change

Both Altmann et al. (1983) and Kroch (1989a) propose that a specific mathematical function, the logistic,

underlies the S-shaped curve which represents the usage of speakers over time. The importance of selecting

a specific function is that statistical techniques can be used to fit a particular set of data to the function and

estimate its parameters, as will be described in detail below. When parameters for different datasets are

estimated in this way, they can be compared, and the results of the comparisons can be used to draw

conclusions about the change under investigation.

The equation of the logistic curve is given in (1) below. In this equation, p is the frequency of the new form,

and varies between 0 and 1, that is, between 0 percent and 100 percent. t is a variable representing time,

and s and k are constants - that is, they are parameters that are fixed (perhaps differently) for each

particular instance of an S-shaped curve:

(1)

An equivalent form of equation (1) is shown in (2). The left-hand side of (2) is called the logistic transform

of the frequency, or logit:

(2)

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747917

Page 3 of 15

15. Variationist Approaches to Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:38 PM

While the logistic of equation (1) is an S-shaped curve, the logit of equation (2) is a straight line, a linear

function of time. 8 s is the slope of the line; k is the y-intercept, and is related to the frequency of the new

form at some fixed point in time, t = 0. Of course, as Kroch and others have pointed out, the logistic model

is an idealization of linguistic change, because there is no time t for which the frequency p of the new form

equals either 0 or 1 in these equations. In other words, the model can only approximate the process of

change at both the beginning, called the point of actuation, where the frequency of the new form jumps from

0 to some small positive value, and at the end, when the old form dies out and the frequency of the new

form jumps from some high value to 1.

Figure 15.2 Bailey's model of linguistic change

Let us now consider how changes begin and how they spread. A concrete example to keep in mind is the

use of auxiliary do in Middle and Early Modern English (Ellegård 1953; Kroch 1989a; Warner 1998), within

three different syntactic contexts: negative declarative clauses, affirmative declarative clauses, and

affirmative questions. In principle, change may be actuated in one of several different ways. Speakers may

start to use the new form simultaneously in all contexts, either at the same initial frequency or at different

initial frequencies; this is simultaneous actuation. For example, auxiliary do may appear for the first time in

all three types of clauses in several texts composed during the same decade, with the same initial low

frequency in all clause types. Or speakers may start to use the new form sequentially, first in the most

favoring context and only subsequently in less favoring contexts; this is sequential actuation. Again, the

initial frequencies may be either the same or different. Once actuation has occurred, the change may in

principle spread in two different ways: either at different rates in different contexts, or at the same rate in

each context. For example, speakers’ use of auxiliary do may increase in frequency more rapidly in negative

clauses than in affirmative declarative clauses and questions. Bailey (1973), for example, claimed that

actuation occurs sequentially, with change spreading more quickly in the most favoring context, less quickly

in the less favoring contexts. This model is illustrated in figure 15.2 , where the three straight lines represent

three plottings of the logit over time, one for each of three different linguistic contexts. Notice that these

three lines all have different slopes and different y-intercepts (the s and k parameters in equations (1) and

(2)), as is clear from the fact that they rise at different rates and will intercept the y-axis at different points.

In contrast to Bailey, Kroch (1989a) proposed the Constant Rate Effect (CRE): while the frequency of use of

competing linguistic forms may differ across contexts at each point in time during the course of the change,

the rate of change for each context is the same. Kroch's model is illustrated above in figure 15.3 . Notice that

these three lines have different y-intercepts (the k parameter), but they all have the same slope (the s

parameter); in other words, they are parallel.

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747917

Page 4 of 15

15. Variationist Approaches to Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:38 PM

Figure 15.3 Kroch's model of linguistic change

It should be pointed out that the word “constant” in the Constant Rate Effect does not refer to a constant

rate of increase in p , the frequency of the new form. As stated above, and as can be seen from the S-shaped

curve in figure 15.1 , which plots the frequency p of the new form over time, the frequency increases slowly

at first, then the rate of increase accelerates, and then it finally tails off. What is “constant” in the Constant

Rate Effect is that the change is the same across linguistic contexts, so that the frequency of the new form

changes in the same way in all contexts. In terms of equations (1) and (2), the parameter k may be different

for each context, but the parameter s is constant for all contexts, as shown by the identical slopes of the

straight lines in figure 15.3 . As Kroch (1989a: 199) states, “Contexts change together because they are

merely surface manifestations of a single underlying change in grammar. Differences in frequency of use of a

new form across contexts reflect functional and stylistic factors, which are constant across time and

independent of grammar.”

Kroch (1989a) presents four cases of linguistic change that have been studied quantitatively - the

replacement of have by have got in British English from 1750 to 1934 (Noble 1985), the rise of the definite

article in Portuguese possessive noun phrases from the fifteenth through the twentieth century (Oliveira e

Silva 1982), the loss of verb second in Middle French (Fontaine 1985; Priestley 1955), and the rise of

auxiliary do in English between 1400 and 1700 (Ellegård 1953) - and shows that all four provide strong

support for the CRE. Additional research has demonstrated that the CRE holds for the replacement of I-final

structure by I-medial structure in the history of English and Yiddish (Pintzuk 1996a and Santorini 1993,

respectively) and the change from OV to VO in the history of Greek (Taylor 1994).

Notice that the grammatical analysis which underlies both the change and the quantitative patterns may be

quite abstract. For example, Kroch (1989a) builds on and extends the work of Adams (1987a, 1987b) for

Middle French to show that three very different surface changes - the loss of subject-verb inversion, the

loss of null subjects, and the rise of left dislocation structures - can all be analyzed as reflexes of the same

underlying grammatical change, the loss of the verb second constraint. These three surface phenomena are

the three different contexts in which variation between options is exhibited. The CRE predicts that all three

surface alternations will proceed at the same rate during the period from 1400 to 1700, as indeed Kroch

(1989a) demonstrates. Similarly, Taylor (1994) shows that in three periods of Classical Greek, the

distribution of clitics and weak pronouns produces the same measure of verb-medial versus verb-final

clause structure as an independent estimate of that ratio derived from the distribution of NP and PP

complements and the rates of postposition.

Conversely, in other cases of syntactic variation and change, identical surface forms may be derived by

different grammatical processes in different contexts. In these cases the CRE is irrelevant, since it holds only

for contexts in which the surface forms are reflexes of the same underlying grammatical alternation. In fact,

it is generally true for these cases that change will proceed at different rates in the different contexts, since

it is unlikely that two separate and unrelated grammatical alternations will advance at the same rate. For

example, HirschbÜhler and Labelle (1994) show that the change from ne infinitival-verb pas to ne pas

infinitival-verb in the history of French affected lexical verbs, modals, and auxiliaries at different times, and

that it proceeded at different rates for the three verb types. These findings thus seem to present a

counterexample to the CRE. However, HirschbÜhler and Labelle use structural evidence to demonstrate that

what appears to be a single grammatical change in three different contexts actually represents two separate

changes: a change in the position of pas , and a loss of verb movement to T. It is only to be expected that

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747917

Page 5 of 15

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: