13. Morphologization from Syntax.pdf

13. Morphologization from Syntax : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

13. Morphologization from Syntax

BRIAN D. JOSEPH

Subject

Linguistics » Historical Linguistics

Key-Topics

morphology , syntax

DOI:

10.1111/b.9781405127479.2004.00015.x

It is clear that the set of changes effected by speakers in their languages include those that are often

labelled “grammaticalization,” “grammaticization,” or even “grammatization.” This notion is variously

defined, 1 but especially in recent years, almost always in such a way as to refer to something that, first of

all, morphemes do, as opposed to (referring to) what is done by speakers, and that, second, echoes the

characterization of Kurylowicz (1965: 52): “an increase of the range of a morpheme advancing from a lexical

to a grammatical or from a less grammatical to a more grammatical status.” Indeed, several other chapters in

this volume - those by Bybee, Fortson, Harrison, Heine, Hock, Mithun, Rankin, and Traugott, to be exact -

are concerned, to one degree or another, with grammaticalization.

As Heine's chapter points out, the notion of “grammaticalization” has been extended by many practitioners

to cover other sorts of change than strictly the movement of an item along a scale (“cline”) of increasing

grammatical status (from content word > grammatical word > clitic > inflectional affix, cf. Hopper and

Traugott 1993: 7), and thus Kuryłowicz's definition is probably too narrow. McMahon (1994a: 160), for

instance, notes that grammaticalization encompasses essentially all types of language change, since

“grammaticalization is not only a syntactic change, but a global change affecting also the morphology,

phonology and semantics.” 2 Still, Kurylowicz's definition is generally accepted as a basic characterization of

grammaticalization, and it is so endorsed by Heine (this volume).

In the present chapter, by contrast, a rather different angle on the emergence of grammatical elements and

related phenomena is taken. In particular, the focus here is on what can be called “morphologization,” in a

particular sense - a set of developments by which some element or elements in a language that are not a

matter of morphology at one stage come to reside in a morphological component - or at least to become

morphological in type 3 - at a later stage. 4 For example, within Romance linguistics 5 it is generally agreed

that the French adverb-forming suffix -ment , as in clairement ‘clearly’ (cf. clair ‘clear/MASC.SG’), is a reflex

of the ablative case of the Latin feminine noun ment - ‘mind’ (nominative singular mens ) as used in adjective

+ noun phrasal combinations serving as adverbials, for example, clarã mente ‘with a clear mind’ (where

clarã is an ablative singular feminine form agreeing with the noun it is modifying); a reanalysis and/or shift

in phrasal status to word-level status seems to have occurred, resulting in monolectal forms in French such

as clairement . 6 Thus what was once in Latin a matter of syntax, that is, a combination of free words forming

a noun phrase that was case marked so as to function adverbially, became in French a matter of

morphology, that is, the output or result of word-formation processes that yield a derived word. But this

case is also a stock example of grammaticalization (see Hopper and Traugott 1993: 130–1), so some

differentiation between grammaticalization and morphologization is needed in order to show their

distinctness (as in Gaeta 1998, for instance).

Thus, the goal of this chapter is to discuss various aspects of morphologization, especially in comparison

with the by-now more familiar notion of grammaticalization, and to present an extended case study

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747915

Page 1 of 15

13. Morphologization from Syntax : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

with the by-now more familiar notion of grammaticalization, and to present an extended case study

examining one example in some detail. The case in point is the change in Medieval and Modern Greek by

which earlier speakers’ use of a periphrastic (i.e., multi-word and thus syntactic) future-marking formation,

consisting of the verb thélõ and a complement verb, yielded to later speakers’ use of a monolectal future in

the modern language -one with an apparently prefixal marker [θa-] attached to an inflected verb form.

Meillet (1912) wrote about this case as a paradigm example of grammaticalization, and it has been

mentioned elsewhere in the grammaticalization literature, as well (e.g., Hopper and Traugott 1993: 24;

McMahon 1994a: 167). 7

1 Scope and Motivation for Two Types of Morphologization

There are two directions for morphologization: either something that is syntactic at one stage can turn into

morphology (the major focus of this chapter), or something that is a matter of phonology at one stage can

become morphological (as discussed by Janda, this volume). These directions could be characterized as

morphologization from above and morphologization from below, respectively, reflecting the customary view

of the components of grammar as hierarchically arranged from the level of sound “up to” the level of

sentence structure, though nothing crucial hinges on this characterization.

Elsewhere, in Joseph and Janda (1988), these two types of morphologization have been referred to as

desyntacticizing and dephonologizing , respectively. Although they can be given these different labels, they

are actually quite similar, having the same outcome, that is, morphology, and the same motivation.

In particular, both reflect a preference on the part of speakers for what Joseph and Janda refer to as

“localized” solutions to the problem of how to account for a given phenomenon in language, for example,

marking for some category or a particular combination of elements. “Localized” solutions range over small

sets of data rather than being widely applicable, and are general only in a very local sense, covering perhaps

just a few forms. Reduplication in Sanskrit provides examples of such local generalizations, since the

patterns of reduplication found, for instance, in the perfect tense formations, including (where V = a vowel

that usually copies the root vocalism) V-, VV-, CV-, and occasionally even CVV-, as well as the highly

specific ãn -, tend to cluster around particular root types, such as V- with roots that begin with v-, ãn - with

certain vowel-initial roots, CV- with alteration of the root-initial consonant with a handful of roots, CV- as

the default case, and so on. Significantly, also, local generalizations tend to result from, and show

fragmentation of, once quite general phenomena - perfect tense reduplication in Proto-Indo-European, for

instance, was almost exclusively 8 CV - and thus they suggest that speakers focus on the analysis of just a

restricted set of data at a time, and therefore come up with quite particularized analyses. 9 That is to say,

speakers view language through a relatively small “window” at any given time, and thus the size of their

focal area is relatively small. 10 This access to only limited data at a given time translates into solutions that

are cast in terms of highly particular properties of stems, affixes, and the like, and which are usually best

accommodated in the morphology, since phonological solutions are usually to be interpreted quite generally,

referring as they do to properties of sound only; thus local generalizations, being defined often in terms of

idiosyncrasies and sometimes extending only over a few forms, are usually morphological in nature, as well

as quite concrete, in that they are based on surface representations and categories that are overtly marked

rather than on abstract properties of phonological form.

In that way, Joseph and Janda claimed, speakers opt for morphological accounts over phonological or

syntactic ones whenever the analysis of a given phenomenon offers any ambiguity as to the extent of its

generality. It was further argued there that grammars should therefore be viewed as being “morphocentric,”

or more accurately, “morpholexicocentric” (see below), with a greater role for the morphological component,

in order to explain this preference speakers give to morphological accounts.

It must be realized that the lexicon is taken here to be connected closely with the morphological component,

and thus is part of what is to be considered “the morphology” of a language. The lexicon, after all, is where

(at least root) morphemes are found and where (at least) idiosyncratic information about morphemes

resides. 11 Thus references to “morphological” phenomena here include “morpholexical” information as well.

Morphology, after all, is concerned with form and the relation of form to meaning in most traditional

views, 12 so any aspect of language that is concerned with form, as the lexicon must necessarily be, is fair

game for being subsumed under - or at least tightly allied with - morphology. Moreover, in many

frameworks, even elements with internal syntax are listed in the lexicon, for example, adjective-noun

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747915

Page 2 of 15

13. Morphologization from Syntax : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

frameworks, even elements with internal syntax are listed in the lexicon, for example, adjective-noun

combinations with specialized meanings such as Cold War , 13 a move that is in keeping with the expanded

view of the role of morphology and the morphological/morpholexical component in language implicit in the

notion of “morphocentricity.” 14

The scope of morphology can thus be quite large, and consequently there is a wide range of phenomena

that can be said to show morphologization, that is, movement into the morphology, assuming of course that

one can devise a heuristic for determining when the boundaries have been crossed (see below, section 3).

2 Morphologization and Grammaticalization Distinguished

As noted at the beginning of this chapter, there is some overlap between the notion of morphologization as

developed here (drawing on Joseph and Janda 1988) and that of grammaticalization, discussed in this volume

and elsewhere. Yet, there are significant differences of approach, method, and substance between the two

that provide a rationale for taking a morphologization viewpoint on various changes and not simply treating

them as instances of grammaticalization. 15

For one thing, there are phenomena in language which are (already) clearly grammatical in their function but

which nonetheless undergo changes in the direction of greater involvement in the morphological component.

For instance, the changes to be discussed concerning the Greek future started with a grammatical usage of

the verb thélõ , which meant ‘want’ as an ordinary lexical verb, in a periphrasis indicating futurity; as

becomes clear below, these changes were such that the realization of the marking for futurity passed from

being a matter of syntax (i.e., word combination) to being a matter of morphology (i.e., word formation).

Thus, there is clearly morphologization in this example by the definition given above, but is there

grammaticalization? There might be, but only if grammaticalization is taken to involve movement along a

“cline” by which expression via morphology, for example with an affix, is “more grammatical” than

expression via syntax (cf. Kurylowicz's definition, given above). However, such a cline is completely

stipulative, for there are free words that have grammatical functions, such as English of or French de ,

various complementizers such as English that and whether or French à , pronouns, etc., as well as affixes

that have no grammatical function at all, such as the empty -al that (descriptively speaking) can be added

for some English speakers to syntactic to form syntactical (note that both are adjectives and that they are

synonymous) or the equally empty -y that (again from a descriptive standpoint) some English speakers add

to competence to give competency , and so on. Thus there is no necessary correlation between an item's

place on the cline and its degree of grammatical involvement. Grammaticalization theorists recognize this

issue to some extent; Lehmann (1985: 306), for instance, gives six criteria - attrition, condensation,

paradigmaticization, coalescence, obligatorification, and fixation 16 - and claims that an item lines up at

equivalent points with regard to each one as it “grammaticalizes.” However, each of these properties is in

principle independent of the others, so that demanding a grouping of all of them involves a stipulation that

one needs to have all six, and in equal measures, to have movement along the grammaticalization cline.

Similarly, as noted in section 1, there are two directions of development that can lead to morphologization,

and desyntacticizing morphologization can readily be linked to dephonologizing morphologization via their

common outcome (morphology) and common motivation (localized generalization by speakers). When viewed

from within a grammaticalization framework, however, it is not at all obvious why

morphological/morpholexical determination for a given phenomenon, as opposed to determination via

regular and general phonological conditions, should be considered to be more grammatical and thus should

have anything to do with or anything in common with, for instance, the movement from syntactically

determined to morphologically/morpholexically determined. For example, marking noun plurals via an affix

that happens to have a regular, exceptionless, phonological effect on a root, such that it would be accounted

for by a purely phonological rule, does not seem a priori to be less grammatical in any meaningful sense

than marking plural via an affix that alters the vocalism of a root in ways that vary from one lexical item or

lexical class to another and thus requires a morpholexically particularistic account; 17 nonetheless,

grammaticalization “theory” wants to link such a change in the nature of the concomitant phonological

effects with the change from phrasal to affixal expression of adverbials or futures or the like as being the

same type of change. 18 Such a linkage is straightforward when viewed from the perspective of

morphologization, since in both cases there is greater involvement of the morphology, but not necessarily so

at all from a grammaticalization perspective.

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747915

Page 3 of 15

13. Morphologization from Syntax : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

at all from a grammaticalization perspective.

Moreover, as noted above, grammaticalization proponents recently have been claiming an ever broader

domain of applicability for this notion, whereas such is not the case with morphologization. Yet there are

changes in language and grammar that do not involve any of the typical characteristics of

grammaticalization. Regular sound change, for instance, under the Neogrammarian view (see Hale, this

volume), is purely phonetically conditioned and almost by definition has no grammatical involvement at all.

Also, a change such as the polarization in word order by which speakers of English have come to

differentiate the perfect I have written the letter from the resultative I have the letter written , moving

away from earlier English variability in ordering for both types, 19 seems not to involve any cluster of the

characteristics said to be typical of grammaticalization. 20 Admittedly, these changes do not involve

morphologization either, but the concept of morphologization makes no claims about such changes, whereas

grammaticalization, in some instantiations at least, does.

Similarly, there are changes in the direction of greater morpholexical involvement that do not involve

grammar, and thus can be accommodated within the concept of morphologization but not

grammaticalization. Relevant here are the sorts of reductions seen, for instance, in German heute ‘today’

from a presumed instrumental phrase hiu tagu* in Old High German, or heuer ‘this year’ from the OHG

instrumental phrase hiu jaru . It is not clear that anything relevant to grammaticalization has taken place, for

this combination of sounds is as grammatical (or not, as the case may be) before the phrase was reduced as

it is afterwards. Yet, as Hopper and Traugott (1993: 23) note regarding heute , “there surely is a difference in

Modern German between heute and an diesem Tage ‘on this day’ that needs to be characterized in some

way”; grammaticalization really does not provide a way, yet morphologization is exactly what is involved

here.

In fact, the only way that hiu tagu > heute might be said to be relevant to grammaticalization is under the

interpretation Hopper (1994) makes concerning what he calls “phonogenesis,” defined by him as “the

process whereby new syntagmatic phonological segments are created out of old morphemes” (p. 31). He

explicitly refers to phonogenesis as “an advanced stage of grammaticalization” (ibid.), noting that there is

generally “phonological reduction that accompanies grammaticalization” (ibid., and see the reference above

in section 2 to Lehmann's “attrition” and “coalescence”). While there is no denying that such developments

occur - and indeed, Hopper presents a large number of well-known and not-so-well-known cases, mostly

from English, by way of illustrating the phenomenon, such as the -nd of friend and fiend reflecting an old

present particpial ending now lacking in any obvious meaning - the terminology and definition seem

particularly inappropriate and the linkage with grammaticalization is at best fortuitous.

For one thing, there is nothing grammatical about such material; if anything, what -nd has undergone might

be termed “degrammaticalization,” at least by the usual definitions of grammaticalization, for there is a

movement out of being, in some sense, a grammatical formative. Admittedly, this criticism may involve

taking the terminology of grammaticalization too much at face value, but given generally accepted

formulations of grammaticalization, an extension of the notion is needed if “phonogenesis” is to be

subsumed under the same rubric as the development of the French adverbial marker -ment . Furthermore,

calling the accretion of material onto a word “phonogenesis” implies that the material added, the element

that was once a morpheme in Hopper's formulation, 21 had no phonic value when it was a morpheme.

However, whether -nd was a recognizable participial suffix or a meaningless string at the end of friend and

fiend , it still contained a sequence of sounds; the morphemic or non-morphemic status of that sequence

does not affect the extent to which this element adds “phonological bulk” (in Hopper's words, p. 29) to a

stem it attaches to. Thus there may be “phono-accretion,” but the sounds were already there and thus had

already undergone “genesis” at the time they constituted a morpheme; the real difference lies in the

morphological status of the sequences in question, that is, phrasal versus word status, or

compound/polymorphemic word versus monomorphemic word status. Such a difference can be readily

characterized in terms of morphologization, but not grammaticalization, and moreover, focusing on

morphological status allows such cases to be linked rather directly with the emergence of the French

adverbial suffix and similar examples in ways that grammaticalization theory can only do by stipulation and

extension of the basic notion. 22

Finally, there are methodological differences between the ways in which morphologization has been studied

and the ways in which grammaticalization has been studied. In particular, grammaticalization has now been

built into an elaborate theoretical framework, so-called “grammaticalization theory,” with a cognitive basis

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747915

Page 4 of 15

13. Morphologization from Syntax : The Handbook of Historical Linguistics : Blackwell Reference Online

12/11/2007 03:36 PM

built into an elaborate theoretical framework, so-called “grammaticalization theory,” with a cognitive basis

and a stake in the putative principle of unidirectionality, by which, it is claimed, changes are always in the

direction of greater grammatical status and not, for instance, in the direction of affix to clitic or free word,

that is, in the opposite direction on the cline of grammaticalization. The existence of such “counter-

grammaticalizations,” described in the literature for several years and recently discussed and summed up,

with extensive literature, in Janda (2001), are particularly troublesome for grammaticalization proponents, 23

but pose no threat for the concept of morphologization, as discussed below in section 5.

Also, grammaticalization studies tend to ignore the somewhat more formal question of where in the

grammar a particular phenomenon is to be located, as if it is always self-evident what the answer to this

question is. Some studies do provide a basis for making a decision, for example Hopper and Traugott (1993:

4–6) regarding the distinction between clitics versus affixes, etc., but many do not, nor do all that recognize

criteria apply them in all cases.

Thus grammaticalization and morphologization indeed offer distinct perspectives on, and represent distinct

ways of viewing, changes that involve grammatical machinery and morphological/morpholexical material. 24

3 How to Tell

As suggested earlier, talking about morphologization implies that there is a way to tell whether some

phenomenon is “in the morphology” or, as is relevant for desyntacticization, “in the syntax” instead. The

most useful heuristics are those enumerated in Zwicky and Pullum (1983) and Zwicky (1985a).

They distinguish among affix, clitic, and word as types of morphological elements, drawing important

distinctions between affix and non-affix and between word and non-word. “Clitics,” then, are elements that

are neither canonical affixes nor canonical words. 25 They further identify a number of traits that are

characteristic of affixes and characteristic of words. For the most part, affixes, as morphological elements,

show various types of idiosyncrasy - they are selective as to what they attach to, they can provoke irregular

effects on the stems they occur with, their ordering is generally fixed, they tend to be prosodically

dependent, they are not subject to syntactic rules (e.g., deletions) unless the whole word they are part of is

affected, 26 and the like - while words, as syntactic elements, show a greater degree of generality, being

unselective in their combinatory possibilities, allowing reordering in response to stylistic factors, having

prosodic independence, showing a one-to-one mapping with semantic rules that give syntactic units

semantic compositionality, etc.

There are other criteria that can be helpful. For example, in the case of the Oscan locative, agreement seems

to solve the issue of what sort of analysis is warranted. Oscan innovated a locative by the agglutination of a

postposition en onto a noun, for example húrtín ‘in the garden’ (Buck 1928: 114), yet what shows that this

is indeed a morphological marker of a case, as opposed to a combination of free words in a noun phrase

that undergo some phonological adjustment, is the fact that the -ín ending occurs on adjectives in

agreement with a locative noun marked in the same way. Thus, this new locative participates in adjective

agreement just like other cases, a feature which shows that the appearance of the -ín is not from a

synchronic merging of a free word onto a stem; if it were a syntactically generated postpositional word, one

would not expect it to occur both on the adjective and on the noun, unless, due to the principle of

compositionality, there were a corresponding semantic contribution from both occurrences.

Still unresolved, admittedly, is the issue of whether compounds are syntax or morphology. The case of

Romance adverbial mente , once again, is instructive. Unlike French, where -ment seems to be an affix (note

that it is bound and provokes an idiosyncratic selection of the adjective stem it is added to 27 ), Spanish

adverbial mente , in certain registers at least, can apply distributively over both adjectives in a conjoined

phrase (apparently contrary to the lexical integrity principle - see n. 26), for example rapida y claramente

‘rapidly and clearly’ (not: ‘*rapid and clearly’) and -mente adverbs can have two accents (thus rápidaménte ).

Moreover, mente survives in Spanish as a free noun meaning ‘mind,’ though it is not at all clear that there is

a synchronic connection between mente ‘mind’ and the adverbial formative. These facts suggest an analysis

whereby -mente adverbs in modern Spanish are compounds, perhaps containing -mente as a bound root. If

that is the case, then the developments with -mente in Spanish would not represent a case of

morphologization, unless compounds are taken to be a matter of morphology (word formation) rather than

of syntax. 28

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747915

Page 5 of 15

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: