11. Analogical Change.pdf

11. Analogical Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

11. Analogical Change

HANS HENRICH HOCK

Subject

Linguistics » Historical Linguistics

DOI:

10.1111/b.9781405127479.2004.00013.x

Since the time of the Neogrammarians, analogy has generally been accepted as a major force in linguistic

change. However, opinions have varied as regards the definition of analogy and analogical change, its

relation to morphological change and to sound change, and the question of whether there are natural

tendencies in analogical change. This chapter attempts to present a critical overview of the major

perspectives and, toward the end, to reconcile some of the contradictions in these perspectives by way of a

hypothesis which views sound change and analogical (and semantic) change as points on a continuum of

changes that may be considered analogical in a larger sense. 1

To facilitate the following discussion I start with a brief presentation of some of the major phenomena which

have been considered analogical. (For further details see Hock 1986, with updates in Hock and Joseph 1996.)

Four-part analogy operates on the basis of a proportional model of the type (1) and generalizes a pattern of

morphological relationship between given forms to other forms which previously did not exhibit this pattern,

as in example (2). Certain conditions increase the success of this type of analogical change. These include

the fact that the “x-“side of the equation should be a synchronically derived form (such as “plural” versus

“singular”) and that the pattern being generalized should be productive. 2 (The term backformation is used in

reference to the much rarer cases in which the equation is “solved” on the side of the synchronically basic

form.)

(1)

(2)

Leveling eliminates (morphophonemic) alternations within paradigms, as in (3). The process is most

successful if the alternations do not signal important morphological distinctions. For instance, in (3) the

vowel alternations in the past tense forms are eliminated, but a difference remains between the present and

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747913

Page 1 of 15

11. Analogical Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

vowel alternations in the past tense forms are eliminated, but a difference remains between the present and

past tense vowels, since the present: past distinction is a relatively important one in English:

(3)

Morphophonemic extension 3 is a rarer alternative to leveling. An example is the British English “intrusive r”

which is inserted between a word-final (non-diphthongal) vowel and a following word-initial vowel and

which traditionally is motivated in terms of a proportion of the type (4):

(4)

Blending telescopes the meanings or functions, as well as the phonetic forms, of two structures into a single

form, as in Lewis Carroll's chortle = chuckle and snort .

Contamination refers to changes in which a particular form influences the pronunciation of a semantically

related form, without changing the meaning of the latter. Examples are especially common in antonyms and

numerals. See, for instance, (5):

(5)

Recomposition and folk etymology are two related processes that assign transparent compound structure to

words; in the former case, this is the historically correct structure (6a), in the latter case it is not (6b):

(6)

Of these processes, the first three tend to apply with much greater systematicity than the remainder.

However, as is well known, the Neogrammarians regarded all analogical change as irregular, in contrast to

sound change which, with certain well-defined exceptions, was considered to be absolutely regular.

Note further that although all of the processes have been considered analogical in much of the literature,

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747913

Page 2 of 15

11. Analogical Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:35 PM

this does not mean that all historical linguists subscribe to that view. At least in part, differences of opinion

reflect the history of the notion “analogy” before it was adopted (or adapted) in historical linguistics.

1 The History and Definition of the Term “Analogy”

As is well known, the term “analogy” goes back to Ancient Greek philosophy and grammar. 4 Even the Ancient

Greek and Roman tradition, however, varied to some degree in the definition of analogy, either as regular

inflection or as proportion (see Best 1973: 16 with references); but Best may be correct in arguing that the

original reference is to proportion.

The use of the term analogy to designate regular inflection (or paradigm regularity) and/or proportion

continues into medieval and early modern times. 5 In addition, however, it also tends to be used in reference

to any observed regularity, including regularity in sound correspondences, as in the following passage from

Schlegel (1808: 6): 6

In this regard we permit no kinds of rules that change or transpose the letters, but demand

complete identity of the word for proof of descent. True, if the intermediate links can be

established historically, then it is possible to derive [It.] giorno from [Lat.] dies [‘day’]; and if

instead of the Latin f we often find h in Spanish, if Latin p very frequently becomes f in the

German form of the same word, and [Lat.] c not infrequently h , then this does establish an

analogy also for other not quite so obvious cases. However, as stated, one must be able to

establish the intermediate links or the general analogy historically; nothing may be invented on

a priori grounds, and the agreement must be very precise and evident to permit even minute

formal variations.

(my translation and emphases)

Given this understanding of analogy, combined with the Romanticist notion of a perfect proto-language, it is

understandable that analogical developments of the sort outlined at the beginning of this chapter generally

were considered instances of “false analogy,” the transfer of a linguistic form from the original pattern - or

analogy - to one that is historically incorrect and therefore a feature of late “decaying” languages. 7

(Curiously, the term “false analogy” is much easier to find in the Neogrammarians’ attacks on earlier

linguists than in the work of these linguists; but see Bhandarkar 1877–8: 14 and passim for genuine uses of

the term.)

The Neogrammarians generally credit Scherer (1868) and Leskien (1876) with laying the foundation for a

proper assessment of analogy as a perfectly normal type of change which can take place at any stage in

history, including in the proto-language; see Osthoff and Brugmann (1878). Koerner has shown recently

(1983b) that a very similar view can be found in the work of Schleicher. 8

Interestingly, the Neogrammarian definition of analogy entailed a significant change in the meaning of the

term - instead of referring to synchronic regularity, it now was used to designate a historical phenomenon

which was considered inherently irregular, in contrast to the “absolute regularity” of sound change.

Nevertheless, for some of the Neogrammarians analogy retained some of its traditional meaning in that its

use was limited to proportional analogy (what is here called four-part analogy and morphophonemic

extension), while other phenomena, such as leveling, are expressly excluded. See Paul (1880: 106–20, 189–

216, and esp. 161 n. 1). 9,10

Paul's approach was accepted by Sturtevant (1947: 97), who, however, includes cases where leveling and

four-part analogy can be said to cooperate and lead to fairly sweeping results (the type Latin honõs,

honõrem → honor, honõrem , for which see Hock 1986: 180). 11 Similarly, Saussure (1916: ch. VI)

characterizes folk etymology as different from (proportional) analogy; Jeffers and Lehiste (1979: 68) restrict

the term “analogy” to proportional analogy; Miranda (1974) argues for a strict distinction between

proportional analogy and leveling; and Kuryłowicz's “laws” work best for proportional analogy, while leveling

is better accounted for by Mańczak (see section 2 below).

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747913

Page 3 of 15

11. Analogical Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

At the same time, as acknowledged by Paul (1880: 161 n. 1), many other linguists operate with a much less

restrictive definition of analogy. For instance, Brugmann (1906: 16–18) includes proportional analogy,

leveling, contamination, and backformation under the heading “Analogiewirkung, Neuschöpfung und

Umbildung”; Wheeler's (1887) classification of analogy embraces blending, contamination, and folk

etymology, in addition to proportional analogy and leveling; and Hermann (1931) vigorously argues against

limiting analogy to proportional phenomena and emphasizes, among others, the factors of contiguity within

the sentence and of phonetic/semantic associations (as in folk etymology). See also Bloomfield (1914: 221–

37); Buck (1933: 45–7).

In fact, most recent publications adopt such a more inclusive definition; see, for example, Anttila (1972: 88–

108 ≈ 1989: 82–94); Hock (1986: ch. 9); Lehmann (1962: 177–92, 1973: 189–92, 1992: 219–34); McMahon

(1994a: 70–91); 12 Trask (1996: 105–12).

Bloomfield (1933: 404–24) goes even farther by including the semantic re-interpretation of meat (‘food’ →

‘meat’) under the heading of analogy, even though the change has no phonetic or phonological effect.

Something similar is found in Boretzky's treatment of analogy (1977: 129–42). While on pp. 134–5 he

operates with a relatively narrow definition of analogy which excludes folk etymology and blending, on pp.

138–42 he envisages a broader definition, which is not limited to phonetic results and includes even

semantic extension. Anttila's definition of analogy in this volume is even broader and certainly includes

metaphorical extension.

Even if semantic changes of this type, which have no phonetic effects, are excluded, it may be legitimately

asked whether we should not include changes in overt form that are purely semantically driven, such as

taboo distortion (as in Mod. Engl. doggone or shoot! ) and onomatopoeic (re)creation (as in Mod. Engl. chirp

or cheep , which in effect have replaced pipe [payp], after the Great English Vowel Shift disrupted the iconic

relationship between earlier pipen and the sound depicted by the word).

Put differently, the question is this: there seems to be a continuum, ranging from the more formally

motivated proportional analogy, leveling, and morpho-phonemic extension, via more semantically driven

blending, contamination, and folk etymology, to the even more clearly semantically motivated taboo

distortion and onomatopoeic (re)creation which have some formal repercussions, to semantic shifts without

such formal repercussions (as in the case of meat ). Where, in this continuum, should we draw the dividing

line, on what grounds should we draw it, and should we draw one at all?

2 Tendencies in Analogical Change

Perhaps the most widely accepted tendency of analogical change is the notion that leveling serves to

establish the principle of “one meaning, one form” and to eliminate variation that does not serve a

morphological purpose; see, for example, Anttila (1972: 107 = 1989: 107, with references to earlier

literature); Hock (1986: 168). The tendency has been named Humboldt's Universal by Vennemann (1969: §

2.23, 1972: 183–5).

Lass (1997: 342–52) claims that the “universal” is empirically untestable “and therefore uninformative” (p.

344). However, his assessment seems overly pessimistic, for it would a priori be possible to falsify the

principle by showing that analogical change more commonly introduces than eliminates variation that has no

morphological purpose. Significantly, this does not seem to be the case.

A more general discussion of the tendencies of analogical change originates in the disagreement between

Kuryłowicz (1945–9, see also 1964b, 1968) and Mańczak (1958, 1966, 1978, 1980). The scholars disagree

in their basic approach (introspective and morphology-oriented for Kuryłowicz, empirical-statistical and

more phonology-oriented for Mańczak), as well as in specific claims. In the following I focus on what I

consider the major issues in this disagreement; for fuller discussion see Best (1973: 61–107); Anttila (1977:

76–80); Collinge (1985: 249–53); Hock (1986: ch. 10); 13 Winters (1995); and the references in these

publications.

The most notable conflict is found between Kuryłowicz's first law of analogy and Mańczak's second

tendency. Pointing to developments of the type (7), Kuryłowicz claims that bipartite markers, such as the

plural marker -∂ plus umlaut, tend to replace simple markers that have the same function, such as the

plural marker -∂ without umlaut. Mańczak, by contrast, states that root alternations are more often

abolished than introduced - the essence of what Vennemann has called Humboldt’ s Universal:

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747913

Page 4 of 15

11. Analogical Change : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

abolished than introduced - the essence of what Vennemann has called Humboldt’ s Universal:

(7)

Following Hock (1986: 235–6) we may resolve this conflict by noting that the two approaches address

different phenomena. The change in (7) involves four-part analogy and is motivated by the fact that the

alternation is serving a morphological “purpose” - of more clearly marking plural forms. On both counts the

changes differ from the more common type (3) above, which involves leveling of alternations that do not

serve such a purpose. 14

While this perspective accounts for the fact that both types of change are possible, skeptics such as Lass

may raise the question of why clear plural marking is “morphologically useful” in German, but not in English,

which has tended to eliminate umlaut plurals (mainly through extension of the s -plural). There is no ready-

made answer to this question, but there is no ready-made answer, either, to the question of what makes a

particular type of formation productive (see n. 2) - even though the concept of productivity is clearly a valid

one, and plays a major role in analogical change.

Kuryłowicz's second law of analogy makes a claim about the general direction of analogical change, namely

that it proceeds from basic form to derived form. Moreover, it states that the relationship between these

forms “is a consequence of their spheres of usage.” As shown in detail by Hock (1986: 213–22), the “sphere

of usage” provision is especially significant. First, it incorporates the observation that productive patterns,

which presumably are used more freely, are more successfully extended in four-part analogy. Second, it

invites us to examine more closely the question of what is basic within particular formal categories and thus

provides a principled explanation for some of Mańczak's rather random observations, such as the claim that

geographic nouns tend to preserve locational cases better than other cases (Hock 1986: 232–4). A further

corollary of the sphere of usage provision is the often-claimed tendency for third persons to be more basic

in analogical change than other forms of the verb (Hock 1986: 220–2) - a tendency which has been called

Watkins's law (Arlotto 1972: 156; Joseph 1980a). 15

In spite of its significance, Kuryłowicz's second law clearly is not without problems. Most obviously, it is

contradicted by backformation, which violates the “basicness” provision of the law. At the same time, it is

well known that backformation is less systematic and successful than four-part analogy, and this behavior

may be considered to confirm the law at least as a tendency. More serious is the fact that the law holds as a

valid tendency only for proportional analogy and fails to make correct predictions for leveling 16 and most

other analogical developments (Hock 1986: 213–14).

Kuryłowicz's fourth law sums up the conventional insight of historical linguists that, when analogical change

results in doublets, such as innovated brothers versus archaic brethren , the newer form tends to take on the

productive function and the older form survives in specialized or otherwise marked usages.

The law's validity has been questioned by Lehmann (1973: 200, 1992: 232) and Kiparsky (1974a). Lehmann

notes that in German, the innovated genitive form of the definite article, dessen , has marked functions

compared to the older form, des , in violation of the law. Kiparsky points to examples such as (8), where

innovated forms are found in more marked, rather than less marked, structures:

(8)

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747913

Page 5 of 15

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: