1. The Comparative Method.pdf

1. The Comparative Method : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:27 PM

1. The Comparative Method

ROBERT L. RANKIN

Subject

Linguistics » Historical Linguistics

Key-Topics

comparative , methods

DOI:

10.1111/b.9781405127479.2004.00003.x

The comparative method is a set of techniques, developed over more than a century and a half, that permits

us to recover linguistic constructs of earlier, usually unattested, stages in a family of related languages. The

recovered ancestral elements may be phonological, morphological, syntactic, lexical, semantic, etc., and may

be units in the system (phonemes, morphemes, words, etc.), or they may possibly be rules, constraints,

conditions, or the like, depending on the model of grammar adopted. The techniques involve comparison of

cognate material from two or more related languages. Systematic comparison yields sets of regularly

corresponding forms from which an antecedent form can often be deduced and its place in the proto-

linguistic system determined. In practice this has nearly always involved beginning with cognate basic

vocabulary, extraction of recurring sound correspondences, and reconstruction of a proto-phonological

system and partial lexicon. 1

1 The Goal of the Comparative Method

Kaufman (1990: 14–15) states: “The central job of comparative-historical linguistics is the identification of

groups of genetically related languages … [and] the reconstruction of their ancestors.” He continues (p. 31):

“it should be clear that while archeology, genetics and comparative ethnology will help flesh out and provide

some shading in the picture of pre-Columbian … Man, it is comparative linguistic study, combined with

some of the results of cross-cultural study, that will supply the bones, sinews, muscles, and mind of our

reconstructed model of early folk and their ways.” Linguistic reconstruction is one of our primary tools for

learning about the prehistoric past. In many ways it is our best, and this is especially true at time depths

where archeology has trouble identifying the ethnicity of its subject matter. Archeology is our best tool for

recovering material culture – settlement patterns, dwelling types, tools, subsistence, and related information

– but it contributes much less to our understanding of what archeologists call ideoculture and socioculture . 2

These are areas in which linguistic reconstruction is potentially much more productive. The comparative

method is our primary tool for arriving at such linguistic reconstructions.

While the principal goal of most linguists who are also historians has been to learn as much as possible

about earlier languages and about past cultures through their languages, other branches of linguistics have

benefited a great deal from the by-products of comparative work. Many who are philosophically synchronic

linguists have looked to comparativists to inform them about the possible types and trajectories of language

change. The study of attested and posited/reconstructed sound changes has played an important role in the

formulation of notions of naturalness in phonological theory, and modern theories of markedness and

optimality often rely, implicitly if not explicitly, on historical and comparative work. The same can be said

for the establishment of the grammaticalization clines that result from much morphosyntactic change. 3 Our

understanding of the complexities of the synchronic polysemy often associated with grammaticalization is

informed by the study of attested and posited intermediate steps in their histories. To a lesser extent the

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g97814051274793

Page 1 of 20

1. The Comparative Method : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:27 PM

same may be said of semantics and semantic change. But such essentially typological studies may not be

considered by some historical linguists to be one of the goals of the comparative method per se. They are

important bonuses that result from a consistent and thorough application of the method to families of

languages, but they will not receive much additional coverage in this chapter.

2 Why Does the Method Work?

The comparative method relies on certain characteristics of language and language change in order to work.

One important factor is, of course, the arbitrariness of the relationship between phonological form and

meaning (non-iconicity). To the extent that the linguistic sign is arbitrary, sound change can operate

unhindered and will normally be rule governed. Where iconicity is present (in sound symbolism, nursery

terms, onomatopoeia) normal change may be impeded or prevented. 4 Linguists therefore avoid comparison

of such items until the basic correspondences among the languages being compared are understood.

A second factor is the regularity of sound change. 5 To the extent that sound change is regular, we can, with

the help of phonetics and an understanding of sound change typology, work backward from more recent to

earlier stages. And indeed most phonological change ends up being change of articulatory habit, that is, rule

change, and thus ultimately regular. Fairly salient interference is required in order to breach such regularity.

Recognition of regularity and of the role it plays in reconstruction has been considered both a strength and a

weakness of Neogrammarian linguistics. It has most often been considered a strength because, of course,

without ultimate regularity there can be no phonological reconstruction. It has sometimes been considered a

weakness of the Neogrammarian position, however. Beginning with Hugo Schuchardt (1885) and continuing

until the present, analogical extension of changes and the pervasive role of dialect borrowing with resultant

diffusion of forms has occupied many linguists, dialectologists, and creolists. 6 Copious amounts of ink have

been spilled in discussions of the extent to which the Neogrammarian “hypothesis” is really “true.” But, as

most Indo-Europeanists have always known, the exceptionlessness of sound change was not so much a

hypothesis for Neogrammarians as it was a definition. Those changes that were sweeping and observed after

several centuries to be essentially exceptionless qualified for the term Lautgesetz (sound law), while changes

that seemed to affect only particular words or groups of words did not so qualify. 7

Most linguists believe that change in articulation begins as a geographically and/or socially limited but

regular, unconscious, and purely phonetic process, which then spreads by several different mechanisms,

including dialect borrowing (social and otherwise) and rule formation during the language acquisition period

in children, until regularity over a greater area is achieved. A perceived dichotomy in the methods of

diffusion has variously been described as sound change versus borrowing and analogy (the terms

traditionally favored by most comparativists), primary versus secondary sound change (Sturtevant 1917: chs

2 and 3), actuation versus implementation (Chen and Wang 1975), and others, although the pairs of terms

do not always correspond 100 percent. The precise extent to which ultimate regularity results from, or is

independent of, dialect borrowing doubtless varies from language family to language family. 8 As a practical

matter, comparative linguistics generally involves compilation and analysis of the reflexes of sound changes

that occurred, diffused, and regularized long ago. Within comparative Indo-European linguistics the problem

of variability within sets of reflexes has not been acute. Whatever the mechanisms that contribute to

ultimate regularity in particular instances, its existence, although sometimes obscured by diffusion and

analogy, is not seriously disputed and is of primary importance for operation of the comparative method.

3 Family Tree and Wave Diagrams of Language Relationship

The comparative method was developed for the study of the well-defined and quite distinct linguistic

subgroups of Indo-European, so comparanda there have tended to be similarly well defined. Obviously such

definition is not always possible (and some might argue that it seldom is). Clearly there are language

families (e.g., northern Athabaskan, Muskogean, some Austronesian) in which some unique subgroups are

difficult to specify with clarity. 9 This has given rise to another red herring frequently encountered in

discussions of the comparative method, namely the assumption that it must be based on some inflexible

notion of Stammbaumtheorie . And here again much ink has been spilled by amateurs wondering which

theory, the family tree ( Stammbaum ) or the supposedly competing wave theory ( Wellentheorie ), is “true.” 10

Both are true. But they are oversimplified graphic representations of different and very complex things, and

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g97814051274793

Page 2 of 20

1. The Comparative Method : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:27 PM

Both are true. But they are oversimplified graphic representations of different and very complex things, and

it seems hyperbole to call them theories in the first place. One emphasizes temporal development and

arrangement, the other contact and spatial arrangement, and each attempts to summarize on a single page

either a stack of comparative grammars or a stack of dialect atlases. Neither is a substitute for a good

understanding by the linguist of both the grammars and the historical, social, and geographical

interrelationships found among his or her target languages. The comparative study of languages or dialects

that are arranged in chains or other adjacent or overlapping continua is certainly a challenge, but it is a

challenge to the linguist rather than to the method. 11

4 Uniformitarianism

Lastly, the method also relies on the more general scientific notion of uniformitarianism , here the

understanding that basic mechanisms of linguistic change in the past (e.g., phonetic change, reanalysis,

extension, etc.) were not substantially different from those observable in the present. Most linguists operate

with this as a given and it has not received detailed treatment in most studies of language change, but

without the assumption of uniformitarianism, reconstruction would not be possible (Allen 1994: 637–8). 12

5 Steps in Application of the Comparative Method

The comparative method proceeds in several recognizable stages, which in practice overlap considerably.

Internal reconstruction is useful when applied to the daughter languages initially and may also be practiced

at various points along the way (see Ringe, this volume). There is relatively little in the way of strict ordering

of procedures. A relatively full comparative treatment of a family of languages would include most or all of

the following, beginning with the discovery of cognates, both lexical and morphological, and concomitant

confirmation of genetic relationship. 13 Most of these topics are discussed below.

i Phonological reconstruction:

a Extraction of phonological correspondence sets.

b Classification of sets by articulation (place/manner).

c Preliminary reconstruction of proto-phonemes.

d Distributional analysis of proto-phonemes; collapse of complementary sets.

e Assignment of phonological/phonetic features to proto-phonemes (the reality debate).

f Possible adjustment of reconstructions in line with typological considerations (in Indo-European,

issues such as laryngeal theory and, more recently, glottalic theory).

ii Reconstruction of vocabulary per se:

a Reconstruction of structured lexical and semantic domains within vocabulary such as kinship or

numeral systems, in which reconstruction of certain members of the system may enable additional

reconstruction of less well-attested or even missing cognate sets within the same system.

b Possible semantic reconstruction of cells in a structured matrix even if lexical material is lacking.

iii Reconstruction of morphology to the extent that morphological reconstruction is merely an

extension of phonological and lexical reconstruction:

a Paradigmaticity may materially aid in reconstruction where cognate morphemes are poorly attested.

iv Reconstruction of syntax .

5.1 Cognate searches

In order to undertake any comparison at all one must have something to compare. The search for cognate

vocabulary is, oddly enough, usually the single most challenging task facing the comparativist. If the linguist

has already established the existence of a genetic relationship between two or more languages (see

Campbell, this volume), she or he has already located a certain number of important cognates. These are

normally searched for among the most basic of inflectional forms and among the most basic vocabulary

items. A list of 100 or 200 basic words is often used initially in cognate searches, the idea being that basic

concepts are the least likely to have been borrowed. We have learned that any such list should be used with

care, however, and then only after careful attention to known areal phenomena in the zone where one is

working. In English around 10 percent of such basic vocabulary is borrowed, mostly from French. In East and

Southeast Asia, though, it is well known that even the most basic numerals are often borrowed from

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g97814051274793

Page 3 of 20

1. The Comparative Method : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:27 PM

Southeast Asia, though, it is well known that even the most basic numerals are often borrowed from

Chinese. In table 1.1 , note that the first four languages are related, while the last three are not. Such known

vulnerabilities should obviously be considered and avoided, something that was often not possible a century

ago but which is often possible today. Atypical syllable structures, clusters, and marginal phonemes are

obviously suspect also.

Table 1.1 Basic numerals in East Asian languages illustrating both cognates and loanwords

Regularly corresponding phonemes in basic vocabulary and in basic grammatical formants (if typology

permits, preferably in paradigms) are the goal. The affixal morphology searched should be largely

inflectional, as derivational morphology is borrowed relatively easily and can wait until basic regularities have

been worked out.

5.2 Phonological reconstruction: comparanda

The question of comparanda in phonological reconstruction is important and is one of the most

underdiscussed questions in the literature: one obviously must know what to compare at all levels. The

degree of abstraction of the comparanda used in phonological reconstruction is significant and can have

important implications, both for relative ease of application of the comparative method, and for the accuracy

of reconstructions. Technically one could compare transcriptions of virtually any degree of abstractness from

a tight phonetic notation that reveals the greatest degree of lectal and individual variability to a highly

abstract underlying and underspecified phonological representation in which only the non-predictable

features are noted. There are good reasons to choose neither of these extreme alternatives, however.

It is not the primary job of the comparativist to document superficial dialect variation, and subphonemic

variability should usually be factored out of transcriptions used for comparison (although it can be very

valuable in charting sound change trajectories). Variable dialect data turn out to be much less variable if

they are first phonemicized. 14 Thus, even though the comparative method is in principle capable of dealing

with any number of variant forms, it is simpler to introduce a degree of abstraction that eliminates as many

as possible without compromising necessary distinctions. Degree of phonological abstraction then becomes

a question the comparativist must address.

The usual way in which the number of comparanda is reduced is to perform a preliminary internal

reconstruction on the data of each of the languages to be compared before attempting to use the

comparative method. This reduces (or eliminates) allomorphy and makes further comparison simpler.

Phonemicization is an obvious first step in such reduction.

Changes in synchronic phonological theory since about 1960 have clouded the picture somewhat. Only two

levels of notation have been significant in most generative phonologies, the underlying phonological and the

surface phonetic. We have already eliminated the phonetic as excessively detailed, but the underlying turns

out to be unsuitable for comparisons also. 15 This is because the procedures generally used for arriving at

synchronic underlying notation, although they often do lead to results that look superficially like

reconstructions, can sometimes lead the analyst in an ahistorical direction. The resultant abstract phoneme

may look like the results of an internal reconstruction, but internally reconstructed and merely abstract

phonemes can differ.

Numerous authors have noted the similarity between the procedures of internal reconstruction and those

used for abstracting underlying segments. It is often claimed that the procedures are really the same (e.g.,

Fox 1995: 210). Both procedures do involve treating allomorphs as cognates (which, internally, they are), but

synchronic phonological theory places a high value on productivity , which may in turn be the result of

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g97814051274793

Page 4 of 20

1. The Comparative Method : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:27 PM

analogical change, whereas internal reconstruction stresses the importance of irregularities, often so rare

that synchronic phonologies would merely assign them an exception feature of some kind. The least

productive and most irregular alternations are often the most revealing for the comparative linguist, but the

most productive and least irregular alternations are the ones that best serve the synchronist. So the two

methodologies may lead in different directions and should be kept distinct.

So it would seem that the comparativist must begin with something not far removed from the conservative

notion of surface phonemes, and that abstraction beyond cover symbols for the most automatic of

alternations must be treated as an avowedly historical procedure and justified by a careful and explicit

application of internal reconstruction. 16 The use of some variety of surface phonemes as comparanda at

once eliminates the most superficial levels of lectal variation while preventing a confusion of internally

reconstructed with merely underlying forms.

5.3 Correspondence sets and phonological reconstruction

Phonological and lexical reconstruction proceeds according to the procedures outlined above. Take, for

example, the cognate sets from several Siouan languages shown in table 1.2 . 17 The sets of stop

correspondences that can be extracted from these are shown in table 1.3 . Major subgroups here are

separated by a solid line and minor subgroups within the central Mississippi Valley subgroup by a broken

line.

The comparative method requires that these sets recur regularly in a great many other basic Siouan words.

With that requirement fulfilled, we see a pattern emerging among the correspondence sets (in spite of the

fact that some of the sets here are incomplete because cognates have not been found in some subgroups).

There are two sets of labial stops, two sets of dentals (we shall return to sets VII and VIII momentarily), and

two sets of velars. And where they differ, they seem to differ by a feature of aspiration or gemination. If we

assume that the gemination is secondary and comes from total assimilation of the h portion of the stop to

what it is adjacent to (i.e., hC > CC in the Dhegihan subgroup), then it appears probable that we should

reconstruct an aspirated and a plain (non-aspirated) set of stops for each of the three places of articulation.

To do this, however, we must answer several questions. Were the Proto-Siouan aspirates pre-aspirated, hC ,

or post-aspirated, Ch? Were the plain stops voiced or voiceless? What kind(s) of general evidence should we

look for and consult in answering these questions?

Table 1.2 Cognate sets from Siouan languages

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g97814051274793

Page 5 of 20

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: